diff --git a/.gitattributes b/.gitattributes index 0e6435e15683bea23764d48c57e8845c1f943112..340af6f4f109e70fee0dc9f37ff365630ce073c2 100644 --- a/.gitattributes +++ b/.gitattributes @@ -1383,3 +1383,11 @@ minigpt2/lib/python3.10/site-packages/more_itertools/__pycache__/more.cpython-31 parrot/lib/python3.10/site-packages/nvidia/cublas/lib/libnvblas.so.12 filter=lfs diff=lfs merge=lfs -text parrot/lib/python3.10/site-packages/numpy/ma/tests/__pycache__/test_core.cpython-310.pyc filter=lfs diff=lfs merge=lfs -text vllm/lib/python3.10/site-packages/pyarrow/_feather.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/libarrow_dataset.so.1801 filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/_parquet_encryption.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/_flight.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/_hdfs.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/_fs.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/libarrow_acero.so.1801 filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/_dataset_parquet_encryption.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text +vllm/lib/python3.10/site-packages/pyarrow/libparquet.so.1801 filter=lfs diff=lfs merge=lfs -text diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/__init__.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..a0eab9c9521644825141f0684c7ad5bac6176d1e --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/__init__.py @@ -0,0 +1,23 @@ +from .backend_config import BackendConfig, BackendPatternConfig, DTypeConfig, DTypeWithConstraints, ObservationType +from .fbgemm import get_fbgemm_backend_config +from .native import get_native_backend_config, get_native_backend_config_dict +from .qnnpack import get_qnnpack_backend_config +from .tensorrt import get_tensorrt_backend_config, get_tensorrt_backend_config_dict +from .executorch import get_executorch_backend_config +from .onednn import get_onednn_backend_config + +__all__ = [ + "get_fbgemm_backend_config", + "get_native_backend_config", + "get_native_backend_config_dict", + "get_qnnpack_backend_config", + "get_tensorrt_backend_config", + "get_tensorrt_backend_config_dict", + "get_executorch_backend_config", + "BackendConfig", + "BackendPatternConfig", + "DTypeConfig", + "DTypeWithConstraints", + "ObservationType", + "get_onednn_backend_config", +] diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/_common_operator_config_utils.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/_common_operator_config_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..4e946a25ffbbf003d39a020ea75fea185551ce46 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/_common_operator_config_utils.py @@ -0,0 +1,637 @@ +import copy +import operator +import torch +import torch.nn.functional as F +import torch.nn as nn +import torch.ao.nn.intrinsic as nni +import torch.ao.nn.intrinsic.qat as nniqat +import torch.ao.nn.qat as nnqat +import torch.ao.nn.quantized.reference as nnqr +from collections import namedtuple +from typing import Callable, Dict, List, Union +from .backend_config import ( + BackendPatternConfig, + DTypeConfig, + DTypeWithConstraints, + ObservationType, +) +from ..fuser_method_mappings import ( + _sequential_wrapper2, + fuse_conv_bn, + fuse_conv_bn_relu, + fuse_linear_bn, + fuse_convtranspose_bn, +) + +__all__: List[str] = [] + +# TODO: rename to be more explicit, e.g. qat_conv_relu +_ConvMetadata = namedtuple( + "_ConvMetadata", + ["root", "transpose", "bn", "reference", "transpose_reference", + "fused_conv_relu", "fused_conv_bn", "fused_conv_bn_relu", + "qat", "relu_qat", "bn_qat", "bn_relu_qat", + "func", "func_transpose"]) +_Conv1dMetadata = _ConvMetadata( + nn.Conv1d, nn.ConvTranspose1d, nn.BatchNorm1d, nnqr.Conv1d, nnqr.ConvTranspose1d, + nni.ConvReLU1d, nni.ConvBn1d, nni.ConvBnReLU1d, + nnqat.Conv1d, nniqat.ConvReLU1d, nniqat.ConvBn1d, nniqat.ConvBnReLU1d, + F.conv1d, F.conv_transpose1d) +_Conv2dMetadata = _ConvMetadata( + nn.Conv2d, nn.ConvTranspose2d, nn.BatchNorm2d, nnqr.Conv2d, nnqr.ConvTranspose2d, + nni.ConvReLU2d, nni.ConvBn2d, nni.ConvBnReLU2d, + nnqat.Conv2d, nniqat.ConvReLU2d, nniqat.ConvBn2d, nniqat.ConvBnReLU2d, + F.conv2d, F.conv_transpose2d) +_Conv3dMetadata = _ConvMetadata( + nn.Conv3d, nn.ConvTranspose3d, nn.BatchNorm3d, nnqr.Conv3d, nnqr.ConvTranspose3d, + nni.ConvReLU3d, nni.ConvBn3d, nni.ConvBnReLU3d, + nnqat.Conv3d, nniqat.ConvReLU3d, nniqat.ConvBn3d, nniqat.ConvBnReLU3d, + F.conv3d, F.conv_transpose3d) + +# Add constraints for fixed qparams ops like sigmoid and tanh to ensure values +# fall within the proper ranges, e.g. [0, 1] for sigmoid, [-1, 1] for tanh +_FIXED_QPARAM_OP_0TO1_CONSTRAINTS = DTypeWithConstraints( + dtype=torch.quint8, + quant_min_lower_bound=0, + quant_max_upper_bound=255, + scale_exact_match=1.0 / 256.0, + zero_point_exact_match=0, +) +_FIXED_QPARAM_OP_NEG1TO1_CONSTRAINTS = DTypeWithConstraints( + dtype=torch.quint8, + quant_min_lower_bound=0, + quant_max_upper_bound=255, + scale_exact_match=2.0 / 256.0, + zero_point_exact_match=128, +) +_FIXED_QPARAMS_OP_TO_CONSTRAINTS: Dict[Union[Callable, str], DTypeWithConstraints] = { + torch.nn.Hardsigmoid: _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + torch.nn.functional.hardsigmoid: _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + "hardsigmoid": _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + "hardsigmoid_": _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + torch.nn.Sigmoid: _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + torch.sigmoid: _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + "sigmoid": _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + "sigmoid_": _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + torch.nn.Softmax: _FIXED_QPARAM_OP_0TO1_CONSTRAINTS, + torch.nn.Tanh: _FIXED_QPARAM_OP_NEG1TO1_CONSTRAINTS, + torch.tanh: _FIXED_QPARAM_OP_NEG1TO1_CONSTRAINTS, + "tanh": _FIXED_QPARAM_OP_NEG1TO1_CONSTRAINTS, + "tanh_": _FIXED_QPARAM_OP_NEG1TO1_CONSTRAINTS, +} + +def _get_binary_op_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + binary_op_configs: List[BackendPatternConfig] = [] + num_tensor_args_to_observation_type_mapping = { + # TODO: this is not used right now since we have extra check in prepare + # will need to change this to NO_OBSERVER later after we implemented + # Tensor dtype inference properly + 0: ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT, + 1: ObservationType.OUTPUT_SHARE_OBSERVER_WITH_INPUT, + 2: ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT, + } + for op_with_quantized_bop_scalar_variant in [operator.add, torch.add, operator.mul, torch.mul]: + bop_patterns = [ + (op_with_quantized_bop_scalar_variant, nn.ReLU), + (op_with_quantized_bop_scalar_variant, F.relu), + (op_with_quantized_bop_scalar_variant, torch.relu), + op_with_quantized_bop_scalar_variant + ] + for bop_pattern in bop_patterns: + binary_op_configs.append( + BackendPatternConfig(bop_pattern) + .set_dtype_configs(dtype_configs) # noqa: E131 + ._set_num_tensor_args_to_observation_type(num_tensor_args_to_observation_type_mapping)) + # matmul + binary_op_configs.append( + BackendPatternConfig(torch.matmul) + .set_dtype_configs(dtype_configs) # noqa: E131 + ) + return binary_op_configs + +def _get_linear_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + """ + Return all configs related to linear modules and ops. + """ + observation_type = ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT + linear_configs: List[BackendPatternConfig] = [] + + # (1) Single linear modules/functions + # ------------------------------------- + # linear module + linear_configs.append( + BackendPatternConfig(torch.nn.Linear) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(torch.nn.Linear) + .set_reference_quantized_module(nnqr.Linear) + .set_qat_module(nnqat.Linear)) + # linear qat module + linear_configs.append( + BackendPatternConfig(nnqat.Linear) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(torch.nn.Linear) + .set_reference_quantized_module(nnqr.Linear)) + # functional linear + linear_configs.append( + BackendPatternConfig(torch.nn.functional.linear) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 1, "bias": 2})) + + # (2) Linear + relu + # ------------------- + # 2.1 linear module + relu fusion config + # linear relu, linear module + relu module + linear_configs.append( + BackendPatternConfig((torch.nn.Linear, torch.nn.ReLU)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(_sequential_wrapper2(nni.LinearReLU)) + .set_fused_module(nni.LinearReLU)) + # linear relu, linear module + functional relu + linear_configs.append( + BackendPatternConfig((torch.nn.Linear, torch.nn.functional.relu)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(_sequential_wrapper2(nni.LinearReLU)) + .set_fused_module(nni.LinearReLU)) + + # 2.2 linear module + relu, fused module configs + # linear relu, fused module + linear_configs.append( + BackendPatternConfig(nni.LinearReLU) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(torch.nn.Linear) + .set_reference_quantized_module(nnqr.Linear) + .set_qat_module(nniqat.LinearReLU)) + # linear relu, qat fused module + linear_configs.append( + BackendPatternConfig(nniqat.LinearReLU) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(torch.nn.Linear) + .set_reference_quantized_module(nnqr.Linear)) + # 2.3 functional linear + relu configs + # linear relu, functional linear + relu module + linear_configs.append( + BackendPatternConfig((F.linear, torch.nn.ReLU)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs)) + # linear relu, functional linear + functional relu + linear_configs.append( + BackendPatternConfig((F.linear, F.relu)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs)) + + # (3) Linear + batchnorm + # ------------------------ + # 3.1 linear bn fusion + linear_configs.append( + BackendPatternConfig((nn.Linear, nn.BatchNorm1d)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(fuse_linear_bn) + .set_fused_module(nni.LinearBn1d)) + + # 3.2 linear bn fused + # linear bn, fused module + linear_configs.append( + BackendPatternConfig(nni.LinearBn1d) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(torch.nn.Linear) + .set_reference_quantized_module(nnqr.Linear) + .set_qat_module(nniqat.LinearBn1d)) + # linear bn, qat fused module + linear_configs.append( + BackendPatternConfig(nniqat.LinearBn1d) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(torch.nn.Linear) + .set_reference_quantized_module(nnqr.Linear)) + return linear_configs + +def _get_conv_configs(dtype_configs): + """ + Return all configs related to conv modules and ops. + """ + conv_configs = [] + observation_type = ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT + for convs in [_Conv1dMetadata, _Conv2dMetadata, _Conv3dMetadata]: + + # (1) Single conv modules/functions + # ----------------------------------- + # conv module + conv_configs.append( + BackendPatternConfig(convs.root) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(convs.root) + .set_reference_quantized_module(convs.reference) + .set_qat_module(convs.qat)) + # conv qat module + conv_configs.append( + BackendPatternConfig(convs.qat) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(convs.root) + .set_reference_quantized_module(convs.reference)) + # functional conv + conv_configs.append( + BackendPatternConfig(convs.func) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 1, "bias": 2})) + + # (2) Conv + relu + # ----------------- + # 2.1 conv module + relu fusion configs + # conv relu fusion, conv module + relu module + conv_configs.append( + BackendPatternConfig((convs.root, torch.nn.ReLU)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(_sequential_wrapper2(convs.fused_conv_relu)) + .set_fused_module(convs.fused_conv_relu)) + # conv relu fusion, conv module + functional relu + conv_configs.append( + BackendPatternConfig((convs.root, F.relu)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(_sequential_wrapper2(convs.fused_conv_relu)) + .set_fused_module(convs.fused_conv_relu)) + # 2.2 conv module + relu fused module configs + # conv relu, fused module + conv_configs.append( + BackendPatternConfig(convs.fused_conv_relu) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(convs.root) + .set_reference_quantized_module(convs.reference) + .set_qat_module(convs.relu_qat)) + # conv relu, qat fused module + conv_configs.append( + BackendPatternConfig(convs.relu_qat) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(convs.root) + .set_reference_quantized_module(convs.reference)) + # 2.3 functional conv + relu configs + # conv relu, functional conv + relu module + conv_configs.append( + BackendPatternConfig((convs.func, torch.nn.ReLU)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs)) + # conv relu, functional conv + functional relu + conv_configs.append( + BackendPatternConfig((convs.func, F.relu)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs)) + + # fused conv relu + conv_configs.append( + BackendPatternConfig(convs.fused_conv_relu) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_qat_module(convs.relu_qat)) + + conv_configs.append( + BackendPatternConfig(convs.relu_qat) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_root_module(convs.root) + .set_reference_quantized_module(convs.reference)) + + # (3) Conv + batchnorm (+ relu) + # ------------------------------- + # 3.1 conv bn fusion configs + # conv + bn fusion + conv_configs.append( + BackendPatternConfig((convs.root, convs.bn)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(fuse_conv_bn) + .set_fused_module(convs.fused_conv_bn)) + # conv + bn + relu module fusion + conv_configs.append( + BackendPatternConfig((convs.root, convs.bn, nn.ReLU)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(fuse_conv_bn_relu) + .set_fused_module(convs.fused_conv_bn_relu)) + # conv + bn + relu functional fusion + conv_configs.append( + BackendPatternConfig((convs.root, convs.bn, F.relu)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_root_module(convs.root) + .set_fuser_method(fuse_conv_bn_relu) + .set_fused_module(convs.fused_conv_bn_relu)) + # TODO: we can add fusion for torch.relu as well + + # 3.2 conv + bn (+ relu) fused module configs + # fused conv bn + conv_configs.append( + BackendPatternConfig(convs.fused_conv_bn) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_qat_module(convs.bn_qat)) + + # fused conv bn relu + conv_configs.append( + BackendPatternConfig(convs.fused_conv_bn_relu) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_qat_module(convs.bn_relu_qat)) + + # conv bn, qat fused module + conv_configs.append( + BackendPatternConfig(convs.bn_qat) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(convs.root) + .set_reference_quantized_module(convs.reference)) + # conv bn relu, qat fused module + conv_configs.append( + BackendPatternConfig(convs.bn_relu_qat) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(convs.root) + .set_reference_quantized_module(convs.reference)) + + # (4) conv transpose and its fusion + # 4.1 conv transpose config + conv_configs.append( + BackendPatternConfig(convs.transpose) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_root_module(convs.transpose) + .set_reference_quantized_module(convs.transpose_reference)) + + # 4.2 conv transpose + bn fusion + conv_configs.append( + BackendPatternConfig((convs.transpose, convs.bn)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(fuse_convtranspose_bn) + .set_root_module(convs.transpose) + .set_reference_quantized_module(convs.transpose_reference)) + + # 4.3 functional conv transpose + conv_configs.append( + BackendPatternConfig(convs.func_transpose) + .set_dtype_configs(dtype_configs) # noqa: E131 + ._set_input_type_to_index({"weight": 1, "bias": 2})) + + return conv_configs + +def _get_cat_config(dtype_configs: List[DTypeConfig]) -> BackendPatternConfig: + return BackendPatternConfig(torch.cat) \ + .set_observation_type(ObservationType.OUTPUT_SHARE_OBSERVER_WITH_INPUT) \ + .set_dtype_configs(dtype_configs) + +def _get_ln_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + ln_configs = [] + ln_configs.append( + BackendPatternConfig(torch.nn.LayerNorm) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs) + ) + ln_configs.append( + BackendPatternConfig(torch.nn.functional.layer_norm) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 2, "bias": 3}) + ) + return ln_configs + +def _get_default_op_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + configs = [] + default_ops = [ + torch.nn.ELU, + torch.nn.LeakyReLU, + torch.nn.Hardswish, + torch.nn.InstanceNorm1d, + torch.nn.InstanceNorm2d, + torch.nn.InstanceNorm3d, + torch.nn.Dropout, + torch.nn.PReLU, + torch.nn.functional.elu, + torch.nn.functional.hardswish, + torch.nn.functional.leaky_relu, + torch.nn.functional.dropout, + ] + for op in default_ops: + configs.append( + BackendPatternConfig(op) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs)) + + configs.append( + BackendPatternConfig(torch.nn.functional.group_norm) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 2, "bias": 3}) + ) + + configs.append( + BackendPatternConfig(torch.nn.functional.instance_norm) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 3, "bias": 4}) + ) + return configs + +def _add_fixed_qparams_to_dtype_configs( + dtype_configs: List[DTypeConfig], + constraints: DTypeWithConstraints, +) -> List[DTypeConfig]: + """ + Return a copy of the list of DTypeConfigs where activations are subject to the specified + constraints required for fixed qparams ops. + + If the data type doesn't match the one in the constraints, simply leave the corresponding + DTypeConfig unchanged. + + If `scale_min_lower_bound` or `scale_max_upper_bound` is specified in the activations, + throw an exception since these settings are incompatible with fixed qparams ops. + """ + new_dtype_configs = [] + for dtype_config in dtype_configs: + dc = copy.deepcopy(dtype_config) + for orig_constraints in [dc.input_dtype_with_constraints, dc.output_dtype_with_constraints]: + if orig_constraints.dtype != constraints.dtype: + continue + if orig_constraints.scale_min_lower_bound is not None: + raise ValueError(f"scale_min_lower_bound is invalid for fixed qparams ops: {dtype_config}") + if orig_constraints.scale_max_upper_bound is not None: + raise ValueError(f"scale_max_upper_bound is invalid for fixed qparams ops: {dtype_config}") + orig_constraints.quant_min_lower_bound = constraints.quant_min_lower_bound + orig_constraints.quant_max_upper_bound = constraints.quant_max_upper_bound + orig_constraints.scale_exact_match = constraints.scale_exact_match + orig_constraints.zero_point_exact_match = constraints.zero_point_exact_match + new_dtype_configs.append(dc) + return new_dtype_configs + +def _get_fixed_qparams_op_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + fixed_qparams_op_configs = [] + for fixed_qparam_op, constraints in _FIXED_QPARAMS_OP_TO_CONSTRAINTS.items(): + new_dtype_configs = _add_fixed_qparams_to_dtype_configs(dtype_configs, constraints) + fixed_qparams_op_configs.append( + BackendPatternConfig(fixed_qparam_op) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(new_dtype_configs)) + return fixed_qparams_op_configs + +def _get_share_qparams_op_configs(dtype_configs): + """ Get the operator config for the operators that works for both float and quantized input + if input is quantized, the output Tensor shares the same quantization parameter + with input. + Example operator: avgpool2d, reshape, transpose, maxpool2d + Example observed operator: + observer_0 - avgpool2d - observer_0 (same observer instance as input) + """ + + def _get_share_qprams_op_backend_config(op): + return BackendPatternConfig(op) \ + .set_observation_type(ObservationType.OUTPUT_SHARE_OBSERVER_WITH_INPUT) \ + .set_dtype_configs(dtype_configs) + + share_qparams_ops = [ + torch.nn.AdaptiveAvgPool1d, + torch.nn.AdaptiveAvgPool2d, + torch.nn.AdaptiveAvgPool3d, + torch.nn.AvgPool1d, + torch.nn.AvgPool2d, + torch.nn.AvgPool3d, + torch.nn.Hardtanh, + torch.nn.Identity, + torch.nn.MaxPool1d, + torch.nn.MaxPool2d, + torch.nn.MaxPool3d, + torch.nn.PixelShuffle, + torch.nn.PixelUnshuffle, + torch.nn.ReLU, + torch.nn.ReLU6, + torch.adaptive_avg_pool1d, + torch.nn.functional.adaptive_avg_pool2d, + torch.nn.functional.adaptive_avg_pool3d, + torch.nn.functional.hardtanh, + torch.nn.functional.hardtanh_, + torch.nn.functional.interpolate, + torch.nn.functional.max_pool1d, + torch.nn.functional.max_pool2d, + torch.nn.functional.max_pool3d, + torch.nn.functional.pixel_shuffle, + torch.nn.functional.pixel_unshuffle, + torch.nn.functional.relu, + torch.nn.functional.relu6, + torch.avg_pool1d, + torch._C._nn.avg_pool2d, + torch._C._nn.avg_pool3d, + torch.clamp, + torch.flatten, + torch.mean, + torch.narrow, + torch.repeat_interleave, + torch.transpose, + torch.squeeze, + torch.stack, + torch.unsqueeze, + operator.floordiv, + "contiguous", + "clamp", + "detach", + "detach_", + "mean", + "permute", + "repeat", + "repeat_interleave", + "reshape", + "resize_", + "relu", + "relu_", + "squeeze", + "squeeze_", + "transpose", + "unsqueeze", + "unsqueeze_", + "view" + ] + return [_get_share_qprams_op_backend_config(op) for op in share_qparams_ops] + +def _get_bn_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + """ Get configs related to batchnorm. """ + bn_configs = [] + bn_to_fused_bn = { + torch.nn.BatchNorm2d: nni.BNReLU2d, + torch.nn.BatchNorm3d: nni.BNReLU3d, + } + for bn in bn_to_fused_bn.keys(): + fused_bn = bn_to_fused_bn[bn] + # bn module + relu module fusion config + bn_configs.append( + BackendPatternConfig((bn, nn.ReLU)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(_sequential_wrapper2(fused_bn)) + .set_fused_module(fused_bn)) + # bn module + F.relu fusion config + bn_configs.append( + BackendPatternConfig((bn, F.relu)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(_sequential_wrapper2(fused_bn)) + .set_fused_module(fused_bn)) + bn_configs.append( + BackendPatternConfig(bn) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs)) + + # fused bn configs + for fused_bn in bn_to_fused_bn.values(): + bn_configs.append( + BackendPatternConfig(fused_bn) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs)) + return bn_configs + +def _get_rnn_op_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + rnn_op_configs = [] + for rnn_op, ref_rnn_op in [ + (nn.GRUCell, nnqr.GRUCell), + (nn.LSTMCell, nnqr.LSTMCell), + (nn.RNNCell, nnqr.RNNCell), + (nn.LSTM, nnqr.LSTM), + (nn.GRU, nnqr.GRU) + ]: + rnn_op_configs.append( + BackendPatternConfig(rnn_op) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(rnn_op) + .set_reference_quantized_module(ref_rnn_op)) + return rnn_op_configs + +def _get_embedding_op_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConfig]: + embedding_op_configs = [] + for embedding_op, qat_embedding_op, ref_embedding_op in [ + (nn.Embedding, nnqat.Embedding, nnqr.Embedding), + (nn.EmbeddingBag, nnqat.EmbeddingBag, nnqr.EmbeddingBag), + ]: + embedding_op_configs.append( + BackendPatternConfig(embedding_op) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_qat_module(qat_embedding_op) + .set_root_module(embedding_op) + .set_reference_quantized_module(ref_embedding_op)) + + # config for qat op + embedding_op_configs.append( + BackendPatternConfig(qat_embedding_op) + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(embedding_op) + .set_reference_quantized_module(ref_embedding_op)) + return embedding_op_configs + +def _get_tensor_info_op_configs(dtype_configs): + """ + These ops work on tensors of different dtypes but return non-tensors + containing information about the input tensor. + """ + + def _get_config(op): + return BackendPatternConfig(op) \ + .set_observation_type(ObservationType.INPUT_OUTPUT_NOT_OBSERVED) \ + .set_dtype_configs(dtype_configs) + + return [_get_config(op) for op in ("shape", "size")] diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/_qnnpack_pt2e.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/_qnnpack_pt2e.py new file mode 100644 index 0000000000000000000000000000000000000000..01e112b688c0428cadb5e31502f18387bcd9282f --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/_qnnpack_pt2e.py @@ -0,0 +1,160 @@ +import operator +import torch +from torch.ao.quantization.backend_config import ( + BackendConfig, + DTypeConfig, + ObservationType, + BackendPatternConfig, +) + +weighted_op_quint8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, + weight_dtype=torch.qint8, + bias_dtype=torch.float, +) +from typing import List + +def get_linear_configs(): + linear_configs = [] + observation_type = ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT + dtype_configs = [weighted_op_quint8_dtype_config] + + # TODO: need to fix the way we insert observers for this pattern + # should be solved in the new fusion API + # reason that this doesn't work: the pattern is a bit complicated and we don't + # have a way to specify which input of the pattern we would like to observe + # pattern: + # bias input weight + # \ | / + # \ | t + # \ | / + # addmm + # we want to observe "weight" as weight, but there is not way to convey this + # information with current pattern language + # + # right now: + # original: + # weight - t \ + # input - addmm + # observed (no hack): + # weight - t - observer \ + # input - observer - addmm + # target: + # weight - observer - t \ + # input - observer - addmm + + # def root_node_getter(node_pattern): + # addmm, bias, act, weight = node_pattern + # return addmm + + # linear_configs.append( + # BackendPatternConfig((torch.ops.aten.addmm.default, MatchAllNode, MatchAllNode, torch.ops.aten.t.default)) + # .set_observation_type(observation_type) # noqa: E131 + # .set_dtype_configs(dtype_configs) + # ._set_root_node_getter(root_node_getter)) + + linear_configs.append( + BackendPatternConfig(torch.ops.aten.addmm.default) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 2, "bias": 0}) + ) + # linear is decomposed to `t - mm` if bias is not present + linear_configs.append( + BackendPatternConfig(torch.ops.aten.mm.default) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 1}) + ) + return linear_configs + +def get_conv_configs(): + conv_configs = [] + observation_type = ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT + dtype_configs = [weighted_op_quint8_dtype_config] + conv_configs.append( + BackendPatternConfig(torch.ops.aten.convolution.default) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 1, "bias": 2}) + ) + conv_configs.append( + BackendPatternConfig((torch.ops.aten.convolution.default, torch.ops.aten.relu.default)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 1, "bias": 2}) + ) + # TODO: remove when functionalization is supported in PT2 mode + conv_configs.append( + BackendPatternConfig((torch.ops.aten.convolution.default, torch.ops.aten.relu_.default)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_input_type_to_index({"weight": 1, "bias": 2}) + ) + return conv_configs + +def get_pooling_configs(): + backend_pattern_configs = [] + observation_type = ObservationType.OUTPUT_SHARE_OBSERVER_WITH_INPUT + dtype_configs = [weighted_op_quint8_dtype_config] + + def root_node_getter(node_pattern): + getitem, maxpool, index = node_pattern + return maxpool + + backend_pattern_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((operator.getitem, torch.ops.aten.max_pool2d_with_indices.default, 0)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + ._set_root_node_getter(root_node_getter) + ) + + return backend_pattern_configs + +def get_relu_configs(): + backend_pattern_configs = [] + observation_type = ObservationType.OUTPUT_SHARE_OBSERVER_WITH_INPUT + dtype_configs = [weighted_op_quint8_dtype_config] + backend_pattern_configs.append( + BackendPatternConfig(torch.ops.aten.relu.default) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs)) + return backend_pattern_configs + +def get_binary_op_configs(): + binary_op_configs: List[BackendPatternConfig] = [] + dtype_configs = [weighted_op_quint8_dtype_config] + num_tensor_args_to_observation_type_mapping = { + # TODO: this is not used right now since we have extra check in prepare + # will need to change this to NO_OBSERVER later after we implemented + # Tensor dtype inference properly + 0: ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT, + 1: ObservationType.OUTPUT_SHARE_OBSERVER_WITH_INPUT, + 2: ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT, + } + for op_with_quantized_bop_scalar_variant in [torch.ops.aten.add.Tensor, torch.ops.aten.add_.Tensor]: + bop_patterns = [ + (op_with_quantized_bop_scalar_variant, torch.ops.aten.relu.default), + op_with_quantized_bop_scalar_variant, + # TODO: remove when functionalization is supported in pt2_mode + (op_with_quantized_bop_scalar_variant, torch.ops.aten.relu_.default), + ] + for bop_pattern in bop_patterns: + binary_op_configs.append( + BackendPatternConfig(bop_pattern) + .set_dtype_configs(dtype_configs) # noqa: E131 + ._set_num_tensor_args_to_observation_type(num_tensor_args_to_observation_type_mapping)) + + return binary_op_configs + +def get_qnnpack_pt2e_backend_config(): + return ( + BackendConfig("qnnpack_pytorch_2.0_export") + .set_backend_pattern_configs(get_linear_configs()) + .set_backend_pattern_configs(get_binary_op_configs()) + .set_backend_pattern_configs(get_conv_configs()) + .set_backend_pattern_configs(get_pooling_configs()) + .set_backend_pattern_configs(get_relu_configs()) + ) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/backend_config.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/backend_config.py new file mode 100644 index 0000000000000000000000000000000000000000..e5a4d2f3afa349688365fe19e498cd3bedcb08e7 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/backend_config.py @@ -0,0 +1,659 @@ +from __future__ import annotations +from dataclasses import dataclass +from typing import Any, Callable, Dict, List, Optional, Type, Union + +import torch +from torch.ao.quantization.utils import Pattern +from enum import Enum + + +__all__ = [ + "BackendConfig", + "BackendPatternConfig", + "DTypeConfig", + "DTypeWithConstraints", + "ObservationType", +] + + +# DTypeConfig dict keys +INPUT_DTYPE_DICT_KEY = "input_dtype" +OUTPUT_DTYPE_DICT_KEY = "output_dtype" +WEIGHT_DTYPE_DICT_KEY = "weight_dtype" +BIAS_DTYPE_DICT_KEY = "bias_dtype" +IS_DYNAMIC_DICT_KEY = "is_dynamic" + +# BackendConfig dict keys +NAME_DICT_KEY = "name" +CONFIGS_DICT_KEY = "configs" + +# BackendPatternConfig dict keys +PATTERN_DICT_KEY = "pattern" +PATTERN_COMPLEX_FORMAT_DICT_KEY = "pattern_complex_format" +OBSERVATION_TYPE_DICT_KEY = "observation_type" +DTYPE_CONFIGS_DICT_KEY = "dtype_configs" +ROOT_MODULE_DICT_KEY = "root_module" +QAT_MODULE_DICT_KEY = "qat_module" +REFERENCE_QUANTIZED_MODULE_DICT_KEY = "reference_quantized_module_for_root" +FUSED_MODULE_DICT_KEY = "fused_module" +FUSER_METHOD_DICT_KEY = "fuser_method" +ROOT_NODE_GETTER_DICT_KEY = "root_node_getter" +EXTRA_INPUTS_GETTER_DICT_KEY = "extra_inputs_getter" +NUM_TENSOR_ARGS_TO_OBSERVATION_TYPE_DICT_KEY = "num_tensor_args_to_observation_type" +INPUT_TYPE_TO_INDEX_DICT_KEY = "input_type_to_index" + + +# TODO: maybe rename this to something that's not related to observer +# e.g. QParamsType +class ObservationType(Enum): + """ An enum that represents different ways of how an operator/operator pattern + should be observed + """ + + OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT = 0 + """this means input and output are observed with different observers, based + on qconfig.activation + example: conv, linear, softmax + """ + + OUTPUT_SHARE_OBSERVER_WITH_INPUT = 1 + """this means the output will use the same observer instance as input, based + on qconfig.activation + example: torch.cat, maxpool + """ + + INPUT_OUTPUT_NOT_OBSERVED = 2 + """this means the input and output are never observed + example: x.shape, x.size + """ + + +@dataclass +class DTypeWithConstraints: + """ + Config for specifying additional constraints for a given dtype, such as quantization + value ranges, scale value ranges, and fixed quantization params, to be used in + :class:`~torch.ao.quantization.backend_config.DTypeConfig`. + + The constraints currently supported are: + + * `quant_min_lower_bound` and `quant_max_upper_bound`: Lower and upper + bounds for the minimum and maximum quantized values respectively. If + the QConfig’s `quant_min` and `quant_max` fall outside this range, + then the QConfig will be ignored. + + * `scale_min_lower_bound` and `scale_max_upper_bound`: Lower and upper + bounds for the minimum and maximum scale values respectively. If the + QConfig’s minimum scale value (currently exposed as `eps`) falls below + the lower bound, then the QConfig will be ignored. Note that the upper + bound is currently not enforced. + + * `scale_exact_match` and `zero_point_exact_match`: Exact match requirements + for scale and zero point, to be used for operators with fixed quantization + parameters such as sigmoid and tanh. If the observer specified in the QConfig + is neither `FixedQParamsObserver` nor `FixedQParamsFakeQuantize`, or if + the quantization parameters don't match, then the QConfig will be ignored. + """ + dtype: Optional[torch.dtype] = None + quant_min_lower_bound: Union[int, float, None] = None + quant_max_upper_bound: Union[int, float, None] = None + scale_min_lower_bound: Union[int, float, None] = None + scale_max_upper_bound: Union[int, float, None] = None + scale_exact_match: Optional[float] = None + zero_point_exact_match: Optional[int] = None + + +@dataclass +class DTypeConfig: + """ + Config object that specifies the supported data types passed as arguments to + quantize ops in the reference model spec, for input and output activations, + weights, and biases. + + For example, consider the following reference model: + + quant1 - [dequant1 - fp32_linear - quant2] - dequant2 + + The pattern in the square brackets refers to the reference pattern of + statically quantized linear. Setting the input dtype as `torch.quint8` + in the DTypeConfig means we pass in `torch.quint8` as the dtype argument + to the first quantize op (quant1). Similarly, setting the output dtype as + `torch.quint8` means we pass in `torch.quint8` as the dtype argument to + the second quantize op (quant2). + + Note that the dtype here does not refer to the interface dtypes of the + op. For example, the "input dtype" here is not the dtype of the input + tensor passed to the quantized linear op. Though it can still be the + same as the interface dtype, this is not always the case, e.g. the + interface dtype is fp32 in dynamic quantization but the "input dtype" + specified in the DTypeConfig would still be quint8. The semantics of + dtypes here are the same as the semantics of the dtypes specified in + the observers. + + These dtypes are matched against the ones specified in the user’s + QConfig. If there is a match, and the QConfig satisfies the constraints + specified in the DTypeConfig (if any), then we will quantize the given + pattern using this DTypeConfig. Otherwise, the QConfig is ignored and + the pattern will not be quantized. + + Example usage:: + + >>> # xdoctest: +SKIP(failing) + >>> dtype_config1 = DTypeConfig( + ... input_dtype=torch.quint8, + ... output_dtype=torch.quint8, + ... weight_dtype=torch.qint8, + ... bias_dtype=torch.float) + + >>> dtype_config2 = DTypeConfig( + ... input_dtype=DTypeWithConstraints( + ... dtype=torch.quint8, + ... quant_min_lower_bound=0, + ... quant_max_upper_bound=255, + ... ), + ... output_dtype=DTypeWithConstraints( + ... dtype=torch.quint8, + ... quant_min_lower_bound=0, + ... quant_max_upper_bound=255, + ... ), + ... weight_dtype=DTypeWithConstraints( + ... dtype=torch.qint8, + ... quant_min_lower_bound=-128, + ... quant_max_upper_bound=127, + ... ), + ... bias_dtype=torch.float) + + >>> dtype_config1.input_dtype + torch.quint8 + + >>> dtype_config2.input_dtype + torch.quint8 + + >>> dtype_config2.input_dtype_with_constraints + DTypeWithConstraints(dtype=torch.quint8, quant_min_lower_bound=0, quant_max_upper_bound=255, \ +scale_min_lower_bound=None, scale_max_upper_bound=None) + """ + input_dtype_with_constraints: DTypeWithConstraints + output_dtype_with_constraints: DTypeWithConstraints + weight_dtype_with_constraints: DTypeWithConstraints + bias_dtype: Optional[torch.dtype] + is_dynamic: Optional[bool] + + def __init__( + self, + input_dtype: Union[torch.dtype, DTypeWithConstraints, None] = None, + output_dtype: Union[torch.dtype, DTypeWithConstraints, None] = None, + weight_dtype: Union[torch.dtype, DTypeWithConstraints, None] = None, + bias_dtype: Optional[torch.dtype] = None, + is_dynamic: Optional[bool] = None, + ): + if isinstance(input_dtype, DTypeWithConstraints): + self.input_dtype_with_constraints = input_dtype + else: + self.input_dtype_with_constraints = DTypeWithConstraints(dtype=input_dtype) + + if isinstance(output_dtype, DTypeWithConstraints): + self.output_dtype_with_constraints = output_dtype + else: + self.output_dtype_with_constraints = DTypeWithConstraints(dtype=output_dtype) + + if isinstance(weight_dtype, DTypeWithConstraints): + self.weight_dtype_with_constraints = weight_dtype + else: + self.weight_dtype_with_constraints = DTypeWithConstraints(dtype=weight_dtype) + + self.bias_dtype = bias_dtype + self.is_dynamic = is_dynamic + + @property + def input_dtype(self) -> Optional[torch.dtype]: + return self.input_dtype_with_constraints.dtype + + @property + def output_dtype(self) -> Optional[torch.dtype]: + return self.output_dtype_with_constraints.dtype + + @property + def weight_dtype(self) -> Optional[torch.dtype]: + return self.weight_dtype_with_constraints.dtype + + @classmethod + def from_dict(cls, dtype_config_dict: Dict[str, Any]) -> DTypeConfig: + """ + Create a ``DTypeConfig`` from a dictionary with the following items (all optional): + "input_dtype": torch.dtype or ``DTypeWithConstraints`` + "output_dtype": torch.dtype or ``DTypeWithConstraints`` + "weight_dtype": torch.dtype or ``DTypeWithConstraints`` + "bias_type": torch.dtype + "is_dynamic": bool + """ + input_dtype = dtype_config_dict.get(INPUT_DTYPE_DICT_KEY, None) + if input_dtype is not None and not isinstance(input_dtype, (torch.dtype, DTypeWithConstraints)): + raise ValueError("Expected input_dtype to be a torch.dtype or DTypeWithConstraints") + output_dtype = dtype_config_dict.get(OUTPUT_DTYPE_DICT_KEY, None) + if output_dtype is not None and not isinstance(output_dtype, (torch.dtype, DTypeWithConstraints)): + raise ValueError("Expected output_dtype to be a torch.dtype or DTypeWithConstraints") + weight_dtype = dtype_config_dict.get(WEIGHT_DTYPE_DICT_KEY, None) + if weight_dtype is not None and not isinstance(weight_dtype, (torch.dtype, DTypeWithConstraints)): + raise ValueError("Expected weight_dtype to be a torch.dtype or DTypeWithConstraints") + bias_dtype = dtype_config_dict.get(BIAS_DTYPE_DICT_KEY, None) + is_dynamic = dtype_config_dict.get(IS_DYNAMIC_DICT_KEY, None) + return cls(input_dtype, output_dtype, weight_dtype, bias_dtype, is_dynamic) + + def to_dict(self) -> Dict[str, Any]: + """ + Convert this ``DTypeConfig`` to a dictionary with the items described in + :func:`~torch.ao.quantization.backend_config.DTypeConfig.from_dict`. + """ + dtype_config_dict: Dict[str, Any] = {} + if self.input_dtype is not None: + dtype_config_dict[INPUT_DTYPE_DICT_KEY] = self.input_dtype_with_constraints + if self.output_dtype is not None: + dtype_config_dict[OUTPUT_DTYPE_DICT_KEY] = self.output_dtype_with_constraints + if self.weight_dtype is not None: + dtype_config_dict[WEIGHT_DTYPE_DICT_KEY] = self.weight_dtype_with_constraints + if self.bias_dtype is not None: + dtype_config_dict[BIAS_DTYPE_DICT_KEY] = self.bias_dtype + if self.is_dynamic is not None: + dtype_config_dict[IS_DYNAMIC_DICT_KEY] = self.is_dynamic + return dtype_config_dict + + +class BackendConfig: + # TODO: refer to NativeBackendConfig once that is implemented + """Config that defines the set of patterns that can be quantized on a given backend, and how reference + quantized models can be produced from these patterns. + + A pattern in this context refers to a module, a functional, an operator, or a directed acyclic graph + of the above. Each pattern supported on the target backend can be individually configured through + :class:`~torch.ao.quantization.backend_config.BackendPatternConfig` in terms of: + + (1) The supported input/output activation, weight, and bias data types + + (2) How observers and quant/dequant ops are inserted in order to construct the reference pattern, and + + (3) (Optionally) Fusion, QAT, and reference module mappings. + + The format of the patterns is described in: + https://github.com/pytorch/pytorch/blob/master/torch/ao/quantization/backend_config/README.md + + Example usage:: + + import torch + from torch.ao.quantization.backend_config import ( + BackendConfig, + BackendPatternConfig, + DTypeConfig, + ObservationType, + ) + + weighted_int8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, + weight_dtype=torch.qint8, + bias_dtype=torch.float) + + def fuse_conv2d_relu(is_qat, conv, relu): + return torch.ao.nn.intrinsic.ConvReLU2d(conv, relu) + + # For quantizing Linear + linear_config = BackendPatternConfig(torch.nn.Linear) \ + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) \ + .add_dtype_config(weighted_int8_dtype_config) \ + .set_root_module(torch.nn.Linear) \ + .set_qat_module(torch.ao.nn.qat.Linear) \ + .set_reference_quantized_module(torch.ao.nn.quantized.reference.Linear) + + # For fusing Conv2d + ReLU into ConvReLU2d + conv_relu_config = BackendPatternConfig((torch.nn.Conv2d, torch.nn.ReLU)) \ + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) \ + .add_dtype_config(weighted_int8_dtype_config) \ + .set_fused_module(torch.ao.nn.intrinsic.ConvReLU2d) \ + .set_fuser_method(fuse_conv2d_relu) + + # For quantizing ConvReLU2d + fused_conv_relu_config = BackendPatternConfig(torch.ao.nn.intrinsic.ConvReLU2d) \ + .set_observation_type(ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT) \ + .add_dtype_config(weighted_int8_dtype_config) \ + .set_root_module(torch.nn.Conv2d) \ + .set_qat_module(torch.ao.nn.intrinsic.qat.ConvReLU2d) \ + .set_reference_quantized_module(torch.ao.nn.quantized.reference.Conv2d) + + backend_config = BackendConfig("my_backend") \ + .set_backend_pattern_config(linear_config) \ + .set_backend_pattern_config(conv_relu_config) \ + .set_backend_pattern_config(fused_conv_relu_config) + + """ + def __init__(self, name: str = ""): + self.name = name + # Store all BackendPatternConfigs in a map to handle duplicates + # Note: the key in this map uses the complex reversed tuple format. + # This is intended only for internal use; users who wish to access + # the original patterns should go through `self.configs` instead. + self._pattern_complex_format_to_config: Dict[Pattern, BackendPatternConfig] = {} + + def __repr__(self): + return f"BackendConfig({self.__dict__})" + + def set_name(self, name: str) -> BackendConfig: + """ + Set the name of the target backend. + """ + self.name = name + return self + + def set_backend_pattern_config(self, config: BackendPatternConfig) -> BackendConfig: + """ + Set the config for an pattern that can be run on the target backend. + This overrides any existing config for the given pattern. + """ + # Avoid circular dependencies + pattern_complex_format = torch.ao.quantization.backend_config.utils \ + ._get_pattern_in_reversed_nested_tuple_format(config) # type: ignore[attr-defined] + self._pattern_complex_format_to_config[pattern_complex_format] = config + return self + + def set_backend_pattern_configs(self, configs: List[BackendPatternConfig]) -> BackendConfig: + """ + Set the configs for patterns that can be run on the target backend. + This overrides any existing config for a given pattern if it was previously registered already. + """ + for conf in configs: + self.set_backend_pattern_config(conf) + return self + + @property + def configs(self) -> List[BackendPatternConfig]: + """ + Return a copy of the list of configs set in this `BackendConfig`. + """ + return list(self._pattern_complex_format_to_config.values()) + + @classmethod + def from_dict(cls, backend_config_dict: Dict[str, Any]) -> BackendConfig: + """ + Create a ``BackendConfig`` from a dictionary with the following items: + + "name": the name of the target backend + + "configs": a list of dictionaries that each represents a `BackendPatternConfig` + + """ + conf = cls(backend_config_dict.get(NAME_DICT_KEY, "")) + for d in backend_config_dict.get(CONFIGS_DICT_KEY, []): + if isinstance(d, BackendPatternConfig): + conf.set_backend_pattern_config(d) + elif isinstance(d, Dict): + conf.set_backend_pattern_config(BackendPatternConfig.from_dict(d)) + else: + raise ValueError(f"Expected backend_config_dict['{CONFIGS_DICT_KEY}'] to be a dictionary") + return conf + + def to_dict(self) -> Dict[str, Any]: + """ + Convert this ``BackendConfig`` to a dictionary with the items described in + :func:`~torch.ao.quantization.backend_config.BackendConfig.from_dict`. + """ + return { + NAME_DICT_KEY: self.name, + CONFIGS_DICT_KEY: [c.to_dict() for c in self.configs], + } + + +class BackendPatternConfig: + """ + Config object that specifies quantization behavior for a given operator pattern. + For a detailed example usage, see :class:`~torch.ao.quantization.backend_config.BackendConfig`. + """ + def __init__(self, pattern: Optional[Pattern] = None): + self.pattern: Optional[Pattern] = pattern + self.observation_type = ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT + self.dtype_configs: List[DTypeConfig] = [] + self.root_module: Optional[Type[torch.nn.Module]] = None + self.qat_module: Optional[Type[torch.nn.Module]] = None + self.reference_quantized_module: Optional[Type[torch.nn.Module]] = None + self.fused_module: Optional[Type[torch.nn.Module]] = None + self.fuser_method: Optional[Callable] = None + + # Temporary/internal configs + self._root_node_getter: Optional[Callable] = None + self._extra_inputs_getter: Optional[Callable] = None + self._num_tensor_args_to_observation_type: Dict[int, ObservationType] = {} + self._input_type_to_index: Dict[str, int] = {} + self._pattern_complex_format: Optional[Pattern] = None + + def __repr__(self): + dict_nonempty = { + k: v for k, v in self.__dict__.items() + if ( + (not isinstance(v, (list, dict)) and v is not None) + or (isinstance(v, (list, dict)) and len(v) > 0) + ) + } + return f"BackendPatternConfig({dict_nonempty})" + + def set_pattern(self, pattern: Pattern) -> BackendPatternConfig: + """ + Set the pattern to configure. + + The pattern can be a float module, functional operator, pytorch operator, or a tuple + combination of the above. Tuple patterns are treated as sequential patterns, and + currently only tuples of 2 or 3 elements are supported. + """ + if self._pattern_complex_format is not None: + raise ValueError("Only one of 'pattern' or 'pattern_complex_format' can be set") + self.pattern = pattern + return self + + def set_observation_type(self, observation_type: ObservationType) -> BackendPatternConfig: + """ + Set how observers should be inserted in the graph for this pattern. + + Observation type here refers to how observers (or quant-dequant ops) will be placed + in the graph. This is used to produce the desired reference patterns understood by + the backend. Weighted ops such as linear and conv require different observers + (or quantization parameters passed to quantize ops in the reference model) for the + input and the output. + + There are two observation types: + + `OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT` (default): the output observer instance + will be different from the input. This is the most common observation type. + + `OUTPUT_SHARE_OBSERVER_WITH_INPUT`: the output observer instance will be the + same as the input. This is useful for operators like `cat`. + + Note: This will be renamed in the near future, since we will soon insert QuantDeQuantStubs + with observers (and fake quantizes) attached instead of observers themselves. + """ + self.observation_type = observation_type + return self + + def add_dtype_config(self, dtype_config: DTypeConfig) -> BackendPatternConfig: + """ + Add a set of supported data types passed as arguments to quantize ops in the + reference model spec. + """ + self.dtype_configs.append(dtype_config) + return self + + def set_dtype_configs(self, dtype_configs: List[DTypeConfig]) -> BackendPatternConfig: + """ + Set the supported data types passed as arguments to quantize ops in the + reference model spec, overriding all previously registered data types. + """ + self.dtype_configs = dtype_configs + return self + + def set_root_module(self, root_module: Type[torch.nn.Module]) -> BackendPatternConfig: + """ + Set the module that represents the root for this pattern. + + When we construct the reference quantized model during the convert phase, + the root modules (e.g. torch.nn.Linear for torch.ao.nn.intrinsic.LinearReLU) + will be swapped to the corresponding reference quantized modules (e.g. + torch.ao.nn.reference.quantized.Linear). This allows custom backends to + specify custom reference quantized module implementations to match the + numerics of their lowered operators. Since this is a one-to-one mapping, + both the root module and the reference quantized module must be specified + in the same BackendPatternConfig in order for the conversion to take place. + """ + self.root_module = root_module + return self + + def set_qat_module(self, qat_module: Type[torch.nn.Module]) -> BackendPatternConfig: + """ + Set the module that represents the QAT implementation for this pattern. + """ + self.qat_module = qat_module + return self + + def set_reference_quantized_module(self, reference_quantized_module: Type[torch.nn.Module]) -> BackendPatternConfig: + """ + Set the module that represents the reference quantized implementation for + this pattern's root module. + + For more detail, see :func:`~torch.ao.quantization.backend_config.BackendPatternConfig.set_root_module`. + """ + self.reference_quantized_module = reference_quantized_module + return self + + def set_fused_module(self, fused_module: Type[torch.nn.Module]) -> BackendPatternConfig: + """ + Set the module that represents the fused implementation for this pattern. + """ + self.fused_module = fused_module + return self + + def set_fuser_method(self, fuser_method: Callable) -> BackendPatternConfig: + """ + Set the function that specifies how to fuse this BackendPatternConfig's pattern. + + The first argument of this function should be `is_qat`, and the rest of the arguments + should be the items in the tuple pattern. The return value of this function should be + the resulting fused module. + + For example, the fuser method for the pattern `(torch.nn.Linear, torch.nn.ReLU)` can be: + + def fuse_linear_relu(is_qat, linear, relu): + return torch.ao.nn.intrinsic.LinearReLU(linear, relu) + + For a more complicated example, see https://gist.github.com/jerryzh168/8bea7180a8ba3c279f2c9b050f2a69a6. + """ + self.fuser_method = fuser_method + return self + + def _set_root_node_getter(self, root_node_getter: Callable) -> BackendPatternConfig: + self._root_node_getter = root_node_getter + return self + + def _set_extra_inputs_getter(self, extra_inputs_getter: Callable) -> BackendPatternConfig: + self._extra_inputs_getter = extra_inputs_getter + return self + + def _set_num_tensor_args_to_observation_type( + self, num_tensor_args_to_observation_type: Dict[int, ObservationType]) -> BackendPatternConfig: + self._num_tensor_args_to_observation_type = num_tensor_args_to_observation_type + return self + + def _set_input_type_to_index(self, input_type_to_index: Dict[str, int]) -> BackendPatternConfig: + self._input_type_to_index = input_type_to_index + return self + + def _set_pattern_complex_format(self, pattern: Pattern) -> BackendPatternConfig: + """ + Set the pattern to configure, using the reversed nested tuple format. + + See the BackendConfig README for more detail: + https://github.com/pytorch/pytorch/blob/master/torch/ao/quantization/backend_config/README.md#advanced-pattern-specification + """ + if self.pattern is not None: + raise ValueError("Only one of 'pattern' or 'pattern_complex_format' can be set") + self._pattern_complex_format = pattern + return self + + @classmethod + def from_dict(cls, backend_pattern_config_dict: Dict[str, Any]) -> BackendPatternConfig: + """ + Create a ``BackendPatternConfig`` from a dictionary with the following items: + + "pattern": the pattern being configured + "observation_type": the :class:`~torch.ao.quantization.backend_config.ObservationType` that specifies how + observers should be inserted for this pattern + "dtype_configs": a list of dictionaries that represents :class:`~torch.ao.quantization.backend_config.DTypeConfig` s + "root_module": a :class:`torch.nn.Module` that represents the root for this pattern + "qat_module": a :class:`torch.nn.Module` that represents the QAT implementation for this pattern + "reference_quantized_module": a :class:`torch.nn.Module` that represents the reference quantized + implementation for this pattern's root module. + "fused_module": a :class:`torch.nn.Module` that represents the fused implementation for this pattern + "fuser_method": a function that specifies how to fuse the pattern for this pattern + "pattern_complex_format": the pattern specified in the reversed nested tuple format (deprecated) + + """ + def _get_dtype_config(obj: Any) -> DTypeConfig: + """ + Convert the given object into a ``DTypeConfig`` if possible, else throw an exception. + """ + if isinstance(obj, DTypeConfig): + return obj + if isinstance(obj, Dict): + return DTypeConfig.from_dict(obj) + raise ValueError( + f"Expected a list of DTypeConfigs in " + f"backend_pattern_config_dict[\"{DTYPE_CONFIGS_DICT_KEY}\"], got '{type(obj)}'" + ) + + conf = cls() + if PATTERN_DICT_KEY in backend_pattern_config_dict: + conf.set_pattern(backend_pattern_config_dict[PATTERN_DICT_KEY]) + if OBSERVATION_TYPE_DICT_KEY in backend_pattern_config_dict: + conf.set_observation_type(backend_pattern_config_dict[OBSERVATION_TYPE_DICT_KEY]) + for d in backend_pattern_config_dict.get(DTYPE_CONFIGS_DICT_KEY, []): + conf.add_dtype_config(_get_dtype_config(d)) + conf.set_root_module(backend_pattern_config_dict.get(ROOT_MODULE_DICT_KEY, None)) + conf.set_qat_module(backend_pattern_config_dict.get(QAT_MODULE_DICT_KEY, None)) + conf.set_reference_quantized_module(backend_pattern_config_dict.get(REFERENCE_QUANTIZED_MODULE_DICT_KEY, None)) + conf.set_fused_module(backend_pattern_config_dict.get(FUSED_MODULE_DICT_KEY, None)) + conf.set_fuser_method(backend_pattern_config_dict.get(FUSER_METHOD_DICT_KEY, None)) + conf._set_root_node_getter(backend_pattern_config_dict.get(ROOT_NODE_GETTER_DICT_KEY, None)) + conf._set_extra_inputs_getter(backend_pattern_config_dict.get(EXTRA_INPUTS_GETTER_DICT_KEY, None)) + conf._set_num_tensor_args_to_observation_type( + backend_pattern_config_dict.get(NUM_TENSOR_ARGS_TO_OBSERVATION_TYPE_DICT_KEY, {})) + conf._set_input_type_to_index(backend_pattern_config_dict.get(INPUT_TYPE_TO_INDEX_DICT_KEY, {})) + if PATTERN_COMPLEX_FORMAT_DICT_KEY in backend_pattern_config_dict: + conf._set_pattern_complex_format(backend_pattern_config_dict[PATTERN_COMPLEX_FORMAT_DICT_KEY]) + return conf + + def to_dict(self) -> Dict[str, Any]: + """ + Convert this ``BackendPatternConfig`` to a dictionary with the items described in + :func:`~torch.ao.quantization.backend_config.BackendPatternConfig.from_dict`. + """ + backend_pattern_config_dict: Dict[str, Any] = { + OBSERVATION_TYPE_DICT_KEY: self.observation_type, + DTYPE_CONFIGS_DICT_KEY: [c.to_dict() for c in self.dtype_configs], + } + if self.pattern is not None: + backend_pattern_config_dict[PATTERN_DICT_KEY] = self.pattern + if self.root_module is not None: + backend_pattern_config_dict[ROOT_MODULE_DICT_KEY] = self.root_module + if self.qat_module is not None: + backend_pattern_config_dict[QAT_MODULE_DICT_KEY] = self.qat_module + if self.reference_quantized_module is not None: + backend_pattern_config_dict[REFERENCE_QUANTIZED_MODULE_DICT_KEY] = self.reference_quantized_module + if self.fused_module is not None: + backend_pattern_config_dict[FUSED_MODULE_DICT_KEY] = self.fused_module + if self.fuser_method is not None: + backend_pattern_config_dict[FUSER_METHOD_DICT_KEY] = self.fuser_method + if self._root_node_getter is not None: + backend_pattern_config_dict[ROOT_NODE_GETTER_DICT_KEY] = self._root_node_getter + if self._extra_inputs_getter is not None: + backend_pattern_config_dict[EXTRA_INPUTS_GETTER_DICT_KEY] = self._extra_inputs_getter + if len(self._num_tensor_args_to_observation_type) > 0: + backend_pattern_config_dict[NUM_TENSOR_ARGS_TO_OBSERVATION_TYPE_DICT_KEY] = self._num_tensor_args_to_observation_type + if len(self._input_type_to_index) > 0: + backend_pattern_config_dict[INPUT_TYPE_TO_INDEX_DICT_KEY] = self._input_type_to_index + if self._pattern_complex_format is not None: + backend_pattern_config_dict[PATTERN_COMPLEX_FORMAT_DICT_KEY] = self._pattern_complex_format + return backend_pattern_config_dict diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/onednn.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/onednn.py new file mode 100644 index 0000000000000000000000000000000000000000..8c14637ae3d3f72b0cb7d84c39d74af3ccb4bfca --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/onednn.py @@ -0,0 +1,542 @@ +import torch +import torch.nn as nn +import torch.ao.nn.intrinsic as nni +import torch.nn.functional as F +import torch.ao.nn.quantized.reference as nnqr +from ._common_operator_config_utils import ( + _get_conv_configs, + _get_linear_configs, + _get_binary_op_configs, + _get_bn_configs, + _get_cat_config, + _get_default_op_configs, + _get_embedding_op_configs, + _get_fixed_qparams_op_configs, + _get_ln_configs, + _get_rnn_op_configs, + _get_share_qparams_op_configs, +) +from .backend_config import ( + BackendPatternConfig, + BackendConfig, + DTypeConfig, + ObservationType, +) +from ..fuser_method_mappings import ( + _sequential_wrapper2, +) +import operator +from torch.ao.quantization.utils import MatchAllNode +import itertools + +# =================== +# | DTYPE CONFIGS | +# =================== + +onednn_weighted_op_int8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, + weight_dtype=torch.qint8, + bias_dtype=torch.float, +) + +onednn_op_quint8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, +) + +onednn_dynamic_int8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.float, + weight_dtype=torch.qint8, + bias_dtype=torch.float, + is_dynamic=True, +) + +onednn_weight_only_qint8_dtype_config = DTypeConfig( + input_dtype=torch.float, + output_dtype=torch.float, + weight_dtype=torch.qint8, +) + +onednn_input_output_only_quint8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, + weight_dtype=torch.float, + bias_dtype=torch.float, +) + +# =================== +# | FUSER METHODS | +# =================== + +def _fuse_linear_bn_leaky_relu(is_qat, linear, bn, leaky_relu): + r"""Given the linear, bn and leaky_relu modules, fuses them and returns the fused module + Args: + is_qat: a flag for whether we are using quantization aware training fusion + or post training quantization fusion + linear: Module instance of type Linear + bn: BatchNorm1d instance that needs to be fused with the linear layer + leaky_relu: LeakyReLU instance that needs to be fused with the linear layer + Examples:: + >>> # xdoctest: +SKIP(failing) + >>> m1 = nn.Linear(20, 10) + >>> b1 = nn.BatchNorm1d(10) + >>> lr = nn.LeakyReLU(0.01) + >>> m2 = _fuse_linear_bn_leaky_relu(m1, b1, lr) + """ + assert(linear.training == bn.training and bn.training == leaky_relu.training),\ + "Linear, BN and LeakyReLU all must be in the same mode (train or eval)." + + if is_qat: + raise NotImplementedError(f"Cannot fuse train modules: {(linear, bn, leaky_relu)}") + else: + map_to_fused_module_eval = { + nn.Linear: nni.LinearLeakyReLU, + } + fused_module = map_to_fused_module_eval.get(type(linear), None) + if fused_module is not None: + fused_linear = nn.utils.fusion.fuse_linear_bn_eval(linear, bn) + fm = fused_module(fused_linear, leaky_relu) + return fm + else: + raise NotImplementedError(f"Cannot fuse eval modules: {(linear, bn, leaky_relu)}") + +# ====================== +# | CONFIGS FOR CONV | +# ====================== +observation_type = ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT + +conv_dtype_configs = [onednn_weighted_op_int8_dtype_config] +conv_configs = _get_conv_configs(conv_dtype_configs) + +# (1) Conv2d + Add + +# conv2d Y +# \ / +# add + +# include: +# conv2d conv2d +# \ / +# add + +def _fuse_conv_add_left(is_qat, add, conv, _): + return nni.ConvAdd2d(conv, add) + +def _conv_add_root_node_getter_left(pattern): + _, conv, _ = pattern + return conv + +def _conv_add_extra_inputs_getter_left(pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + _, conv, extra_input = pattern + return [extra_input] + +# conv2d +# \ +# bn Y +# \ / +# add + +def _fuse_conv_bn_add_left(is_qat, add, bn_conv, _): + bn, conv = bn_conv + if is_qat: + raise NotImplementedError(f"Cannot fuse train modules: {(conv, bn, add)}") + else: + fused_conv = nn.utils.fusion.fuse_conv_bn_eval(conv, bn) + return nni.ConvAdd2d(fused_conv, add) + +def _conv_bn_add_root_node_getter_left(add_pattern): + _, bn_conv, _ = add_pattern + bn, conv = bn_conv + return conv + +def _conv_bn_add_extra_inputs_getter_left(add_pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + _, bn_conv, extra_input = add_pattern + bn, conv = bn_conv + return [extra_input] + +conv_add_left_optioins = itertools.product( + [True, False], # with_bn + [torch.add, operator.add], # add_op +) + +for with_bn, add_op in conv_add_left_optioins: + if with_bn: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((add_op, (nn.BatchNorm2d, nn.Conv2d), MatchAllNode)) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_bn_add_left) + ._set_root_node_getter(_conv_bn_add_root_node_getter_left) + ._set_extra_inputs_getter(_conv_bn_add_extra_inputs_getter_left) + .set_fused_module(nni.ConvAdd2d)) + else: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((add_op, nn.Conv2d, MatchAllNode)) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_add_left) + ._set_root_node_getter(_conv_add_root_node_getter_left) + ._set_extra_inputs_getter(_conv_add_extra_inputs_getter_left) + .set_fused_module(nni.ConvAdd2d)) + +# Y conv2d +# \ / +# add + +def _fuse_conv_add_right(is_qat, add, _, conv): + return nni.ConvAdd2d(conv, add) + +def _conv_add_root_node_getter_right(pattern): + add, _, conv = pattern + return conv + +def _conv_add_extra_inputs_getter_right(pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + _, extra_input, conv = pattern + return [extra_input] + +# conv2d +# / +# Y bn +# \ / +# add + +def _fuse_conv_bn_add_right(is_qat, add, _, bn_conv): + bn, conv = bn_conv + if is_qat: + raise NotImplementedError(f"Cannot fuse train modules: {(conv, bn, add)}") + else: + fused_conv = nn.utils.fusion.fuse_conv_bn_eval(conv, bn) + return nni.ConvAdd2d(fused_conv, add) + +def _conv_bn_add_root_node_getter_right(pattern): + add, _, bn_conv = pattern + bn, conv = bn_conv + return conv + +def _conv_bn_add_extra_inputs_getter_right(pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + _, extra_input, bn_conv = pattern + bn, conv = bn_conv + return [extra_input] + +conv_add_optioins = itertools.product( + [True, False], # with_bn + [torch.add, operator.add], # add_op +) + +for with_bn, add_op in conv_add_optioins: + if with_bn: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((add_op, MatchAllNode, (nn.BatchNorm2d, nn.Conv2d))) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_bn_add_right) + ._set_root_node_getter(_conv_bn_add_root_node_getter_right) + ._set_extra_inputs_getter(_conv_bn_add_extra_inputs_getter_right) + .set_fused_module(nni.ConvAdd2d)) + else: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((add_op, MatchAllNode, nn.Conv2d)) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_add_right) + ._set_root_node_getter(_conv_add_root_node_getter_right) + ._set_extra_inputs_getter(_conv_add_extra_inputs_getter_right) + .set_fused_module(nni.ConvAdd2d)) + +conv_configs.append( + BackendPatternConfig(nni.ConvAdd2d) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(conv_dtype_configs) + .set_root_module(nn.Conv2d) + .set_reference_quantized_module(nnqr.Conv2d)) + +# (2) Conv2d + Add + Relu + +# conv2d Y +# \ / +# add +# \ +# relu + +def _fuse_conv_add_relu_left(is_qat, relu, add_pattern): + add, conv, _ = add_pattern + return nni.ConvAddReLU2d(conv, add, relu) + +def _conv_add_relu_root_node_getter_left(pattern): + relu, add_pattern = pattern + _, conv, _ = add_pattern + return conv + +def _conv_add_relu_extra_inputs_getter_left(pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + relu, add_pattern = pattern + _, conv, extra_input = add_pattern + return [extra_input] + +# conv2d +# \ +# bn Y +# \ / +# add +# \ +# relu + +def _fuse_conv_bn_add_relu_left(is_qat, relu, add_pattern): + add, bn_conv, _ = add_pattern + bn, conv = bn_conv + if is_qat: + raise NotImplementedError(f"Cannot fuse train modules: {(conv, bn, add, relu)}") + else: + fused_conv = nn.utils.fusion.fuse_conv_bn_eval(conv, bn) + return nni.ConvAddReLU2d(fused_conv, add, relu) + +def _conv_bn_add_relu_root_node_getter_left(pattern): + relu, add_pattern = pattern + _, bn_conv, _ = add_pattern + bn, conv = bn_conv + return conv + +def _conv_bn_add_relu_extra_inputs_getter_left(pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + relu, add_pattern = pattern + _, bn_conv, extra_input = add_pattern + bn, conv = bn_conv + return [extra_input] + +conv_add_relu_left_optioins = itertools.product( + [True, False], # with_bn + [torch.add, operator.add], # add_op +) + +for with_bn, add_op in conv_add_relu_left_optioins: + if with_bn: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((nn.ReLU, (add_op, (nn.BatchNorm2d, nn.Conv2d), MatchAllNode))) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_bn_add_relu_left) + ._set_root_node_getter(_conv_bn_add_relu_root_node_getter_left) + ._set_extra_inputs_getter(_conv_bn_add_relu_extra_inputs_getter_left) + .set_fused_module(nni.ConvAddReLU2d)) + else: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((nn.ReLU, (add_op, nn.Conv2d, MatchAllNode))) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_add_relu_left) + ._set_root_node_getter(_conv_add_relu_root_node_getter_left) + ._set_extra_inputs_getter(_conv_add_relu_extra_inputs_getter_left) + .set_fused_module(nni.ConvAddReLU2d)) + +# Y conv2d +# \ / +# add +# \ +# relu + +def _fuse_conv_add_relu_right(is_qat, relu, add_pattern): + add, _, conv = add_pattern + return nni.ConvAddReLU2d(conv, add, relu) + +def _conv_add_relu_root_node_getter_right(pattern): + relu, add_pattern = pattern + _, _, conv = add_pattern + return conv + +def _conv_add_relu_extra_inputs_getter_right(pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + relu, add_pattern = pattern + _, extra_input, conv = add_pattern + return [extra_input] + +# conv2d +# / +# Y bn +# \ / +# add +# \ +# relu + +def _fuse_conv_bn_add_relu_right(is_qat, relu, add_pattern): + add, _, bn_conv = add_pattern + bn, conv = bn_conv + if is_qat: + raise NotImplementedError(f"Cannot fuse train modules: {(conv, bn, add, relu)}") + else: + fused_conv = nn.utils.fusion.fuse_conv_bn_eval(conv, bn) + return nni.ConvAddReLU2d(fused_conv, add, relu) + +def _conv_bn_add_relu_root_node_getter_right(pattern): + relu, add_pattern = pattern + _, _, bn_conv = add_pattern + bn, conv = bn_conv + return conv + +def _conv_bn_add_relu_extra_inputs_getter_right(pattern): + """ get inputs pattern for extra inputs, inputs for root node + are assumed to be copied over from root node to the fused node + """ + relu, add_pattern = pattern + _, extra_input, bn_conv = add_pattern + bn, conv = bn_conv + return [extra_input] + +conv_add_relu_optioins = itertools.product( + [True, False], # with_bn + [torch.add, operator.add], # add_op +) + +for with_bn, add_op in conv_add_relu_optioins: + if with_bn: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((nn.ReLU, (add_op, MatchAllNode, (nn.BatchNorm2d, nn.Conv2d)))) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_bn_add_relu_right) + ._set_root_node_getter(_conv_bn_add_relu_root_node_getter_right) + ._set_extra_inputs_getter(_conv_bn_add_relu_extra_inputs_getter_right) + .set_fused_module(nni.ConvAddReLU2d)) + else: + conv_configs.append( + BackendPatternConfig() + ._set_pattern_complex_format((nn.ReLU, (add_op, MatchAllNode, nn.Conv2d))) # noqa: E131 + .set_observation_type(observation_type) + .set_dtype_configs(conv_dtype_configs) + .set_fuser_method(_fuse_conv_add_relu_right) + ._set_root_node_getter(_conv_add_relu_root_node_getter_right) + ._set_extra_inputs_getter(_conv_add_relu_extra_inputs_getter_right) + .set_fused_module(nni.ConvAddReLU2d)) + +conv_configs.append( + BackendPatternConfig(nni.ConvAddReLU2d) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(conv_dtype_configs) + .set_root_module(nn.Conv2d) + .set_reference_quantized_module(nnqr.Conv2d)) + +# ======================== +# | CONFIGS FOR LINEAR | +# ======================== + +linear_dtype_configs = [ + onednn_weighted_op_int8_dtype_config, + onednn_dynamic_int8_dtype_config, +] +linear_configs = _get_linear_configs(linear_dtype_configs) + +def _add_eltwise_fusion_configs(configs, root_module, root_op, post_module, post_op, + dtype_configs, fuser_method, fused_module, observation_type, + ref_quant_module): + # 1 base module + op module fusion config + configs.append( + BackendPatternConfig((root_module, post_module)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(fuser_method) + .set_fused_module(fused_module)) + # base module + functional post op + configs.append( + BackendPatternConfig((root_module, post_op)) + .set_dtype_configs(dtype_configs) # noqa: E131 + .set_fuser_method(fuser_method) + .set_fused_module(fused_module)) + + # 2 fused module configs + configs.append( + BackendPatternConfig(fused_module) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs) + .set_root_module(root_module) + .set_reference_quantized_module(ref_quant_module)) + + # 3 functional base op + post op configs + configs.append( + BackendPatternConfig((root_op, post_module)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs)) + configs.append( + BackendPatternConfig((root_op, post_op)) + .set_observation_type(observation_type) # noqa: E131 + .set_dtype_configs(dtype_configs)) + +# Configs for linear + leaky_relu fusion +_add_eltwise_fusion_configs(linear_configs, nn.Linear, F.linear, + nn.LeakyReLU, F.leaky_relu, linear_dtype_configs, + _sequential_wrapper2(nni.LinearLeakyReLU), + nni.LinearLeakyReLU, observation_type, nnqr.Linear) + +# Configs for linear module + batchnorm + leaky_relu +linear_configs.append( + BackendPatternConfig((nn.Linear, nn.BatchNorm1d, nn.LeakyReLU)) + .set_dtype_configs(linear_dtype_configs) # noqa: E131 + .set_fuser_method(_fuse_linear_bn_leaky_relu) + .set_fused_module(nni.LinearLeakyReLU)) + +# Configs for linear + tanh fusion +_add_eltwise_fusion_configs(linear_configs, nn.Linear, F.linear, + nn.Tanh, torch.tanh, linear_dtype_configs, + _sequential_wrapper2(nni.LinearTanh), + nni.LinearTanh, observation_type, nnqr.Linear) + +# =========================== +# | CONFIGS FOR OTHER OPS | +# =========================== + +binary_op_dtype_configs = [onednn_op_quint8_dtype_config] +default_op_dtype_configs = [onednn_op_quint8_dtype_config] +fixed_qparams_op_dtype_configs = [onednn_op_quint8_dtype_config] +share_qparams_op_dtype_configs = [onednn_op_quint8_dtype_config] +rnn_op_dtype_configs = [onednn_dynamic_int8_dtype_config] +embedding_op_dtype_configs = [onednn_weight_only_qint8_dtype_config] +layer_norm_op_dtype_configs = [onednn_input_output_only_quint8_dtype_config] + +# ===================== +# | BACKEND CONFIGS | +# ===================== + +def get_onednn_backend_config() -> BackendConfig: + """ + Return the `BackendConfig` for PyTorch's native ONEDNN backend. + """ + return BackendConfig("onednn") \ + .set_backend_pattern_configs(conv_configs) \ + .set_backend_pattern_configs(linear_configs) \ + .set_backend_pattern_configs(_get_binary_op_configs(binary_op_dtype_configs)) \ + .set_backend_pattern_config(_get_cat_config(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_default_op_configs(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_fixed_qparams_op_configs(fixed_qparams_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_share_qparams_op_configs(share_qparams_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_bn_configs(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_ln_configs(layer_norm_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_rnn_op_configs(rnn_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_embedding_op_configs(embedding_op_dtype_configs)) + +__all__ = [ + "get_onednn_backend_config", +] diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/qnnpack.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/qnnpack.py new file mode 100644 index 0000000000000000000000000000000000000000..772a25c65574481d70186e9d968039756b2fa0ae --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/qnnpack.py @@ -0,0 +1,160 @@ +import torch +from ._common_operator_config_utils import ( + _get_binary_op_configs, + _get_bn_configs, + _get_cat_config, + _get_conv_configs, + _get_default_op_configs, + _get_embedding_op_configs, + _get_fixed_qparams_op_configs, + _get_linear_configs, + _get_rnn_op_configs, + _get_share_qparams_op_configs, +) +from .backend_config import BackendConfig, DTypeConfig, DTypeWithConstraints + +__all__ = [ + "get_qnnpack_backend_config", +] + +# =================== +# | DTYPE CONFIGS | +# =================== + +qnnpack_weighted_op_quint8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, + weight_dtype=torch.qint8, + bias_dtype=torch.float, +) + +qnnpack_default_op_quint8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, +) + +qnnpack_default_op_fp16_dtype_config = DTypeConfig( + input_dtype=torch.float16, + output_dtype=torch.float16, + weight_dtype=torch.float16, + bias_dtype=torch.float16, +) + +qnnpack_default_dynamic_int8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.float, + weight_dtype=torch.qint8, + bias_dtype=torch.float, + is_dynamic=True, +) + +qnnpack_default_dynamic_float16_dtype_config = DTypeConfig( + input_dtype=torch.float16, + output_dtype=torch.float, + weight_dtype=torch.float16, + bias_dtype=torch.float, + is_dynamic=True, +) + +qnnpack_weight_only_quint8_dtype_config = DTypeConfig( + input_dtype=torch.float, + output_dtype=torch.float, + weight_dtype=torch.quint8, +) + +qnnpack_weight_only_quint4x2_dtype_config = DTypeConfig( + input_dtype=torch.float, + output_dtype=torch.float, + weight_dtype=torch.quint4x2, +) + +# xnnpack compatible dtype configs + +# We restrict scale values to be 2 ** -12 to ensure the +# requantization scale never falls below the xnnpack lower +# threshold. Additionally, for qint8 weight, we restrict +# the quantization values to [-127, +127], excluding -128. +# For more detail, refer to the description of +# `default_symmetric_qnnpack_qconfig`. + +# TODO: add additional restriction on qscheme to ensure it +# is either per_tensor_symmetric or per_channel_symmetric + +qnnpack_act_qint8_scale_min_2_neg_12 = DTypeWithConstraints( + dtype=torch.qint8, + scale_min_lower_bound=2 ** -12, +) + +qnnpack_weight_qint8_neg_127_to_127_scale_min_2_neg_12 = DTypeWithConstraints( + dtype=torch.qint8, + quant_min_lower_bound=-127, + quant_max_upper_bound=127, + scale_min_lower_bound=2 ** -12, +) + +qnnpack_weighted_op_qint8_symmetric_dtype_config = DTypeConfig( + input_dtype=qnnpack_act_qint8_scale_min_2_neg_12, + output_dtype=qnnpack_act_qint8_scale_min_2_neg_12, + weight_dtype=qnnpack_weight_qint8_neg_127_to_127_scale_min_2_neg_12, + bias_dtype=torch.float, +) + +qnnpack_default_op_qint8_symmetric_dtype_config = DTypeConfig( + input_dtype=qnnpack_act_qint8_scale_min_2_neg_12, + output_dtype=qnnpack_act_qint8_scale_min_2_neg_12, +) + + +# ===================== +# | BACKEND CONFIGS | +# ===================== + +def get_qnnpack_backend_config() -> BackendConfig: + """ + Return the `BackendConfig` for PyTorch's native QNNPACK backend. + """ + conv_dtype_configs = [ + qnnpack_weighted_op_qint8_symmetric_dtype_config, + qnnpack_weighted_op_quint8_dtype_config, + ] + linear_dtype_configs = [ + qnnpack_weighted_op_qint8_symmetric_dtype_config, + qnnpack_weighted_op_quint8_dtype_config, + qnnpack_default_dynamic_int8_dtype_config, + qnnpack_default_dynamic_float16_dtype_config, + ] + binary_op_dtype_configs = [ + qnnpack_default_op_qint8_symmetric_dtype_config, + qnnpack_default_op_quint8_dtype_config, + ] + default_op_dtype_configs = [ + qnnpack_default_op_qint8_symmetric_dtype_config, + qnnpack_default_op_quint8_dtype_config, + ] + fixed_qparams_op_dtype_configs = [ + qnnpack_default_op_qint8_symmetric_dtype_config, + qnnpack_default_op_quint8_dtype_config, + ] + share_qparams_op_dtype_configs = [ + qnnpack_default_op_qint8_symmetric_dtype_config, + qnnpack_default_op_quint8_dtype_config, + ] + rnn_op_dtype_configs = [ + qnnpack_default_dynamic_int8_dtype_config, + qnnpack_default_dynamic_float16_dtype_config, + ] + embedding_op_dtype_configs = [ + qnnpack_weight_only_quint8_dtype_config, + qnnpack_weight_only_quint4x2_dtype_config, + ] + return BackendConfig("qnnpack") \ + .set_backend_pattern_configs(_get_conv_configs(conv_dtype_configs)) \ + .set_backend_pattern_configs(_get_linear_configs(linear_dtype_configs)) \ + .set_backend_pattern_configs(_get_binary_op_configs(binary_op_dtype_configs)) \ + .set_backend_pattern_config(_get_cat_config(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_default_op_configs(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_fixed_qparams_op_configs(fixed_qparams_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_share_qparams_op_configs(share_qparams_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_bn_configs(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_rnn_op_configs(rnn_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_embedding_op_configs(embedding_op_dtype_configs)) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/x86.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/x86.py new file mode 100644 index 0000000000000000000000000000000000000000..b4f165958f2791d3e6e2f63eceecdcd9e6f6d50c --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/backend_config/x86.py @@ -0,0 +1,113 @@ +import torch +from ._common_operator_config_utils import ( + _get_binary_op_configs, + _get_bn_configs, + _get_cat_config, + _get_conv_configs, + _get_default_op_configs, + _get_embedding_op_configs, + _get_fixed_qparams_op_configs, + _get_linear_configs, + _get_rnn_op_configs, + _get_share_qparams_op_configs, + _get_tensor_info_op_configs, +) +from .backend_config import BackendConfig, DTypeConfig + +__all__ = [ + "get_x86_backend_config", +] + +# =================== +# | DTYPE CONFIGS | +# =================== + +# X86 aligns with FBGEMM for now + +x86_weighted_op_int8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, + weight_dtype=torch.qint8, + bias_dtype=torch.float, +) + +x86_default_op_quint8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.quint8, +) + +x86_default_op_fp16_dtype_config = DTypeConfig( + input_dtype=torch.float16, + output_dtype=torch.float16, + weight_dtype=torch.float16, + bias_dtype=torch.float16, +) + +x86_default_dynamic_int8_dtype_config = DTypeConfig( + input_dtype=torch.quint8, + output_dtype=torch.float, + weight_dtype=torch.qint8, + bias_dtype=torch.float, + is_dynamic=True, +) + +x86_default_dynamic_float16_dtype_config = DTypeConfig( + input_dtype=torch.float16, + output_dtype=torch.float, + weight_dtype=torch.float16, + bias_dtype=torch.float, + is_dynamic=True, +) + +x86_weight_only_quint8_dtype_config = DTypeConfig( + input_dtype=torch.float, + output_dtype=torch.float, + weight_dtype=torch.quint8, +) + +x86_weight_only_quint4x2_dtype_config = DTypeConfig( + input_dtype=torch.float, + output_dtype=torch.float, + weight_dtype=torch.quint4x2, +) + + +# ===================== +# | BACKEND CONFIGS | +# ===================== + +def get_x86_backend_config() -> BackendConfig: + """ + Return the `BackendConfig` for PyTorch's native x86 backend. + """ + conv_dtype_configs = [x86_weighted_op_int8_dtype_config] + linear_dtype_configs = [ + x86_weighted_op_int8_dtype_config, + x86_default_dynamic_int8_dtype_config, + x86_default_dynamic_float16_dtype_config, + ] + binary_op_dtype_configs = [x86_weighted_op_int8_dtype_config] + default_op_dtype_configs = [x86_default_op_quint8_dtype_config] + fixed_qparams_op_dtype_configs = [x86_weighted_op_int8_dtype_config] + share_qparams_op_dtype_configs = [x86_default_op_quint8_dtype_config] + tensor_info_op_dtype_configs = [x86_default_op_quint8_dtype_config] + rnn_op_dtype_configs = [ + x86_default_dynamic_int8_dtype_config, + x86_default_dynamic_float16_dtype_config, + ] + embedding_op_dtype_configs = [ + x86_weight_only_quint8_dtype_config, + x86_weight_only_quint4x2_dtype_config, + ] + return BackendConfig("x86") \ + .set_backend_pattern_configs(_get_conv_configs(conv_dtype_configs)) \ + .set_backend_pattern_configs(_get_linear_configs(linear_dtype_configs)) \ + .set_backend_pattern_configs(_get_binary_op_configs(binary_op_dtype_configs)) \ + .set_backend_pattern_config(_get_cat_config(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_default_op_configs(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_fixed_qparams_op_configs(fixed_qparams_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_share_qparams_op_configs(share_qparams_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_tensor_info_op_configs(tensor_info_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_bn_configs(default_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_rnn_op_configs(rnn_op_dtype_configs)) \ + .set_backend_pattern_configs(_get_embedding_op_configs(embedding_op_dtype_configs)) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/__init__.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/__init__.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..0def4a2a2ae2b5a1ad39cfa8cf410e25496cb45b Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/__init__.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_decomposed.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_decomposed.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..2ab5015e0176d759a2d68f01c65e9b6f42a43012 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_decomposed.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_equalize.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_equalize.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..d7f884667412cb0ac87af006b45426e2d66c5563 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_equalize.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_lower_to_native_backend.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_lower_to_native_backend.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..49a6efc6d997c1e1f316e08f1af66e132174af81 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/_lower_to_native_backend.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/convert.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/convert.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..83aba559597401cc2f54c7903604711072ae807f Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/convert.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/custom_config.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/custom_config.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..fff94a023cb8c7ba63d459da8e65acc267c30a98 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/custom_config.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/fuse.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/fuse.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..1917dcf578a7bd2e24d26571a19397a6bb4f0fde Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/fuse.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/fuse_handler.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/fuse_handler.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..90ef34a21ddbc77071c077a4b0fd8060df6c1c43 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/fuse_handler.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/graph_module.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/graph_module.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..fc8c196e18d22989dea286c0540bd0639ae0cbd3 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/graph_module.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lower_to_fbgemm.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lower_to_fbgemm.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..cb4dfe07fe46dfa43ec9cf673ddc54ba5016f055 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lower_to_fbgemm.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lower_to_qnnpack.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lower_to_qnnpack.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..4d718fe99f28aeca2c111fa057aa9ccf5ecaf65a Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lower_to_qnnpack.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lstm_utils.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lstm_utils.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..9614d36a571848fa915b8bd2f9bc448ac91b0516 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/lstm_utils.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/match_utils.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/match_utils.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..26015677c487d7eeb4603c5379742ccc61a26d60 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/match_utils.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/pattern_utils.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/pattern_utils.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..72fc3890816920341867b20b154cc27ee6ae4004 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/pattern_utils.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/prepare.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/prepare.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..37098d9c736012cc923cb78a77a92497309f9bdb Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/prepare.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/qconfig_mapping_utils.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/qconfig_mapping_utils.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..57b6529d4e72563567cfa4f65d3716378d6bf065 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/qconfig_mapping_utils.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/quantize_handler.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/quantize_handler.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..3b2bb33c29bf0b00beab6d7a69e7789ffefd3ffa Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/quantize_handler.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/tracer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/tracer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..069825873f82d473516e0d90db05936d648cfbb7 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/tracer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/utils.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/utils.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..350ea088045c16c6234fa6393978eab58bb391d9 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/__pycache__/utils.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_decomposed.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_decomposed.py new file mode 100644 index 0000000000000000000000000000000000000000..c64defa89f7004134f0701f6e27bcba92cefb530 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_decomposed.py @@ -0,0 +1,480 @@ +import torch +from torch.library import Library, impl +from torch.ao.quantization.utils import determine_qparams, validate_qmin_qmax +from typing import Tuple + + +# Note: decomposed means decomposed quantized tensor, using decomposed so that the +# name is not too long +quantized_decomposed_lib = Library("quantized_decomposed", "DEF") + +_DTYPE_TO_QVALUE_BOUNDS = { + torch.uint8: (0, 255), + torch.int8: (-128, 127), + torch.int16: (-(2**15), 2**15 - 1), + torch.int32: (-(2**31), 2**31 - 1) +} + +# Helper to check the passed in quant min and max are valid for the dtype +def _quant_min_max_bounds_check(quant_min, quant_max, dtype): + if dtype not in _DTYPE_TO_QVALUE_BOUNDS: + raise ValueError(f"Unsupported dtype: {dtype}") + quant_min_lower_bound, quant_max_upper_bound = _DTYPE_TO_QVALUE_BOUNDS[dtype] + + assert quant_min >= quant_min_lower_bound, \ + "quant_min out of bound for dtype, " \ + f"quant_min_lower_bound: {quant_min_lower_bound} quant_min: {quant_min}" + + assert quant_max <= quant_max_upper_bound, \ + "quant_max out of bound for dtype, " \ + f"quant_max_upper_bound: {quant_max_upper_bound} quant_max: {quant_max}" + +quantized_decomposed_lib.define( + "quantize_per_tensor(Tensor input, float scale, int zero_point, " + "int quant_min, int quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "quantize_per_tensor", "CompositeExplicitAutograd") +def quantize_per_tensor( + input: torch.Tensor, + scale: float, + zero_point: int, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine quantization for the Tensor using the same quantization parameters to map + from floating point to quantized values + + Args: + input (torch.Tensor): original float32 or bfloat16 Tensor + scale (float): quantization parameter for affine quantization + zero_point (int): quantization parameter for affine quantization + quant_min (int): minimum quantized value for output Tensor + quant_max (int): maximum quantized value for output Tensor + dtype (torch.dtype): requested dtype (e.g. torch.uint8) for output Tensor + + Returns: + Tensor with requested dtype (e.g. torch.uint8), note the quantization parameters + are not stored in the Tensor, we are storing them in function arguments instead + """ + if input.dtype == torch.bfloat16: + input = input.to(torch.float32) + + assert input.dtype == torch.float32, f"Expecting input to have dtype torch.float32, but got dtype: {input.dtype}" + _quant_min_max_bounds_check(quant_min, quant_max, dtype) + + inv_scale = 1.0 / scale + return torch.clamp(torch.round(input * inv_scale) + zero_point, quant_min, quant_max).to(dtype) + +quantized_decomposed_lib.define( + "quantize_per_tensor.tensor(Tensor input, Tensor scale, Tensor zero_point, " + "int quant_min, int quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "quantize_per_tensor.tensor", "CompositeExplicitAutograd") +def quantize_per_tensor_tensor( + input: torch.Tensor, + scale: torch.Tensor, + zero_point: torch.Tensor, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine quantization for the Tensor using the same quantization parameters to map + from floating point to quantized values + Same as `quantize_per_tensor` but scale and zero_point are Scalar Tensor instead of + scalar values + """ + assert zero_point.numel() == 1, f"Expecting zero_point tensor to be one element, but received : {zero_point.numel()}" + assert scale.numel() == 1, f"Expecting scale tensor to be one element, but received : {scale.numel()}" + return quantize_per_tensor(input, scale.item(), zero_point.item(), quant_min, quant_max, dtype) + +@impl(quantized_decomposed_lib, "quantize_per_tensor.tensor", "Meta") +def quantize_per_tensor_tensor_meta(input, scale, zero_point, quant_min, quant_max, dtype): + assert zero_point.numel() == 1, f"Expecting zero_point tensor to be one element, but received : {zero_point.numel()}" + assert scale.numel() == 1, f"Expecting scale tensor to be one element, but received : {scale.numel()}" + assert input.dtype == torch.float32, f"Expecting input to have dtype torch.float32, but got dtype: {input.dtype}" + return torch.empty_like(input, dtype=dtype) + +# TODO: remove other variants and keep this one +quantized_decomposed_lib.define( + "quantize_per_tensor.tensor2(Tensor input, Tensor scale, Tensor zero_point, " + "Tensor quant_min, Tensor quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "quantize_per_tensor.tensor2", "CompositeExplicitAutograd") +def quantize_per_tensor_tensor2( + input: torch.Tensor, + scale: torch.Tensor, + zero_point: torch.Tensor, + quant_min: torch.Tensor, + quant_max: torch.Tensor, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine quantization for the Tensor using the same quantization parameters to map + from floating point to quantized values + Same as `quantize_per_tensor` but scale and zero_point are Scalar Tensor instead of + scalar values + """ + assert zero_point.numel() == 1, f"Expecting zero_point tensor to be one element, but received : {zero_point.numel()}" + assert scale.numel() == 1, f"Expecting scale tensor to be one element, but received : {scale.numel()}" + return quantize_per_tensor(input, scale.item(), zero_point.item(), quant_min.item(), quant_max.item(), dtype) + +@impl(quantized_decomposed_lib, "quantize_per_tensor.tensor2", "Meta") +def quantize_per_tensor_tensor2_meta(input, scale, zero_point, quant_min, quant_max, dtype): + return quantize_per_tensor_tensor_meta(input, scale, zero_point, quant_min, quant_max, dtype) + +# Note: quant_min/quant_max/dtype are not used in the operator, but for now it's kept in +# the signature as metadata for the input Tensor, this might be useful for pattern +# matching in the future +# We will revisit this later if we found there are no use cases for it +quantized_decomposed_lib.define( + "dequantize_per_tensor(Tensor input, float scale, int zero_point, " + "int quant_min, int quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "dequantize_per_tensor", "CompositeExplicitAutograd") +def dequantize_per_tensor( + input: torch.Tensor, + scale: float, + zero_point: int, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine dequantization for the Tensor using the same quantization parameters to map + from quantized values to floating point values + + Args: + input (torch.Tensor): Tensor with dtype matching `dtype` argument, + e.g. (`torch.uint8`), it is a per tensor quantized Tensor if combined with + quantization parameters in the argument of this function (scale/zero_point) + + scale (float): quantization parameter for affine quantization + + zero_point (int): quantization parameter for affine quantization + + quant_min (int): minimum quantized value for input Tensor (not used in computation, + reserved for pattern matching) + + quant_max (int): maximum quantized value for input Tensor (not used in computation, + reserved for pattern matching) + + dtype (torch.dtype): dtype for input Tensor (not used in computation, + reserved for pattern matching) + + Returns: + dequantized float32 Tensor + """ + assert input.dtype == dtype, f"Expecting input to have dtype: {dtype}, but got {input.dtype}" + if dtype in _DTYPE_TO_QVALUE_BOUNDS: + # TODO: investigate why + # (input - zero_point).to(torch.float32) * scale + # failed the test + return (input.to(torch.float32) - zero_point) * scale + else: + raise ValueError(f"Unsupported dtype in dequantize_per_tensor: {dtype}") + + +quantized_decomposed_lib.define( + "dequantize_per_tensor.tensor(Tensor input, Tensor scale, Tensor zero_point, " + "int quant_min, int quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "dequantize_per_tensor.tensor", "CompositeExplicitAutograd") +def dequantize_per_tensor_tensor( + input: torch.Tensor, + scale: torch.Tensor, + zero_point: torch.Tensor, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine dequantization for the Tensor using the same quantization parameters to map + from quantized values to floating point values + Same as `dequantize_per_tensor` but scale and zero_point are Scalar Tensor instead of + scalar values + """ + assert zero_point.numel() == 1, f"Expecting zero_point tensor to be one element, but received : {zero_point.numel()}" + assert scale.numel() == 1, f"Expecting scale tensor to be one element, but received : {scale.numel()}" + return dequantize_per_tensor(input, scale.item(), zero_point.item(), quant_min, quant_max, dtype) + +@impl(quantized_decomposed_lib, "dequantize_per_tensor.tensor", "Meta") +def dequantize_per_tensor_tensor_meta(input, scale, zero_point, quant_min, quant_max, dtype): + assert zero_point.numel() == 1, f"Expecting zero_point tensor to be one element, but received : {zero_point.numel()}" + assert scale.numel() == 1, f"Expecting scale tensor to be one element, but received : {scale.numel()}" + assert input.dtype == dtype, f"Expecting input to have dtype: {dtype}" + if dtype in _DTYPE_TO_QVALUE_BOUNDS: + return torch.empty_like(input, dtype=torch.float32) + else: + raise ValueError(f"Unsupported dtype in dequantize_per_tensor: {dtype}") + +# TODO: remove other variants and keep this one +quantized_decomposed_lib.define( + "dequantize_per_tensor.tensor2(Tensor input, Tensor scale, Tensor zero_point, " + "Tensor quant_min, Tensor quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "dequantize_per_tensor.tensor2", "CompositeExplicitAutograd") +def dequantize_per_tensor_tensor2( + input: torch.Tensor, + scale: torch.Tensor, + zero_point: torch.Tensor, + quant_min: torch.Tensor, + quant_max: torch.Tensor, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine dequantization for the Tensor using the same quantization parameters to map + from quantized values to floating point values + Same as `dequantize_per_tensor` but scale and zero_point are Scalar Tensor instead of + scalar values + """ + assert zero_point.numel() == 1, f"Expecting zero_point tensor to be one element, but received : {zero_point.numel()}" + assert scale.numel() == 1, f"Expecting scale tensor to be one element, but received : {scale.numel()}" + return dequantize_per_tensor(input, scale.item(), zero_point.item(), quant_min.item(), quant_max.item(), dtype) + +@impl(quantized_decomposed_lib, "dequantize_per_tensor.tensor2", "Meta") +def dequantize_per_tensor_tensor2_meta(input, scale, zero_point, quant_min, quant_max, dtype): + return dequantize_per_tensor_tensor_meta(input, scale, zero_point, quant_min, quant_max, dtype) + +quantized_decomposed_lib.define( + "choose_qparams.tensor(Tensor input, int quant_min, int quant_max, " + "float eps, ScalarType dtype) -> (Tensor, Tensor)") + +@impl(quantized_decomposed_lib, "choose_qparams.tensor", "CompositeExplicitAutograd") +def choose_qparams_tensor( + input: torch.Tensor, + qmin: int, + qmax: int, + eps: float, + dtype: torch.dtype +) -> Tuple[torch.Tensor, torch.Tensor]: + """ Given an input Tensor, derive the per tensor affine quantization parameter + (scale and zero_point) for target quantized Tensor from the Tensor + + Args: + input (torch.Tensor): floating point input Tensor + quant_min (int): minimum quantized value for target quantized Tensor + quant_max (int): maximum quantized value for target quantized Tensor + dtype (torch.dtype): dtype for target quantized Tensor + + Returns: + scale (float): quantization parameter for the target quantized Tensor + zero_point (int): quantization parameter for the target quantized Tensor + """ + assert input.dtype == torch.float32, f"Expecting input to have dtype torch.float32, but got dtype: {input.dtype}" + assert dtype in _DTYPE_TO_QVALUE_BOUNDS, \ + f"Expecting target dtype to be one of {_DTYPE_TO_QVALUE_BOUNDS.keys()}, but got: {dtype}" + validate_qmin_qmax(qmin, qmax) + + min_val, max_val = torch.aminmax(input) + + return determine_qparams( + min_val, max_val, qmin, qmax, dtype, torch.Tensor([eps]), has_customized_qrange=False) + +quantized_decomposed_lib.define( + "choose_qparams_symmetric.tensor(Tensor input, int quant_min, int quant_max, " + "float eps, ScalarType dtype) -> (Tensor, Tensor)") + +@impl(quantized_decomposed_lib, "choose_qparams_symmetric.tensor", "CompositeExplicitAutograd") +def choose_qparams_symmetric_tensor( + input: torch.Tensor, + qmin: int, + qmax: int, + eps: float, + dtype: torch.dtype +) -> Tuple[torch.Tensor, torch.Tensor]: + """ Given an input Tensor, derive the per tensor affine quantization parameter + (scale and zero_point) for target quantized Tensor from the Tensor + + Args: + input (torch.Tensor): floating point input Tensor + quant_min (int): minimum quantized value for target quantized Tensor + quant_max (int): maximum quantized value for target quantized Tensor + dtype (torch.dtype): dtype for target quantized Tensor + + Returns: + scale (float): quantization parameter for the target quantized Tensor + zero_point (int): quantization parameter for the target quantized Tensor + """ + assert input.dtype == torch.float32, f"Expecting input to have dtype torch.float32, but got dtype: {input.dtype}" + assert dtype in _DTYPE_TO_QVALUE_BOUNDS, \ + f"Expecting target dtype to be one of {_DTYPE_TO_QVALUE_BOUNDS.keys()}, but got: {dtype}" + validate_qmin_qmax(qmin, qmax) + + min_val, max_val = torch.aminmax(input) + return determine_qparams( + min_val, + max_val, + qmin, + qmax, + dtype, + torch.Tensor([eps]), + has_customized_qrange=False, + qscheme=torch.per_tensor_symmetric + ) + +@impl(quantized_decomposed_lib, "choose_qparams.tensor", "Meta") +def choose_qparams_tensor_meta( + input: torch.Tensor, + quant_min: int, + quant_max: int, + eps: float, + dtype: torch.dtype +) -> Tuple[torch.Tensor, torch.Tensor]: + assert input.dtype == torch.float32, f"Expecting input to have dtype torch.float32, but got dtype: {input.dtype}" + assert quant_min < quant_max, f"Expecting quant_min to be smaller than quant_max but received min: \ + {quant_min} max: {quant_max}" + return torch.empty(1, dtype=torch.double, device=input.device), torch.empty(1, dtype=torch.int64, device=input.device) + +@impl(quantized_decomposed_lib, "choose_qparams_symmetric.tensor", "Meta") +def choose_qparams_symmetric_tensor_meta( + input: torch.Tensor, + quant_min: int, + quant_max: int, + eps: float, + dtype: torch.dtype +) -> Tuple[torch.Tensor, torch.Tensor]: + return torch.empty(1, dtype=torch.double, device=input.device), torch.empty(1, dtype=torch.int64, device=input.device) + +# Helper function used to implement per-channel quantization against any axis +def _permute_to_axis_zero(x, axis): + new_axis_list = list(range(x.dim())) + new_axis_list[axis] = 0 + new_axis_list[0] = axis + y = x.permute(tuple(new_axis_list)) + return y, new_axis_list + +quantized_decomposed_lib.define( + "quantize_per_channel(Tensor input, Tensor scales, Tensor zero_points, int axis, " + "int quant_min, int quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "quantize_per_channel", "CompositeExplicitAutograd") +def quantize_per_channel( + input: torch.Tensor, + scales: torch.Tensor, + zero_points: torch.Tensor, + axis: int, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine per channel quantization for the Tensor using the same quantization + parameters for each channel/axis to map from floating point to quantized values + + Args: + input (torch.Tensor): original float32 or bfloat16 Tensor + scales (torch.Tensor): a list of scale quantization parameter for + affine quantization, one per channel + zero_point (torch.Tensor): a list of zero_point quantization parameter for + affine quantization, one per channel + quant_min (int): minimum quantized value for output Tensor + quant_max (int): maximum quantized value for output Tensor + dtype (torch.dtype): requested dtype (e.g. torch.uint8) for output Tensor + + Returns: + Tensor with requested dtype (e.g. torch.uint8), note the quantization parameters + are not stored in the Tensor, we are storing them in function arguments instead + """ + if input.dtype == torch.bfloat16: + input = input.to(torch.float32) + + assert input.dtype == torch.float32, f"Expecting input to have dtype torch.float32, but got dtype: {input.dtype}" + assert axis < input.dim(), f"Expecting axis to be < {input.dim()}" + _quant_min_max_bounds_check(quant_min, quant_max, dtype) + input, permute_axis_list = _permute_to_axis_zero(input, axis) + res = torch.zeros_like(input) + + for i in range(input.size(0)): + res[i] = torch.clamp( + torch.round(input[i] * (1.0 / scales[i])) + zero_points[i], + quant_min, + quant_max + ) + + out = res.permute(tuple(permute_axis_list)) + return out.to(dtype) + +@impl(quantized_decomposed_lib, "quantize_per_channel", "Meta") +def quantize_per_channel_meta( + input: torch.Tensor, + scales: torch.Tensor, + zero_points: torch.Tensor, + axis: int, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + assert input.dtype == torch.float32, f"Expecting input to have dtype torch.float32, but got dtype: {input.dtype}" + assert axis < input.dim(), f"Expecting axis to be < {input.dim()}" + _quant_min_max_bounds_check(quant_min, quant_max, dtype) + return torch.empty_like(input, dtype=dtype) + +# Note: quant_min/quant_max/dtype are not used in the operator, but for now it's kept in +# the signature as metadata for the input Tensor, this might be useful for pattern +# matching in the future +# We will revisit this later if we found there are no use cases for it +quantized_decomposed_lib.define( + "dequantize_per_channel(Tensor input, Tensor scales, Tensor zero_points, int axis, " + "int quant_min, int quant_max, ScalarType dtype) -> Tensor") + +@impl(quantized_decomposed_lib, "dequantize_per_channel", "CompositeExplicitAutograd") +def dequantize_per_channel( + input: torch.Tensor, + scales: torch.Tensor, + zero_points: torch.Tensor, + axis: int, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + """ Affine per channel dequantization for the Tensor using the same quantization + parameters for each channel/axis to map from quantized values to floating point values + + Args: + input (torch.Tensor): Tensor with dtype matching `dtype` argument, + e.g. (`torch.uint8`), it is a per channel quantized Tensor if combined with + quantization parameter in the argument of this function (scales/zero_points/axis) + + scales (torch.Tensor): a list of scale quantization parameter for + affine quantization, one per channel + + zero_points (torch.Tensor): a list of zero_point quantization parameter for + affine quantization, one per channel + + quant_min (int): minimum quantized value for output Tensor (not used in computation, + reserved for pattern matching) + + quant_max (int): maximum quantized value for output Tensor (not used in computation, + reserved for pattern matching) + + dtype (torch.dtype): requested dtype for output Tensor (not used in computation, + reserved for pattern matching) + + Returns: + dequantized float32 Tensor + """ + assert input.dtype == dtype, f"Expecting input to have dtype {dtype}, but got dtype: {input.dtype}" + assert axis < input.dim(), f"Expecting axis to be < {input.dim()}" + _quant_min_max_bounds_check(quant_min, quant_max, dtype) + input, permute_axis_list = _permute_to_axis_zero(input, axis) + res = torch.zeros_like(input, dtype=torch.float32) + + for i in range(input.size(0)): + # TODO: investigate why + # (input[i] - zero_points[i]).to(torch.float32) * scales[i] + # failed the test + res[i] = (input[i].to(torch.float32) - zero_points[i]) * scales[i] + + out = res.permute(tuple(permute_axis_list)) + return out + +@impl(quantized_decomposed_lib, "dequantize_per_channel", "Meta") +def dequantize_per_channel_meta( + input: torch.Tensor, + scales: torch.Tensor, + zero_points: torch.Tensor, + axis: int, + quant_min: int, + quant_max: int, + dtype: torch.dtype +) -> torch.Tensor: + assert input.dtype == dtype, f"Expecting input to have dtype {dtype}, but got dtype: {input.dtype}" + assert axis < input.dim(), f"Expecting axis to be < {input.dim()}" + _quant_min_max_bounds_check(quant_min, quant_max, dtype) + return torch.empty_like(input, dtype=torch.float32) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_lower_to_native_backend.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_lower_to_native_backend.py new file mode 100644 index 0000000000000000000000000000000000000000..e769b6c52655035840267be8bf41e3db8f0b2b5a --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_lower_to_native_backend.py @@ -0,0 +1,1168 @@ +import torch +from torch.fx import map_arg, Node +from torch.fx.graph import Graph +import torch.nn as nn +import torch.nn.functional as F +import torch.ao.nn.intrinsic as nni +import torch.ao.nn.intrinsic.quantized as nniq +import torch.ao.nn.intrinsic.quantized.dynamic as nniqd +import torch.ao.nn.quantized as nnq +import torch.ao.nn.quantized.dynamic as nnqd +import torch.ao.nn.quantized.reference as nnqr +from torch.ao.nn.quantized.modules.utils import WeightedQuantizedModule +from torch.fx import GraphModule +from .utils import ( + collect_producer_nodes, + get_linear_prepack_op_for_dtype, + get_new_attr_name_with_prefix, + get_qconv_prepack_op, + graph_module_from_producer_nodes, +) +from ..utils import _parent_name +from ..qconfig import QConfigAny +from ..quantization_mappings import get_quantized_operator +from .utils import create_node_from_old_node_preserve_meta +from typing import Dict, Tuple, Type, List, Callable, Any, Union, Set, Optional +import operator + +QOP_TO_ARG_NAMES_TO_SKIP = { + torch._ops.ops.quantized.hardswish: ['inplace'], + torch._ops.ops.quantized.elu: ['inplace'], + torch._ops.ops.quantized.dropout: ['inplace'], + torch._ops.ops.quantized.instance_norm: + ['running_mean', 'running_var', 'use_input_stats', 'momentum'], +} + +def _is_node_in_list(node, modules, func_list, method_list, module_type_list): + is_call_function = node.op == "call_function" and node.target in func_list + is_call_method = node.op == "call_method" and node.target in method_list + is_call_module = node.op == "call_module" and type(modules[str(node.target)]) in module_type_list + return is_call_function, is_call_method, is_call_module + +def is_fixed_qparams_node(node, modules): + func_list = [ + torch.nn.functional.hardsigmoid, + torch.nn.functional.sigmoid, + torch.sigmoid, + torch.tanh, + ] + method_list = [ + "hardsigmoid", + "hardsigmoid_", + "sigmoid", + "sigmoid_", + "tanh", + "tanh_", + ] + module_type_list = [ + torch.nn.Hardsigmoid, + torch.nn.Sigmoid, + torch.nn.Tanh, + torch.nn.Softmax, + ] + return _is_node_in_list(node, modules, func_list, method_list, module_type_list) + +def is_default_node(node, modules): + func_list = [ + torch.nn.functional.elu, + torch.nn.functional.hardswish, + torch.nn.functional.instance_norm, + torch.nn.functional.layer_norm, + torch.nn.functional.leaky_relu, + torch.nn.functional.dropout, + ] + method_list: List[Any] = [] + module_type_list = [ + nnqr.ConvTranspose1d, + nnqr.ConvTranspose2d, + nnqr.ConvTranspose3d, + torch.nn.ELU, + torch.nn.LeakyReLU, + torch.nn.Hardswish, + torch.nn.InstanceNorm1d, + torch.nn.InstanceNorm2d, + torch.nn.InstanceNorm3d, + torch.nn.LayerNorm, + torch.nn.Dropout, + torch.nn.PReLU, + torch.nn.BatchNorm2d, + torch.nn.BatchNorm3d, + torch.ao.nn.intrinsic.BNReLU2d, + torch.ao.nn.intrinsic.BNReLU3d, + ] + return _is_node_in_list(node, modules, func_list, method_list, module_type_list) + +def is_copy_node(node, modules): + func_list = [ + torch.adaptive_avg_pool1d, + torch.nn.functional.adaptive_avg_pool2d, + torch.nn.functional.adaptive_avg_pool3d, + torch.nn.functional.hardtanh, + torch.nn.functional.hardtanh_, + torch.nn.functional.interpolate, + torch.nn.functional.max_pool1d, + torch.nn.functional.max_pool2d, + torch.nn.functional.max_pool3d, + torch.nn.functional.relu, + torch.nn.functional.relu6, + torch.avg_pool1d, + torch._C._nn.avg_pool2d, + torch._C._nn.avg_pool3d, + torch.clamp, + torch.flatten, + torch.mean, + operator.floordiv, + # F.channel_shuffle and torch.channel_shuffle are essentially the same thing + # so we only need to put one of them here + torch.channel_shuffle, + ] + method_list = [ + "clamp", + "mean", + "relu", + "relu_", + ] + module_type_list = [ + torch.nn.AdaptiveAvgPool1d, + torch.nn.AdaptiveAvgPool2d, + torch.nn.AdaptiveAvgPool3d, + torch.nn.AvgPool1d, + torch.nn.AvgPool2d, + torch.nn.AvgPool3d, + torch.nn.Hardtanh, + torch.nn.MaxPool1d, + torch.nn.MaxPool2d, + torch.nn.MaxPool3d, + torch.nn.ReLU, + torch.nn.ReLU6, + torch.nn.ChannelShuffle, + ] + return _is_node_in_list(node, modules, func_list, method_list, module_type_list) + +def is_general_tensor_shape_node(node, modules): + func_list = [ + torch.narrow, + torch.transpose, + torch.repeat_interleave, + torch.squeeze, + torch.stack, + torch.unsqueeze, + torch.nn.functional.pixel_shuffle, + torch.nn.functional.pixel_unshuffle, + ] + method_list = [ + "contiguous", + "detach", + "detach_", + "permute", + "repeat", + "repeat_interleave", + "reshape", + "resize_", + "shape", + "size", + "squeeze", + "squeeze_", + "transpose", + "unsqueeze", + "unsqueeze_", + "view", + ] + module_type_list = [ + torch.nn.Identity, + torch.nn.PixelShuffle, + torch.nn.PixelUnshuffle, + ] + return _is_node_in_list(node, modules, func_list, method_list, module_type_list) + +def is_other_node(node, modules): + func_list = [ + torch.cat, + ] + method_list: List[Any] = [] + module_type_list: List[Any] = [] + return _is_node_in_list(node, modules, func_list, method_list, module_type_list) + +def is_special_pattern_node(node, modules): + res_function, res_method, res_module = False, False, False + for checker in [is_fixed_qparams_node, is_default_node, is_copy_node, is_general_tensor_shape_node, is_other_node]: + is_call_function, is_call_method, is_call_module = checker(node, modules) + res_function = res_function or is_call_function + res_method = res_method or is_call_method + res_module = res_module or is_call_module + return res_function, res_method, res_module + +def is_dequantize_node(node): + return isinstance(node, Node) and node.op == "call_method" and node.target == "dequantize" + +def is_getattr_tensor_metadata_node(node): + return node.op == "call_function" and \ + node.target == getattr and \ + node.args[1] in ["shape"] + +def is_get_tensor_info_node(node): + return node.op == "call_method" and \ + node.target in ["shape", "size"] + +def should_skip_lowering(op: torch.fx.node.Node, qconfig_map: Dict[str, QConfigAny]): + """ + Return True if the op is configured with a None qconfig, False otherwise. + Note: maybe need to generalize this to also check for the dtype, and we + only lower when dtype matches, but right now fbgemm/qnnpack only support + a single dtype, so it is OK for now. + """ + return op.name in qconfig_map and qconfig_map[op.name] is None + +# Mapping from reference module class to the replacement static quantized module class for lowering +STATIC_LOWER_MODULE_MAP: Dict[Type[nn.Module], Type[WeightedQuantizedModule]] = { + nnqr.Linear: nnq.Linear, + nnqr.Conv1d: nnq.Conv1d, + nnqr.Conv2d: nnq.Conv2d, + nnqr.Conv3d: nnq.Conv3d, +} + +# Mapping from reference module class to the replacement dynamic quantized module class for lowering +DYNAMIC_LOWER_MODULE_MAP: Dict[Type[nn.Module], Type[nn.Module]] = { + nnqr.Linear: nnqd.Linear, + nnqr.GRUCell: nnqd.GRUCell, + nnqr.LSTMCell: nnqd.LSTMCell, + nnqr.RNNCell: nnqd.RNNCell, + nnqr.LSTM: nnqd.LSTM, + nnqr.GRU: nnqd.GRU, +} + +# Mapping from reference module class to the replacement weight only quantized module class for lowering +# TODO: correct the namespace for these modules +WEIGHT_ONLY_LOWER_MODULE_MAP: Dict[Type[nn.Module], Type[nn.Module]] = { + nnqr.Embedding: nnq.Embedding, + nnqr.EmbeddingBag: nnq.EmbeddingBag, +} + +# TODO: merge with STATIC_LOWER_MODULE_MAP after we merge +# _lower_static_weighted_ref_module and special_pattern_replacement +SPECIAL_PATTERN_LOWER_MODULE_MAP = { + nn.BatchNorm2d: nnq.BatchNorm2d, + nn.BatchNorm3d: nnq.BatchNorm3d, + nnqr.ConvTranspose1d: nnq.ConvTranspose1d, + nnqr.ConvTranspose2d: nnq.ConvTranspose2d, + nnqr.ConvTranspose3d: nnq.ConvTranspose3d, + nn.ELU: nnq.ELU, + nn.LeakyReLU: nnq.LeakyReLU, + nn.Hardswish: nnq.Hardswish, + nn.InstanceNorm1d: nnq.InstanceNorm1d, + nn.InstanceNorm2d: nnq.InstanceNorm2d, + nn.InstanceNorm3d: nnq.InstanceNorm3d, + nn.LayerNorm: nnq.LayerNorm, + nn.Dropout: nnq.Dropout, + nn.Softmax: nnq.Softmax, + nn.PReLU: nnq.PReLU, + nni.BNReLU2d: nniq.BNReLU2d, + nni.BNReLU3d: nniq.BNReLU3d, +} + +# Mapping from fused module class to a 2-tuple of: +# 1) The inner reference module class +# 2) The replacement static quantized module class for lowering +STATIC_LOWER_FUSED_MODULE_MAP: Dict[Type[nn.Module], Tuple[Type[nn.Module], Type[WeightedQuantizedModule]]] = { + nni.LinearReLU: (nnqr.Linear, nniq.LinearReLU), + # TODO: LinearLeakyReLU is registered as global but it is only fused and + # lowered when ondnn's backend config is used. Maybe need to separate + # registration and lowering functions for different backends in the future. + nni.LinearLeakyReLU: (nnqr.Linear, nniq.LinearLeakyReLU), + nni.LinearTanh: (nnqr.Linear, nniq.LinearTanh), + nni.ConvReLU1d: (nnqr.Conv1d, nniq.ConvReLU1d), + nni.ConvReLU2d: (nnqr.Conv2d, nniq.ConvReLU2d), + nni.ConvReLU3d: (nnqr.Conv3d, nniq.ConvReLU3d), +} + +# The difference between STATIC_LOWER_FUSED_MODULE_TWO_INPUTS_MAP and STATIC_LOWER_FUSED_MODULE_MAP: +# The refer node inside STATIC_LOWER_FUSED_MODULE_TWO_INPUTS_MAP has 2 inputs. +# Mapping from fused module class to a 2-tuple of: +# 1) The inner reference module class +# 2) The replacement static quantized module class for lowering +STATIC_LOWER_FUSED_MODULE_TWO_INPUTS_MAP: Dict[Type[nn.Module], Tuple[Type[nn.Module], Type[WeightedQuantizedModule]]] = { + nni.ConvAdd2d: (nnqr.Conv2d, nniq.ConvAdd2d), + nni.ConvAddReLU2d: (nnqr.Conv2d, nniq.ConvAddReLU2d), +} + +# Mapping from fused module class to a 2-tuple of: +# 1) The inner reference module class +# 2) The replacement dynamic quantized module class for lowering +DYNAMIC_LOWER_FUSED_MODULE_MAP: Dict[Type[nn.Module], Tuple[Type[nn.Module], Type[nn.Module]]] = { + nni.LinearReLU: (nnqr.Linear, nniqd.LinearReLU), +} + +# Mapping from a functional to lower to a 2-tuple of +# 1) The quantized version of the op +# 2) The quantized version of the op fused with relu, if it exists, else None +STATIC_LOWER_FUNCTIONAL_MAP: Dict[Callable, Tuple[Callable, Optional[Callable]]] = { + F.linear: (torch.ops.quantized.linear, torch.ops.quantized.linear_relu), + F.conv1d: (torch.ops.quantized.conv1d, torch.ops.quantized.conv1d_relu), + F.conv2d: (torch.ops.quantized.conv2d, torch.ops.quantized.conv2d_relu), + F.conv3d: (torch.ops.quantized.conv3d, torch.ops.quantized.conv3d_relu), + F.conv_transpose1d: (torch.ops.quantized.conv_transpose1d, None), + F.conv_transpose2d: (torch.ops.quantized.conv_transpose2d, None), + F.conv_transpose3d: (torch.ops.quantized.conv_transpose3d, None), +} + +WEIGHT_PREPACK_OPS: Set[Callable] = { + torch._ops.ops.quantized.linear_prepack, + torch._ops.ops.quantized.linear_prepack_fp16, + torch._ops.ops.quantized.conv1d_prepack, + torch._ops.ops.quantized.conv2d_prepack, + torch._ops.ops.quantized.conv3d_prepack, + torch.ops.quantized.conv_transpose1d_prepack, + torch.ops.quantized.conv_transpose2d_prepack, + torch.ops.quantized.conv_transpose3d_prepack, +} + +# Mapping from a functional to a dictionary, where the key is a 2-tuple of +# (input_activation_dtype, weight_dtype) and the value is a 2-tuple of +# 1) The dynamically quantized version of the op +# 2) The dynamically quantized version of the op fused with relu, if it exists, else None +DYNAMIC_LOWER_FUNCTIONAL_MAP: Dict[Callable, Dict[Tuple[torch.dtype, torch.dtype], Tuple[Callable, Optional[Callable]]]] = { + F.linear: { + (torch.quint8, torch.qint8): (torch.ops.quantized.linear_dynamic, + torch.ops.quantized.linear_relu_dynamic), + (torch.float16, torch.float16): (torch.ops.quantized.linear_dynamic_fp16, + torch.ops.quantized.linear_relu_dynamic_fp16) + }, + # dynamic conv + relu is not available yet + F.conv1d: { + (torch.quint8, torch.qint8): (torch.ops.quantized.conv1d_dynamic, None), + }, + F.conv2d: { + (torch.quint8, torch.qint8): (torch.ops.quantized.conv2d_dynamic, None), + }, + F.conv3d: { + (torch.quint8, torch.qint8): (torch.ops.quantized.conv3d_dynamic, None), + }, +} + +CONV_FUNCTIONAL_OPS: Set[Callable] = { + F.conv1d, + F.conv2d, + F.conv3d, +} + +CONV_TRANSPOSE_FUNCTIONAL_OPS: Set[Callable] = { + F.conv_transpose1d, + F.conv_transpose2d, + F.conv_transpose3d, +} + +QBIN_OP_MAPPING: Dict[Union[Callable, str], Callable] = { + operator.add: torch.ops.quantized.add, + torch.add: torch.ops.quantized.add, + operator.mul: torch.ops.quantized.mul, + torch.mul: torch.ops.quantized.mul, + torch.matmul: torch.ops.quantized.matmul, +} +QBIN_RELU_OP_MAPPING: Dict[Union[Callable, str], Callable] = { + operator.add: torch.ops.quantized.add_relu, + torch.add: torch.ops.quantized.add_relu, + operator.mul: torch.ops.quantized.mul_relu, + torch.mul: torch.ops.quantized.mul_relu, +} + +def _save_packed_weight(self, destination, prefix, keep_vars): + for attr_name in dir(self): + if "_packed_weight" in attr_name and \ + isinstance(getattr(self, attr_name), torch._C.ScriptObject): # type: ignore[attr-defined] + packed_weight = getattr(self, attr_name) + destination[prefix + attr_name] = packed_weight + +def _load_packed_weight(self, state_dict, prefix, local_metadata, strict, + missing_keys, unexpected_keys, error_msgs): + attrs_to_pop = [] + for attr_name in state_dict: + if attr_name.startswith("_packed_weight") and isinstance(state_dict[attr_name], torch._C.ScriptObject): # type: ignore[attr-defined] # noqa: B950 + setattr(self, attr_name, state_dict[attr_name]) + attrs_to_pop.append(attr_name) + + # pop the packed param attributesn + for attr_name in attrs_to_pop: + state_dict.pop(attr_name) + +def fold_weight( + quantized_model: GraphModule, + node_name_to_scope: Dict[str, Tuple[str, type]] +) -> GraphModule: + """ + Trace back from the weight node util we hit getattr, reconstruct the + graph module with the traced nodes and run the graph module to pack the + weight. then replace the original chain of ops with the packed weight. + """ + packed_weights = {} + # map from folded node name to the prepacked weight name + folded_nodes = {} + # get packed weights + for node in quantized_model.graph.nodes: + if node.op == 'call_function' and node.target in WEIGHT_PREPACK_OPS: + nodes_to_fold = collect_producer_nodes(node) + if nodes_to_fold is not None: + for node_to_fold in nodes_to_fold: + folded_nodes[node_to_fold.name] = node + + prepacking_module = graph_module_from_producer_nodes( + quantized_model, nodes_to_fold) + packed_weight = prepacking_module() + packed_weights[node.name] = packed_weight + + # remove folded nodes and replace the prepacking node with getattr + folded_graph = Graph() + env: Dict[Any, Any] = {} + + def load_arg(a): + return map_arg(a, lambda node: env[node.name]) + + for node in quantized_model.graph.nodes: + prepack_node = folded_nodes.get(node.name, None) + if prepack_node is node: + packed_weight = packed_weights[node.name] + # add a prepacked attribute to root + op_node = next(iter(prepack_node.users)) + module_path, _ = node_name_to_scope[op_node.name] + get_new_packed_weight_name = \ + get_new_attr_name_with_prefix(module_path + '_packed_weight_') + packed_weight_name = get_new_packed_weight_name(quantized_model) + setattr(quantized_model, packed_weight_name, packed_weight) + # replace prepack node with a getattr node + env[node.name] = folded_graph.create_node( + 'get_attr', packed_weight_name, (), {}) + elif prepack_node is not None: + # remove the foled node + continue + else: + # copy other nodes + env[node.name] = folded_graph.node_copy(node, load_arg) + + quantized_model = GraphModule(quantized_model, folded_graph) + quantized_model._register_state_dict_hook(_save_packed_weight) + quantized_model._register_load_state_dict_pre_hook(_load_packed_weight, with_module=True) + return quantized_model + +def _get_module(node: Node, modules: Dict[str, nn.Module]) -> Optional[nn.Module]: + """ + Return the `torch.nn.Module` that corresponds to the specified node's target. + If no such node exists, return None. + """ + if node.op == "call_module" and str(node.target) in modules: + return modules[str(node.target)] + else: + return None + +def _match_static_pattern( + node: Node, + modules: Dict[str, nn.Module], + qconfig_map: Dict[str, QConfigAny], + matching_modules_or_ops: List[Callable], + dequantize_node_arg_indices: List[int] +) -> Union[Tuple[Node, Node, Node], Tuple[None, None, None]]: + """ + Match the pattern (dequantize - ref node - quantize) against the node provided. + + If there is a match, return a 3-tuple of: + 1) q_node: the quantize node, + 2) relu_node: a relu node wrapping the ref_node, and + 3) ref_node: a reference module or functional node to replace with its quantized counterpart + Otherwise, if there is no match, return a 3-tuple of (None, None, None). + + Parameters: + node: The `torch.fx.Node` to match against. + modules: A mapping from node names to modules in the model graph, used for module lookup. + qconfig_map: A mapping from node names to the qconfigs associated with the nodes. + If the corresponding qconfig for the reference node is None, then return no match. + matching_modules_or_ops: Either a list of functions or a list of `torch.nn.Module`s. + If the reference node is not in this list, then return no match. + dequantize_node_arg_indices: A list of indices in the reference node args where dequantize + nodes may be present. An empty list means skipping the check for dequantize nodes. + """ + SKIP_LOWERING_VALUE = (None, None, None) + + # Match quantize node + if node.op != "call_function" or node.target != torch.quantize_per_tensor: + return SKIP_LOWERING_VALUE + q_node = node + ref_node = q_node.args[0] + assert(isinstance(ref_node, Node)) + + # Handle cases where the node is wrapped in a ReLU + if (ref_node.op == "call_function" and ref_node.target in (F.relu, torch.relu)) or\ + (ref_node.op == "call_module" and type(_get_module(ref_node, modules)) == nn.ReLU): + relu_node = ref_node + ref_node = relu_node.args[0] + assert(isinstance(ref_node, Node)) + else: + relu_node = None + if should_skip_lowering(ref_node, qconfig_map): + return SKIP_LOWERING_VALUE + + # Match reference module or functional + if isinstance(matching_modules_or_ops[0], type) and issubclass(matching_modules_or_ops[0], nn.Module): + expected_op = "call_module" + match_key = type(_get_module(ref_node, modules)) + else: + expected_op = "call_function" + match_key = ref_node.target + if ref_node.op != expected_op or match_key not in matching_modules_or_ops: + return SKIP_LOWERING_VALUE + + # Match dequantize node(s). Both of the following conditions must pass: + # (1) All `torch.fx.Node`s at the matching indices must be a dequantize node + # (2) There must be at least one dequantize node + matched_dequantize = False + for i in dequantize_node_arg_indices: + assert i < len(ref_node.args),\ + f"Dequantize index {i} exceeded reference node's arg length {len(ref_node.args)}" + arg = ref_node.args[i] + if is_dequantize_node(arg): + matched_dequantize = True + elif isinstance(arg, Node): + return SKIP_LOWERING_VALUE + if not matched_dequantize: + return SKIP_LOWERING_VALUE + + return (q_node, relu_node, ref_node) + +def _match_static_pattern_with_two_inputs( + node: Node, + modules: Dict[str, nn.Module], + qconfig_map: Dict[str, QConfigAny], + matching_modules_or_ops: List[Callable] +) -> Union[Tuple[Node, Node], Tuple[None, None]]: + """ + (dequantize \ + Match the pattern (dequantize - ref node - quantize) against the node provided. + + If there is a match, return a 2-tuple of: + 1) q_node: the quantize node, + 2) ref_node: a reference module or functional node to replace with its quantized counterpart + Otherwise, if there is no match, return a 2-tuple of (None, None). + + Parameters: + node: The `torch.fx.Node` to match against. + modules: A mapping from node names to modules in the model graph, used for module lookup. + qconfig_map: A mapping from node names to the qconfigs associated with the nodes. + If the corresponding qconfig for the reference node is None, then return no match. + matching_modules_or_ops: Either a list of functions or a list of `torch.nn.Module`s. + If the reference node is not in this list, then return no match. + """ + SKIP_LOWERING_VALUE = (None, None) + + # Match quantize node + if node.op != "call_function" or node.target != torch.quantize_per_tensor: + return SKIP_LOWERING_VALUE + q_node = node + ref_node = q_node.args[0] + assert(isinstance(ref_node, Node)) + + if should_skip_lowering(ref_node, qconfig_map): + return SKIP_LOWERING_VALUE + + # Match reference module or functional + if isinstance(matching_modules_or_ops[0], type) and issubclass(matching_modules_or_ops[0], nn.Module): + expected_op = "call_module" + match_key = type(_get_module(ref_node, modules)) + else: + # This pass only support op of "call_module" + return SKIP_LOWERING_VALUE + + if ref_node.op != expected_op or match_key not in matching_modules_or_ops: + return SKIP_LOWERING_VALUE + + # Check ref_node has 2 input nodes, both are dq node. + if len(ref_node.args) != 2: + return SKIP_LOWERING_VALUE + for i in range(len(ref_node.args)): + arg = ref_node.args[i] + if not is_dequantize_node(arg): + return SKIP_LOWERING_VALUE + + return (q_node, ref_node) + +def _lower_static_weighted_ref_module( + model: GraphModule, + qconfig_map: Dict[str, QConfigAny]): + """ + Traverse the graph and find dequantize - ref module - quantize patterns + and replace them with the quantized version of the ref module. + """ + modules = dict(model.named_modules(remove_duplicate=False)) + nodes = list(model.graph.nodes) + for n in model.graph.nodes: + # Step 0: Find nodes that match this pattern (dequantize - ref module - quantize) + matching_modules = list(STATIC_LOWER_MODULE_MAP.keys()) + list(STATIC_LOWER_FUSED_MODULE_MAP.keys()) + (q_node, relu_node, ref_node) = _match_static_pattern( + n, modules, qconfig_map, matching_modules, dequantize_node_arg_indices=[0]) # type: ignore[arg-type] + if q_node is None: + continue + assert(ref_node is not None) + (_, scale_node, zero_point_node, _) = q_node.args + ref_module = _get_module(ref_node, modules) + ref_class = type(ref_module) + assert(isinstance(scale_node, Node)) + assert(isinstance(zero_point_node, Node)) + assert(issubclass(ref_class, nn.Module)) + + # Step 1: Change this pattern to use the corresponding quantized module + # For fused modules, we also check whether the inner module is a reference module + # If so, we replace the entire fused module with the corresponding quantized module + if ref_class in STATIC_LOWER_FUSED_MODULE_MAP: + inner_ref_class, q_class = STATIC_LOWER_FUSED_MODULE_MAP[ref_class] + if type(ref_module[0]) != inner_ref_class: # type: ignore[index] + continue + else: + q_class = STATIC_LOWER_MODULE_MAP[ref_class] + output_scale = getattr(model, scale_node.target) + output_zero_point = getattr(model, zero_point_node.target) + q_module = q_class.from_reference(ref_module, output_scale, output_zero_point) + # replace reference module with quantized module + parent_name, module_name = _parent_name(ref_node.target) + setattr(modules[parent_name], module_name, q_module) + + # Step 2: Reroute around dq_node, and remove q_node and its args + assert(len(ref_node.args) == 1) + dq_node = ref_node.args[0] + assert(isinstance(dq_node, Node)) + ref_node.replace_input_with(dq_node, dq_node.args[0]) + q_node.replace_all_uses_with(ref_node) + model.graph.erase_node(q_node) + model.graph.erase_node(scale_node) + model.graph.erase_node(zero_point_node) + +def _lower_static_weighted_ref_module_with_two_inputs( + model: GraphModule, + qconfig_map: Dict[str, QConfigAny]): + """ + Traverse the graph and find patterns + dequantize dequantize + \\ // + ref module + \\ + quantize + and replace them with the quantized version of the ref module. + """ + modules = dict(model.named_modules(remove_duplicate=False)) + nodes = list(model.graph.nodes) + for n in model.graph.nodes: + # (dequantize \ + # Step 0: Find nodes that match this pattern (dequantize - ref module - quantize) + matching_modules = list(STATIC_LOWER_FUSED_MODULE_TWO_INPUTS_MAP.keys()) + (q_node, ref_node) = _match_static_pattern_with_two_inputs( + n, modules, qconfig_map, matching_modules) # type: ignore[arg-type] + if q_node is None: + continue + assert(ref_node is not None) + (_, scale_node, zero_point_node, _) = q_node.args + ref_module = _get_module(ref_node, modules) + ref_class = type(ref_module) + assert(isinstance(scale_node, Node)) + assert(isinstance(zero_point_node, Node)) + assert(issubclass(ref_class, nn.Module)) + + # Step 1: Change this pattern to use the corresponding quantized module + # For fused modules, we also check whether the inner module is a reference module + # If so, we replace the entire fused module with the corresponding quantized module + if ref_class in STATIC_LOWER_FUSED_MODULE_TWO_INPUTS_MAP: + inner_ref_class, q_class = STATIC_LOWER_FUSED_MODULE_TWO_INPUTS_MAP[ref_class] + if type(ref_module[0]) != inner_ref_class: # type: ignore[index] + continue + else: + continue + output_scale = getattr(model, scale_node.target) + output_zero_point = getattr(model, zero_point_node.target) + q_module = q_class.from_reference(ref_module, output_scale, output_zero_point) + # replace reference module with quantized module + parent_name, module_name = _parent_name(ref_node.target) + setattr(modules[parent_name], module_name, q_module) + + # Step 2: Reroute around dq_node, and remove q_node and its args + assert(len(ref_node.args) == 2) + for arg in ref_node.args: + if not is_dequantize_node(arg): + continue + dq_node = arg + assert(isinstance(dq_node, Node)) + ref_node.replace_input_with(dq_node, dq_node.args[0]) + + q_node.replace_all_uses_with(ref_node) + model.graph.erase_node(q_node) + model.graph.erase_node(scale_node) + model.graph.erase_node(zero_point_node) + +def _lower_dynamic_weighted_ref_module(model: GraphModule): + """ + Traverse the graph and find quantize_per_tensor_dynamic - dequantize - ref_module patterns + and replace them with the dynamically quantized version of the ref module. + """ + named_modules = dict(model.named_modules(remove_duplicate=False)) + for n in model.graph.nodes: + if n.op != "call_module" or \ + type(named_modules[str(n.target)]) not in \ + set(DYNAMIC_LOWER_MODULE_MAP.keys()).union( + set(DYNAMIC_LOWER_FUSED_MODULE_MAP.keys())): + continue + ref_node = n + dq_node = ref_node.args[0] + if dq_node.op != "call_method" or dq_node.target != "dequantize": + continue + + input_dynamic_q_node = dq_node.args[0] + + if input_dynamic_q_node.op != "call_function" or \ + input_dynamic_q_node.target != torch.quantize_per_tensor_dynamic: + continue + + activation_dtype = input_dynamic_q_node.args[1] + is_fp16 = activation_dtype == torch.float16 + is_int8 = activation_dtype in [torch.quint8, torch.qint8] + if not is_int8 and not is_fp16: + continue + + ref_module = named_modules[str(ref_node.target)] + ref_class = type(ref_module) + if ref_class in DYNAMIC_LOWER_FUSED_MODULE_MAP: + inner_ref_class, q_class = DYNAMIC_LOWER_FUSED_MODULE_MAP[ref_class] + if type(ref_module[0]) != inner_ref_class: + continue + else: + q_class = DYNAMIC_LOWER_MODULE_MAP.get(ref_class) # type: ignore[assignment] + # TODO: maybe define a WeightedDynamicallyQuantizedModule + q_module = q_class.from_reference(ref_module) # type: ignore[attr-defined] + + # replace reference module with dynamically quantized module + parent_name, module_name = _parent_name(ref_node.target) + setattr(named_modules[parent_name], module_name, q_module) + ref_node.replace_input_with(dq_node, input_dynamic_q_node.args[0]) + +def _lower_weight_only_weighted_ref_module(model: GraphModule): + """ + Traverse the graph and find ref_module patterns + and replace them with the weight only quantized version of the ref module. + """ + named_modules = dict(model.named_modules(remove_duplicate=False)) + for n in model.graph.nodes: + if n.op != "call_module" or \ + type(named_modules[str(n.target)]) not in \ + set(WEIGHT_ONLY_LOWER_MODULE_MAP.keys()): + continue + ref_node = n + ref_module = named_modules[str(ref_node.target)] + ref_class = type(ref_module) + q_class = WEIGHT_ONLY_LOWER_MODULE_MAP.get(ref_class) + # TODO: WeightedQuantizedModule is currently assuming static quant apis + # with output_scale, output_zero_point in from_reference, we may want to + # relax that, or rename this + # TODO: maybe define a WeightedWeightOnlyQuantizedModule + q_module = q_class.from_reference(ref_module) # type: ignore[union-attr] + + # replace reference module with dynamically quantized module + parent_name, module_name = _parent_name(ref_node.target) + setattr(named_modules[parent_name], module_name, q_module) + +def _lower_static_weighted_ref_functional( + model: GraphModule, + qconfig_map: Dict[str, QConfigAny]): + """ + Traverse the graph and replace functional reference patterns with their quantized versions. + """ + modules = dict(model.named_modules(remove_duplicate=False)) + nodes = list(model.graph.nodes) + for n in model.graph.nodes: + # Step 0: Find nodes that match this pattern (dequantize - functional op - quantize) + matching_ops = list(STATIC_LOWER_FUNCTIONAL_MAP.keys()) + (q_node, relu_node, func_node) = _match_static_pattern( + n, modules, qconfig_map, matching_ops, dequantize_node_arg_indices=[0, 1]) + if q_node is None: + continue + assert(func_node is not None) + (_, output_scale_node, output_zp_node, _) = q_node.args + (input_dq_node, weight_dq_node, *remaining_func_args) = func_node.args + assert(isinstance(output_zp_node, Node)) + assert(isinstance(input_dq_node, Node)) + assert(isinstance(weight_dq_node, Node)) + quantized_weight = weight_dq_node.args[0] + assert(isinstance(quantized_weight, Node)) + if quantized_weight.op != "call_function" or\ + quantized_weight.target not in (torch.quantize_per_tensor, torch.quantize_per_channel): + continue + + # Step 1: Replace quantized weights with packed weights, which will be folded later + # Use the right prepack op and prepare the corresponding args + # Linear prepack args: (quantized weights[, bias]) + # Conv prepack args: (quantized weights[, bias, stride, padding, dilation, groups]) + prepack_args = [quantized_weight] + remaining_func_args + if func_node.target == F.linear: + weight_dtype = quantized_weight.args[-1] + prepack_op = get_linear_prepack_op_for_dtype(weight_dtype) + elif func_node.target in CONV_FUNCTIONAL_OPS: + prepack_op = get_qconv_prepack_op(func_node.target) # type: ignore[arg-type] + # For conv1d, the stride, padding, and dilation args may be ints, + # in which case we need to convert them to tuples + if func_node.target == F.conv1d: + for i in [2, 3, 4]: + if len(prepack_args) > i and isinstance(prepack_args[i], int): + prepack_args[i] = (prepack_args[i],) + elif func_node.target in CONV_TRANSPOSE_FUNCTIONAL_OPS: + prepack_op = get_qconv_prepack_op(func_node.target) # type: ignore[arg-type] + # For conv_transpose1d, the stride, padding, and dilation args may be ints, + # in which case we need to convert them to tuples + if func_node.target == F.conv_transpose1d: + # Note prepack_args[5] is groups. + for i in [2, 3, 4, 6]: + if len(prepack_args) > i and isinstance(prepack_args[i], int): + prepack_args[i] = (prepack_args[i],) + # swap dilation and groups + # prepack op has arguments: {w, b, stride, padding, output_padding, dilation, groups} + # transposed conv op has arguments: {x, w, b, stride, padding, output_padding, groups, dilation} + if (len(prepack_args) > 6): + prepack_args[5], prepack_args[6] = prepack_args[6], prepack_args[5] + else: + raise ValueError(f"Lowering is not supported for op '{func_node.target}'") + with model.graph.inserting_before(output_scale_node): + # kwargs of the func node are needed for prepack op (i.e., quantized::linear_prepack) + # They are not needed for compute op (i.e., quantized::linear) + kwargs = func_node.kwargs + # F.linear uses 'bias' key for bias while qlinear_prepack uses 'B' for bias + if func_node.target == F.linear and 'bias' in kwargs: + kwargs = kwargs.copy() + kwargs['B'] = kwargs['bias'] + del kwargs['bias'] + packed_weight = model.graph.create_node("call_function", prepack_op, tuple(prepack_args), kwargs) + + # Step 2: Replace reference pattern with the corresponding quantized op + (q_func, q_relu_func) = STATIC_LOWER_FUNCTIONAL_MAP[func_node.target] # type: ignore[index] + # conv_transpose does not support fusion with relu yet. q_relu_func is None in such cases + if q_relu_func is not None: + func_node.target = q_relu_func if relu_node is not None else q_func + else: + func_node.target = q_func + func_node.args = (input_dq_node.args[0], packed_weight, output_scale_node, output_zp_node) + # kwargs for func_node has been moved to kwargs for prepack op + func_node.kwargs = {} + q_node.replace_all_uses_with(func_node) + # Move func_node after output_zp_node in the graph + output_zp_node.append(func_node) + + # Clean up: Remove quantize node, and the relu node if it exists + model.graph.erase_node(q_node) + if relu_node is not None and q_relu_func is not None: + model.graph.erase_node(relu_node) + +def _lower_dynamic_weighted_ref_functional( + model: GraphModule, + qconfig_map: Dict[str, QConfigAny]): + """ + Traverse the graph and replace functional reference patterns with their dynamically + quantized versions. + Examples: + quantize_per_tensor_dynamic - dequantize - functional linear --> linear_dynamic + to(torch.float16) - dequantize - functional linear --> linear_dynamic_fp16 + """ + modules = dict(model.named_modules(remove_duplicate=False)) + nodes = list(model.graph.nodes) + # we want to search in reserved order so that we can match the larger patterns first + # e.g. we want to match linear - relu before linear. + for n in reversed(model.graph.nodes): + + # Step 0: Find nodes that match this pattern + # (quantize_per_tensor_dynamic - dequantize - dynamically quantized op) + # We search for the pattern backwards, starting with the quantize node + # Quantize node args: (func, scale, zp, dtype) + func_node = n + # Handle cases where the functional op is wrapped in a ReLU + if func_node.op == "call_function" and func_node.target == F.relu or \ + func_node.op == "call_module" and \ + type(modules[str(func_node.target)]) == torch.nn.ReLU: + relu_node = func_node + func_node = relu_node.args[0] + else: + relu_node = None + if should_skip_lowering(func_node, qconfig_map): + continue + # Linear args: (dequantized inputs, dequantized weights[, bias]) + # Conv args: (dequantized inputs, dequantized weights[, bias, stride, padding, dilation, groups]) + if func_node.op != "call_function" or func_node.target not in DYNAMIC_LOWER_FUNCTIONAL_MAP: + continue + (input_dq_node, weight_dq_node, *remaining_func_args) = func_node.args + if input_dq_node.op != "call_method" or input_dq_node.target != "dequantize" or \ + weight_dq_node.op != "call_method" or weight_dq_node.target != "dequantize": + continue + + input_dynamic_q_node = input_dq_node.args[0] + + if input_dynamic_q_node.op != "call_function" or \ + input_dynamic_q_node.target != torch.quantize_per_tensor_dynamic: + continue + + reduce_range_node = None + (pattern_input, activation_dtype, reduce_range_node) = input_dynamic_q_node.args + is_fp16 = activation_dtype == torch.float16 + is_int8 = activation_dtype in [torch.quint8, torch.qint8] + if not is_int8 and not is_fp16: + continue + + quantized_weight = weight_dq_node.args[0] + weight_dtype = quantized_weight.args[-1] + + # Step 1: Try to select reference pattern with the corresponding quantized op + dynamic_quant_dtype_key = (activation_dtype, weight_dtype) + if dynamic_quant_dtype_key not in DYNAMIC_LOWER_FUNCTIONAL_MAP[func_node.target]: + print(f"Didn't find dtype combination {dynamic_quant_dtype_key} during " + f"dynamic quantized op lowering for {func_node.target}") + continue + (q_func, q_relu_func) = DYNAMIC_LOWER_FUNCTIONAL_MAP[func_node.target][dynamic_quant_dtype_key] + + if q_func is None or q_relu_func is None: + print("Didn't find corresponding quantized function or quantized relu function " + f"for {func_node.target}, {dynamic_quant_dtype_key}") + continue + + # Step 2: Replace quantized weights with packed weights, which will be folded later + # Use the right prepack op and prepare the corresponding args + # Linear prepack args: (quantized weights[, bias]) + # Conv prepack args: (quantized weights[, bias, stride, padding, dilation, groups]) + prepack_args = [quantized_weight] + remaining_func_args + if func_node.target == F.linear: + prepack_op = get_linear_prepack_op_for_dtype(weight_dtype) + elif func_node.target in CONV_FUNCTIONAL_OPS: + prepack_op = get_qconv_prepack_op(func_node.target) + # For conv1d, the stride, padding, and dilation args may be ints, + # in which case we need to convert them to tuples + if func_node.target == F.conv1d: + for i in [2, 3, 4]: + if len(prepack_args) > i and isinstance(prepack_args[i], int): + prepack_args[i] = (prepack_args[i],) + else: + raise ValueError(f"Lowering is not supported for op '{func_node.target}'") + with model.graph.inserting_before(func_node): + packed_weight = model.graph.create_node("call_function", prepack_op, tuple(prepack_args), {}) + + # Step 3: Replace reference pattern with the corresponding quantized op + func_node.target = q_relu_func if relu_node is not None else q_func + if is_int8: + func_node.args = (pattern_input, packed_weight, reduce_range_node) + else: + func_node.args = (pattern_input, packed_weight) + + if relu_node is not None: + relu_node.replace_all_uses_with(func_node) + + # Step 4: Remove the relu node if it exists + if relu_node is not None: + model.graph.erase_node(relu_node) + +def _lower_quantized_binary_op( + model: GraphModule, + qconfig_map: Dict[str, QConfigAny]): + binary_ops_to_lower: List[Callable] = [operator.add, torch.add, operator.mul, torch.mul, torch.matmul] + modules = dict(model.named_modules(remove_duplicate=False)) + for n in model.graph.nodes: + # Step 0: Find nodes that match this pattern (dequantize - ref module - quantize) + (q_node, relu_node, bop_node) = _match_static_pattern( + n, modules, qconfig_map, binary_ops_to_lower, dequantize_node_arg_indices=[0, 1]) + if q_node is None: + continue + assert(bop_node is not None) + (_, scale_node, zero_point_node, _) = q_node.args + + # Step 1: Remove dequant nodes + num_dq_nodes = 0 + for arg in bop_node.args: + if not is_dequantize_node(arg): + continue + dq_node = arg + assert(isinstance(dq_node, Node)) + dn_input = dq_node.args[0] + bop_node.replace_input_with(dq_node, dn_input) + num_dq_nodes += 1 + assert(num_dq_nodes > 0) + + # Step 2: Swap binary op to quantized binary op + assert bop_node.target in QBIN_OP_MAPPING + binop_to_qbinop = QBIN_OP_MAPPING if relu_node is None else QBIN_RELU_OP_MAPPING + qbin_op = binop_to_qbinop[bop_node.target] + # prepare the args for quantized binary op + # (x, y) + qop_node_args = list(bop_node.args) + # (x, y, scale, zero_point) + # add scale and zero_point arguments for Tensor - Tensor operation + if num_dq_nodes == 2: + qop_node_args.extend([scale_node, zero_point_node]) + # insert a call to quantized binary op and remove the original binary op + with model.graph.inserting_after(q_node): + qop_node = create_node_from_old_node_preserve_meta( + model.graph, + ("call_function", qbin_op, tuple(qop_node_args), {}), + bop_node) + q_node.replace_all_uses_with(qop_node) + + # Step 3: Remove quantize node, binary op node, and relu node if any + model.graph.erase_node(q_node) + if relu_node is not None: + model.graph.erase_node(relu_node) + model.graph.erase_node(bop_node) + +def special_pattern_replacement(model: GraphModule): + modules = dict(model.named_modules(remove_duplicate=False)) + for n in model.graph.nodes: + q_node = n + is_quantize = q_node.target == torch.quantize_per_tensor + is_to_fp16 = q_node.op == "call_method" and q_node.target == "to" and \ + len(q_node.args) == 2 and q_node.args[1] == torch.float16 + if not (is_quantize or is_to_fp16): + continue + ref_node = q_node.args[0] + # get output scale/zero_point/dtype from the quantize node + # ref_node, scale_node, zero_point_node, dtype = q_node.args + # TODO: add safety checks that users for the ref_node and dq_node needs to be one + is_call_function, is_call_method, is_call_module = is_fixed_qparams_node(ref_node, modules) + if is_to_fp16 and (is_call_function or is_call_method or is_call_module): + # TODO: add a warning or error out here? (bc-breaking if error out) + # warnings.warn( + # "Only reference patterns are currently supported for {dtype} dtype with {op} op" + # "".format(dtype=dtypes, op=ref_node)) + continue + + is_call_function, is_call_method, is_call_module = is_default_node(ref_node, modules) + if is_to_fp16 and (is_call_function or is_call_method or is_call_module): + # TODO: add a warning or error out here? (bc-breaking if error out) + continue + + # This check includes all supported ops + is_call_function, is_call_method, is_call_module = is_special_pattern_node(ref_node, modules) + if not (is_call_module or is_call_function or is_call_method): + continue + assert len(ref_node.args) > 0 or len(ref_node.kwargs) > 0 + dq_node_or_nodes = ref_node.args[0] if len(ref_node.args) > 0 else next(iter(ref_node.kwargs.values())) + assert isinstance(dq_node_or_nodes, (Node, tuple, list)) + is_dequantize = False + if isinstance(dq_node_or_nodes, Node): + is_dequantize = dq_node_or_nodes.op == 'call_method' and \ + dq_node_or_nodes.target == 'dequantize' + elif isinstance(dq_node_or_nodes, (tuple, list)): + is_dequantize = all( + x.op == 'call_method' and x.target == 'dequantize' + for x in dq_node_or_nodes) + + if not is_dequantize: + continue + + # TODO: enable we have patterns that needs to swap the modules + if is_call_module: + ref_module = modules[ref_node.target] + if type(ref_module) in SPECIAL_PATTERN_LOWER_MODULE_MAP and is_quantize: + qmodule_cls = SPECIAL_PATTERN_LOWER_MODULE_MAP.get(type(ref_module)) + scale_node = q_node.args[1] + zero_point_node = q_node.args[2] + output_scale = getattr(model, scale_node.target) + output_zero_point = getattr(model, zero_point_node.target) + + qmodule = qmodule_cls.from_reference(ref_module, output_scale, output_zero_point) # type:ignore[union-attr] + # replace reference module with quantized module + parent_name, module_name = _parent_name(ref_node.target) + setattr(modules[parent_name], module_name, qmodule) + + # reroute around dq node: + dq_nodes: List[Node] = [] + if isinstance(dq_node_or_nodes, Node): + dq_nodes = [dq_node_or_nodes] + elif isinstance(dq_node_or_nodes, (tuple, list)): + dq_nodes = list(dq_node_or_nodes) + + for dq_node in dq_nodes: + dn_input = dq_node.args[0] + ref_node.replace_input_with(dq_node, dn_input) + + # store q node args + qnode_qparams = list(q_node.args)[1:] + # replace uses of q node with input and remove q node + q_node_input = q_node.args[0] + q_node.replace_all_uses_with(q_node_input) + model.graph.erase_node(q_node) + + is_call_function, is_call_method, is_call_module = is_default_node(ref_node, modules) + if is_call_function: + # pass scale/zer_point arguments from quantize_per_tensor to the default node operator + # insert an op after the zero_point node so that the scale/zero_point + # nodes are is available + qop = get_quantized_operator(ref_node.target) + args = list(ref_node.args) + kwargs = dict(ref_node.kwargs) + if qop in QOP_TO_ARG_NAMES_TO_SKIP: + args_to_skip = QOP_TO_ARG_NAMES_TO_SKIP[qop] + for arg in args_to_skip: + if arg in kwargs: + kwargs.pop(arg) + kwargs["output_scale"] = qnode_qparams[0] + kwargs["output_zero_point"] = qnode_qparams[1] + with model.graph.inserting_after(qnode_qparams[1]): + qop_node = create_node_from_old_node_preserve_meta( + model.graph, + ("call_function", qop, tuple(args), kwargs), + ref_node) + ref_node.replace_all_uses_with(qop_node) + model.graph.erase_node(ref_node) + else: + # remove scale/zero_point node for quantize node + for n in qnode_qparams: + if isinstance(n, Node): + model.graph.erase_node(n) + + return model + +def _lower_getattr_tensor_metadta_op(model: GraphModule): + """ Modified the graph of the model inplace, to skip extra dequantize op before + the general tensor shape ops when possible + """ + for n in model.graph.nodes: + if is_getattr_tensor_metadata_node(n): + maybe_dq = n.args[0] + if maybe_dq.op != "call_method" or maybe_dq.target != "dequantize": + continue + # skip the dequantize node + args = list(n.args) + args[0] = n.args[0].args[0] + n.args = tuple(args) + +def _lower_get_tensor_info_op(model: GraphModule): + """ Modified the graph of the model inplace, to skip extra dequantize op before + the general tensor shape ops when possible + """ + for n in model.graph.nodes: + if not is_get_tensor_info_node(n): + continue + maybe_dq = n.args[0] + if maybe_dq.op != "call_method" or maybe_dq.target != "dequantize": + continue + # skip the dequantize node + args = list(n.args) + args[0] = n.args[0].args[0] + n.args = tuple(args) + +def _lower_to_native_backend( + model: GraphModule, + qconfig_map: Dict[str, QConfigAny], + node_name_to_scope: Dict[str, Tuple[str, type]] +) -> GraphModule: + """ Lower a quantized reference model (with reference quantized operator patterns) + to the native backend in PyTorch (fbgemm/qnnpack), both backends shares the same + operator signature so they can be lowered with the same function + """ + _lower_static_weighted_ref_module(model, qconfig_map) + _lower_static_weighted_ref_module_with_two_inputs(model, qconfig_map) + _lower_dynamic_weighted_ref_module(model) + _lower_weight_only_weighted_ref_module(model) + _lower_static_weighted_ref_functional(model, qconfig_map) + _lower_dynamic_weighted_ref_functional(model, qconfig_map) + _lower_quantized_binary_op(model, qconfig_map) + _lower_getattr_tensor_metadta_op(model) + _lower_get_tensor_info_op(model) + special_pattern_replacement(model) + model.graph.eliminate_dead_code() + model = fold_weight(model, node_name_to_scope) + model.graph.eliminate_dead_code() + model.recompile() + model.graph.lint() + return model diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/__init__.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/__init__.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..2c11872f615b8d259c5f706d6cc2e21212215f67 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/__init__.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/detector.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/detector.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..043766d392b7f806b81963ec487c915e8780693d Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/detector.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/model_report_observer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/model_report_observer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..841c58d80ff22ffd29964cc39c1de17681f5277e Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/model_report_observer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/model_report_visualizer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/model_report_visualizer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..96b852284f884b4f8930a104c3cef878f3fc8a7f Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/__pycache__/model_report_visualizer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/detector.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/detector.py new file mode 100644 index 0000000000000000000000000000000000000000..71986fd17fbe8e8c5061634dc69e4871ae61f4e7 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/_model_report/detector.py @@ -0,0 +1,1539 @@ +from typing import Any, Dict, Set, Tuple, Callable, List + +import torch +import torch.nn as nn +import torch.ao.nn.qat as nnqat +from abc import ABC, abstractmethod +from torch.ao.quantization.fake_quantize import FakeQuantize +from torch.ao.quantization.fx.graph_module import GraphModule +from torch.ao.quantization.fx._model_report.model_report_observer import ModelReportObserver +from torch.ao.quantization.qconfig import ( + QConfig, + default_qconfig, + _assert_valid_qconfig, +) +from torch.ao.quantization.observer import ( + ObserverBase, + default_dynamic_quant_observer, + default_per_channel_weight_observer, + default_observer, + default_weight_observer, +) +from torch.ao.quantization.fx._equalize import ( + default_equalization_qconfig, + EqualizationQConfig, +) +from torch.ao.quantization.observer import _is_activation_post_process + +# Names for observer insert keys +DETECTOR_TARGET_NODE_KEY = "target_node" +DETECTOR_OBS_TO_INSERT_KEY = "observer_to_insert" +DETECTOR_IS_POST_OBS_KEY = "is_post_observer" +DETECTOR_OBS_ARGS_KEY = "observer_args" + +# Mapping related code +class DetectorQConfigInfo: + r""" + This class contains the QConfig information for a single module. + The list of variables / values this contains can grow depending on the + extensibility of the qconfig mapping feature set but this currently includes: + - if activation observer is dynamic + - if weight observer is per channel + + + Args: + module_fqn (str): The fully qualified name (fqn) of the module that this + information contains info relevant to qconfig for + """ + + def __init__(self, module_fqn: str): + super().__init__() + self.module_fqn = module_fqn + + # populate this section with all the variables we might find important + # change from none if your detector is actually using this + self.is_activation_dynamic = False + self.is_weight_per_channel = False + + # equalization related options + self.is_equalization_recommended = False + + def generate_quantization_qconfig(self, module: torch.nn.Module) -> QConfig: + r""" + Args: + module (torch.nn.Module) The module we are generating + the qconfig for + + Returns the generated quantization QConfig according to what a valid configuration is + """ + # Apply suggestions to new qconfig + module_qconfig = default_qconfig + + # keep track of dynamic and per_channel recommendations + recommendations_list = [] + # append as if a list of combinations + recommendations_list.append((self.is_activation_dynamic, self.is_weight_per_channel)) + recommendations_list.append((self.is_activation_dynamic, False)) # only trying dynamic rec + recommendations_list.append((False, self.is_weight_per_channel)) # only trying dynamic + + # now we try each of the combinations + for rec in recommendations_list: + # rec[0] -> dynamic recommended + # rec[1] -> per channel recommended + activation = default_dynamic_quant_observer if rec[0] else default_observer + weight = default_per_channel_weight_observer if rec[1] else default_weight_observer + test_config = QConfig(activation, weight) + try: + _assert_valid_qconfig(test_config, module) + module_qconfig = test_config + break + except AssertionError: + # if not a valid configuration, we move on to the next one in priority + continue + + # return the QConfig chosen + return module_qconfig + + def generate_equalization_qconfig(self) -> EqualizationQConfig: + r""" + This returns the equalization configuration for a module. + + For now, it just returns the default, but as more equalization options become + possible, this method can get more fleshed out with more nuanced granularity. + + + Returns the generated equalization QConfig according to what a valid configuration is + """ + # in this case, we just return default equalization config + # we know this is valid because only valid modules would even + # have this option + return default_equalization_qconfig + +# Adding base class for detectors +class DetectorBase(ABC): + r""" Base Detector Module + Any detector class should derive from this class. + + Concrete detectors should follow the same general API, which includes: + - A method to calculate and return observer insertion points + - Should return both the fqns and the Observer class to insert + - A method to return a report based on the detector + - Should return a str-based report and dict info in Tuple[str,Dict] format + """ + + def __init__(self): + super().__init__() + self.detector_config_info = None + + @abstractmethod + def determine_observer_insert_points(self, model) -> Dict: + r""" + Args + model (nn.Module or subclass): model to find observer insertion points + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to a Dict. + This dict maps string keys to detector specific information + """ + pass + + @abstractmethod + def get_detector_name(self) -> str: + r""" Returns the name of the current detector """ + pass + + + @abstractmethod + def get_qconfig_info(self, model) -> Dict[str, DetectorQConfigInfo]: + r""" Returns the DetectorQConfigInfo for each module_fqn relevant + Args + model (nn.Module or subclass): model to find observer insertion points + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to: + A DetectorQConfigInfo with the information to generate a QConfig for a specific module + """ + pass + + def _get_targeting_node(self, prepared_fx_model: GraphModule, target_fqn: str) -> torch.fx.node.Node: + r""" + Takes in a GraphModule and the target_fqn and finds the node whose target is this fqn. + + If it's not found, it means it is most likely inside a fused layer + We just go one layer up in terms of the fqn we are searching for until we find parent node + If we get to empty string, then we know that it doesn't exist + + The reason for the recursion is that if the model that we are looking for got fused, + we will have module fqn as e.g. x.linear.0 but the graph will only have a node for the fused module, + which would have fqn as x.linear so they will not match. + To handle this, if we don't match, we then take off the last bit of the fqn e.g. x.linear.0 -> x.linear, + or more generally foo.bar.baz -> foo.bar and search again, this will allow us to locate the correct module + even in cases with fusion + + Args: + prepared_fx_model (GraphModule): The prepared Fx GraphModule + target_fqn (str): The fqn of the layer we are trying to target + + Returns the node object we are trying to add observers around + """ + for node in prepared_fx_model.graph.nodes: + # if the node's target is our target, return it + if node.target == target_fqn: + return node + + # getting here means node not found + # if no "." we are already at base and failed + parent_fqn_sep_index = target_fqn.rfind(".") + if parent_fqn_sep_index == -1: + raise ValueError("passed in target_fqn not found in graph's targets.") + else: + # recursively call it with parent fqn + return self._get_targeting_node(prepared_fx_model, target_fqn[:parent_fqn_sep_index]) + + @abstractmethod + def generate_detector_report(self, model) -> Tuple[str, Dict[str, Any]]: + r""" + Args + model (nn.Module or subclass): model to find observer insertion points + + Returns a Tuple of two elements: + Str: string report of the suggested improvements + Dict: contains useful data collected by the observer pertinent to this report + """ + pass + +class PerChannelDetector(DetectorBase): + r""" This class is used to detect if any Linear or Conv layers in a model utilize per_channel quantization. + Only Linear and Conv layers can use per_channel as of now so only these two are currently checked. + + per_channel quantization can lead to major benefits in the form of accuracy. + Therefore, if the backend used by the user supports it, it is recommended to use + + Args: + backend (str, optional): the backend the user wishes to use in production + Default value is current torch.backends.quantized.engine + """ + + # Keys for return dictionary + BACKEND_KEY = "backend" + PER_CHAN_SUPPORTED_KEY = "per_channel_quantization_supported" + PER_CHAN_USED_KEY = "per_channel_quantization_used" + + # Default map for representing supported per channel quantization modules for different backends + DEFAULT_BACKEND_PER_CHANNEL_SUPPORTED_MODULES: Dict[str, Set[Any]] = { + "fbgemm": {nn.Linear, nn.Conv1d, nn.Conv2d, nn.Conv3d, nnqat.Linear, nnqat.Conv1d, nnqat.Conv2d, nnqat.Conv3d}, + "qnnpack": {nn.Linear, nn.Conv1d, nn.Conv2d, nn.Conv3d, nnqat.Linear, nnqat.Conv1d, nnqat.Conv2d, nnqat.Conv3d}, + "onednn": {nn.Linear, nn.Conv1d, nn.Conv2d, nn.Conv3d, nnqat.Linear, nnqat.Conv1d, nnqat.Conv2d, nnqat.Conv3d}, + "x86": {nn.Linear, nn.Conv1d, nn.Conv2d, nn.Conv3d, nnqat.Linear, nnqat.Conv1d, nnqat.Conv2d, nnqat.Conv3d}, + } + + def __init__(self, backend: str = torch.backends.quantized.engine): + super().__init__() + + # store the backend information + self.backend_chosen = backend + self.supported_modules = set() + if self.backend_chosen in self.DEFAULT_BACKEND_PER_CHANNEL_SUPPORTED_MODULES: + self.supported_modules = self.DEFAULT_BACKEND_PER_CHANNEL_SUPPORTED_MODULES[self.backend_chosen] + else: + raise ValueError(f"Not configured to work with {self.backend_chosen}. Try a different default backend") + + def get_detector_name(self) -> str: + r""" returns the string name of this detector""" + return "per_channel_detector" + + def get_qconfig_info(self, model) -> Dict[str, DetectorQConfigInfo]: + r""" Returns the DetectorQConfigInfo for each module_fqn relevant + Args + model (nn.Module or subclass): model to find observer insertion points + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to: + A DetectorQConfigInfo with the information to generate a QConfig for a specific module + """ + # run the helper function to populate the dictionary + per_channel_info = self._detect_per_channel_helper(model) + + # we actually have a qconfig info object we are populating + module_fqn_to_detector_qconfig_info = {} + + for module_fqn in per_channel_info: + # create a detector info instance + detector_qconfig_info = DetectorQConfigInfo(module_fqn) + + # see if per channel quantization is supported + per_chan_supported: bool = per_channel_info[module_fqn][self.PER_CHAN_SUPPORTED_KEY] + detector_qconfig_info.is_weight_per_channel = per_chan_supported + module_fqn_to_detector_qconfig_info[module_fqn] = detector_qconfig_info + + return module_fqn_to_detector_qconfig_info + + def determine_observer_insert_points(self, model: nn.Module) -> Dict: + r""" + There is no observers inserted for the PerChannelDetector. + + Returns an empty dictionary since no observers are added or needed + """ + return {} + + + def _detect_per_channel_helper(self, model: nn.Module): + r""" + determines if per_channel quantization is supported in modules and submodules. + + Returns a dictionary in the higher level _detect_per_channel function. + Each entry maps the fully-qualified-name to information on whether per_channel quantization. + + Args: + model: The current module that is being checked to see if it is per_channel quantizable + + Returns dictionary mapping fqns to if per_channel quantization is possible + """ + # create dict we will return + per_channel_info: Dict = {} + + # get the fully qualified name and check if in list of modules to include and list of modules to ignore + for fqn, module in model.named_modules(): + + is_in_include_list = sum([isinstance(module, x) for x in self.supported_modules]) > 0 + + # check if the module per_channel is supported + # based on backend + per_channel_supported = False + + if is_in_include_list: + per_channel_supported = True + + # assert statement for MyPy + q_config_file = module.qconfig + assert isinstance(q_config_file, QConfig) + + # this object should either be fake quant or observer + q_or_s_obj = module.qconfig.weight.p.func() + assert isinstance(q_or_s_obj, (FakeQuantize, ObserverBase)) + + per_channel_used = False # will be true if found in qconfig + + if hasattr(q_or_s_obj, "ch_axis"): # then we know that per_channel quantization used + + # all fake quants have channel axis so need to check is_per_channel + if isinstance(q_or_s_obj, FakeQuantize): + if hasattr(q_or_s_obj, "is_per_channel") and q_or_s_obj.is_per_channel: + per_channel_used = True + elif isinstance(q_or_s_obj, ObserverBase): + # should be an observer otherwise + per_channel_used = True + else: + raise ValueError("Should be either observer or fake quant") + + per_channel_info[fqn] = { + self.PER_CHAN_SUPPORTED_KEY: per_channel_supported, + self.PER_CHAN_USED_KEY: per_channel_used, + self.BACKEND_KEY: self.backend_chosen + } + + return per_channel_info + + def generate_detector_report(self, model: nn.Module) -> Tuple[str, Dict[str, Any]]: + r"""Checks if any Linear or Conv layers in the model utilize per_channel quantization. + Only Linear and Conv layers can use per_channel as of now so only these two are currently checked. + + Looks at q_config format and backend to determine if per_channel can be utilized. + Uses the DEFAULT_BACKEND_PER_CHANNEL_SUPPORTED_MODULES structure to determine support + + Args: + model: The prepared and calibrated model we want to check if using per_channel + + Returns a tuple with two elements: + String report of potential actions to improve model (if per_channel quantization is available in backend) + Dictionary mapping per_channel quantizable elements to: + whether per_channel quantization is supported by the backend + if it is being utilized in the current model + """ + + # run the helper function to populate the dictionary + per_channel_info = self._detect_per_channel_helper(model) + + # String to let the user know of further optimizations + further_optims_str = f"Further Optimizations for backend {self.backend_chosen}: \n" + + optimizations_possible = False + for fqn in per_channel_info: + fqn_dict = per_channel_info[fqn] + if fqn_dict[self.PER_CHAN_SUPPORTED_KEY] and not fqn_dict[self.PER_CHAN_USED_KEY]: + optimizations_possible = True + further_optims_str += f"Module {fqn} can be configured to use per_channel quantization.\n" + + if optimizations_possible: + further_optims_str += ( + "To use per_channel quantization, make sure the qconfig has a per_channel weight observer." + ) + else: + further_optims_str += "No further per_channel optimizations possible." + + # return the string and the dictionary form of same information + return (further_optims_str, per_channel_info) + + +class DynamicStaticDetector(DetectorBase): + r""" + Determines whether dynamic or static quantization is more appropriate for a given module. + + Takes advantage of the ModelReportObserver that records range information. + Stationary distribution of data are strictly above tolerance level for the comparison statistic: + + S = average_batch_activation_range/epoch_activation_range + + Nonstationary distributions are below or at the tolerance level for this metric. + + If the distribution of data right after the module is non-stationary, recommend dynamic quantization + Otherwise recommend static quantization + + Args: + tolerance (float, optional): The threshold where S metric is stationary above and non-stationary otherwise. Default: 0.5 + """ + # names for the pre and post observers that are inserted + DEFAULT_PRE_OBSERVER_NAME = "model_report_pre_observer" + DEFAULT_POST_OBSERVER_NAME = "model_report_post_observer" + + # naming conventions for stationary vs non-stationary data + STATIONARY_STR = "stationary" + NON_STATIONARY_STR = "non-stationary" + + # naming for activation + INPUT_ACTIVATION_PREFIX = "input_activation_" + OUTPUT_ACTIVATION_PREFIX = "output_activation_" + + # naming conventions for the keys of the return module info + TOLERANCE_KEY = "dynamic_static_tolerance" + DEFAULT_DYNAMIC_REC_KEY = "dynamic_recommended" + PRE_OBS_COMP_STAT_KEY = INPUT_ACTIVATION_PREFIX + "dynamic_static_comp_stat" + POST_OBS_COMP_STAT_KEY = OUTPUT_ACTIVATION_PREFIX + "dynamic_static_comp_stat" + PRE_OBS_DATA_DIST_KEY = INPUT_ACTIVATION_PREFIX + "dynamic_static_data_classification" + POST_OBS_DATA_DIST_KEY = OUTPUT_ACTIVATION_PREFIX + "dynamic_static_data_classification" + IS_CURRENTLY_SUPPORTED_KEY = "is_dynamic_supported" + + # modules that are supported both dynamic and static for this report function + DEFAULT_DYNAMIC_STATIC_CHECK_SUPPORTED = {nn.Linear} + + # modules that will be supported soon for both + DEFAULT_DYNAMIC_STATIC_FUTURE_SUPPORTED = {nn.Conv1d, nn.Conv2d, nn.Conv3d} + + def __init__(self, tolerance=0.5): + super().__init__() + + # set tolerance level and initialize a set to keep track of useful fqn locations + self.tolerance = tolerance + self.useful_observer_fqns: Set[str] = set() + + def determine_observer_insert_points(self, prepared_fx_model: GraphModule) -> Dict[str, Dict[str, Any]]: + r""" + Determines where observers need to be inserted for the Dynamic vs Static detector. + For this detector, we want to place observers on either side of linear layers in the model. + + Currently inserts observers for: + linear layers + + Args: + prepared_fx_model (GraphModule): The prepared Fx GraphModule + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to a Dict with: + key "target_node" -> the node we are trying to observe with this observer (torch.fx.node.Node) + key "observer_to_insert" -> the observer we wish to insert (ObserverBase) + key "is_post_observer" -> True if this is meant to be a post-observer for target_node, False if pre-observer + key "observer_args" -> The arguments that are meant to be passed into the observer + """ + + # observer for this detector is ModelReportObserver + obs_ctr = ModelReportObserver + + # return dict + obs_fqn_to_info: Dict[str, Dict[str, Any]] = {} + + for fqn, module in prepared_fx_model.named_modules(): + # make sure module is supported + if self._is_supported(module, insert=True): + # if it's a supported type, we want to get node and add observer insert locations + targeted_node = self._get_targeting_node(prepared_fx_model, fqn) + + # add entry for pre-observer + pre_obs_fqn = fqn + "." + self.DEFAULT_PRE_OBSERVER_NAME + + obs_fqn_to_info[pre_obs_fqn] = { + DETECTOR_TARGET_NODE_KEY: targeted_node, + DETECTOR_OBS_TO_INSERT_KEY: obs_ctr(), + DETECTOR_IS_POST_OBS_KEY: False, + DETECTOR_OBS_ARGS_KEY: targeted_node.args + } + + # add entry for post-observer + post_obs_fqn = fqn + "." + self.DEFAULT_POST_OBSERVER_NAME + + obs_fqn_to_info[post_obs_fqn] = { + DETECTOR_TARGET_NODE_KEY: targeted_node, + DETECTOR_OBS_TO_INSERT_KEY: obs_ctr(), + DETECTOR_IS_POST_OBS_KEY: True, + DETECTOR_OBS_ARGS_KEY: (targeted_node,) + } + + return obs_fqn_to_info + + def get_detector_name(self) -> str: + r""" returns the string name of this detector""" + return "dynamic_vs_static_detector" + + + def get_qconfig_info(self, model) -> Dict[str, DetectorQConfigInfo]: + r""" Returns the DetectorQConfigInfo for each module_fqn relevant + Args + model (nn.Module or subclass): model to find observer insertion points + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to: + A DetectorQConfigInfo with the information to generate a QConfig for a specific module + """ + # run the helper function to populate the dictionary + dynamic_static_info = self._generate_dict_info(model) + + # we actually have a qconfig info object we are populating + module_fqn_to_detector_qconfig_info = {} + + for module_fqn in dynamic_static_info: + # create a detector info instance + detector_qconfig_info = DetectorQConfigInfo(module_fqn) + + # see if per channel quantization is supported + dynamic_static_recommended: bool = dynamic_static_info[module_fqn][self.DEFAULT_DYNAMIC_REC_KEY] + detector_qconfig_info.is_activation_dynamic = dynamic_static_recommended + module_fqn_to_detector_qconfig_info[module_fqn] = detector_qconfig_info + + return module_fqn_to_detector_qconfig_info + + def _is_supported(self, module: nn.Module, insert: bool = False) -> bool: + r"""Returns whether the given module is supported for observers + + Args + module: The module to check and ensure is supported + insert: True if this is check for observer insertion, false if for report gen + + Returns True if the module is supported by observer, False otherwise + """ + # check to see if module is of a supported type + is_supported_type = sum([isinstance(module, x) for x in self.DEFAULT_DYNAMIC_STATIC_CHECK_SUPPORTED]) > 0 + + # check if it will be supported + future_supported_type = sum([isinstance(module, x) for x in self.DEFAULT_DYNAMIC_STATIC_FUTURE_SUPPORTED]) > 0 + + # supported + supported = is_supported_type or future_supported_type + + # this is check for observer insertion + if insert: + return supported + else: + # this is for report gen and we also need to check if it contains observers + has_obs = hasattr(module, self.DEFAULT_PRE_OBSERVER_NAME) and hasattr(module, self.DEFAULT_POST_OBSERVER_NAME) + return supported and has_obs + + def _generate_dict_info(self, model: GraphModule) -> Dict[str, Any]: + r""" + Helper function for generate_detector_report that does the generation of the dictionary. + This process is done as specified in generate_detector_report documentation + + Args: + model (GraphModule): The prepared and calibrated GraphModule with inserted ModelReportObservers + + Returns a Dictionary mapping modules with ModelReportObservers around them to: + whether dynamic quantization is recommended + their S metric of input to module + whether input to module is stationary or non-stationary + their S metric of output of module + whether output of module is stationary or non-stationary + the tolerance level to decided whether input/output is stationary or non-stationary + whether it is currently supported or planned for the future + """ + # store modules dynamic vs static information + module_dynamic_static_info = {} + + # This for loop goes through the modules, and extracts all relevant information into module_dynamic_static_info + # This information primary includes whether the data distributions around a supported module is stationary or not + # Based on this, it is recorded whether dynamic or static quantization is recommended + + # loop through all submodules included nested ones + for fqn, module in model.named_modules(): + # if module is Linear has the ModelReportObserver attached to it + if self._is_supported(module): + # get pre and post observers for the module + pre_obs = getattr(module, self.DEFAULT_PRE_OBSERVER_NAME) + post_obs = getattr(module, self.DEFAULT_POST_OBSERVER_NAME) + + # get the statistics for each module + pre_stat = pre_obs.get_batch_to_epoch_ratio() + post_stat = post_obs.get_batch_to_epoch_ratio() + + # record module, pre and post stat, and whether to do dynamic or static based off it + # true if post observer data distribution is non-stationary, false if it's stationary + dynamic_recommended = post_stat <= self.tolerance + + # specify the classifications for whether data distributions considered stationary or non-stationary + pre_obs_dist_classif = self.STATIONARY_STR if pre_stat > self.tolerance else self.NON_STATIONARY_STR + post_obs_dist_classif = self.STATIONARY_STR if post_stat > self.tolerance else self.NON_STATIONARY_STR + + # check if current support or future support + is_supported_type = sum([isinstance(module, x) for x in self.DEFAULT_DYNAMIC_STATIC_CHECK_SUPPORTED]) > 0 + + # store the set of important information for this module + module_info = { + self.TOLERANCE_KEY: self.tolerance, + self.DEFAULT_DYNAMIC_REC_KEY: dynamic_recommended, + self.PRE_OBS_COMP_STAT_KEY: pre_stat, + self.PRE_OBS_DATA_DIST_KEY: pre_obs_dist_classif, + self.POST_OBS_COMP_STAT_KEY: post_stat, + self.POST_OBS_DATA_DIST_KEY: post_obs_dist_classif, + self.IS_CURRENTLY_SUPPORTED_KEY: is_supported_type, + } + + module_dynamic_static_info[fqn] = module_info + + return module_dynamic_static_info + + def generate_detector_report(self, model: GraphModule) -> Tuple[str, Dict[str, Any]]: + r""" + Determines whether dynamic or static quantization is more appropriate for a given module. + + Takes advantage of the ModelReportObserver that records range information. + Stationary distribution of data are strictly above tolerance level for the comparison statistic: + + S = average_batch_activation_range/epoch_activation_range + + Nonstationary distributions are below or at the tolerance level for this metric. + + If the distribution of data right after the module is non-stationary, recommend dynamic quantization + Otherwise recommend static quantization + + This will then generate suggestions for dynamic vs static quantization focused around Linear. + + Args: + model (GraphModule): The prepared and calibrated GraphModule with inserted ModelReportObservers + + Returns a tuple with two elements: + String report of of whether dynamic or static quantization is recommended for certain modules + Dictionary mapping modules with ModelReportObservers around them to: + whether dynamic quantization is recommended + their S metric of input to module + whether input to module is stationary or non-stationary + their S metric of output of module + whether output of module is stationary or non-stationary + the tolerance level to decided whether input/output is stationary or non-stationary + whether it is currently supported or planned for the future + """ + + # get the dictionary of the information to format the string report + module_dynamic_static_info = self._generate_dict_info(model) + + dynamic_vs_static_string = "Dynamic vs. Static Quantization suggestions: \n" + + modules_added: bool = False # check to make sure at least 1 module added. + + dynamic_benefit = " You will get more accurate results if you use dynamic quantization" + static_benefit = " You can increase model efficiency if you use static quantization" + future_support_str = ". This layer is not yet supported for dynamic quantization" + # This for loop goes through the information collected in module_dynamic_static_info and: + # Populates the string based report with the information from module_dynamic_static_info + # Compiles the complete report by appending relevant formatted strings + + for module_fqn in module_dynamic_static_info.keys(): + + # there is at least 1 module for suggestion + modules_added = True + module_info = module_dynamic_static_info[module_fqn] + suggestion_string_template = "For module {} it is suggested to use {} quantization because {}.\n" + + # decide what string formatting values will be + quantization_type = "" + quantization_reasoning = "the distribution of data before {} is {} and the distribution after is {}." + + benefit_str = "" + + # strings for if dynamic quantized per tensor is needed + recommend_per_tensor = ". We recommend to add a {} before this module if it is static." + rec_lay_to_add = "dynamic quantize per tensor layer" + dynamic_per_tensor_string = recommend_per_tensor.format(rec_lay_to_add) + dynamic_per_tensor_reasoning_string = ( + " This is because the input to this module has a non-stationary distribution" + ) + + # start composing explanation + if module_info[self.DEFAULT_DYNAMIC_REC_KEY]: + quantization_type = "dynamic" + # check if currently supported or future supported + benefit_str = dynamic_benefit + if not module_info[self.IS_CURRENTLY_SUPPORTED_KEY]: + benefit_str += future_support_str + else: + quantization_type = "static" + benefit_str = static_benefit + + # now set the quantization explanation string + quantization_reasoning = ( + quantization_reasoning.format( + module_fqn, module_info[self.PRE_OBS_DATA_DIST_KEY], module_info[self.POST_OBS_DATA_DIST_KEY] + ) + + benefit_str + ) + + # if we have a non-stationary input -> linear -> stationary we suggested static + # however, we want to also recommend they add a dynamic quantize per tensor right if this change is made + if ( + module_info[self.PRE_OBS_DATA_DIST_KEY] == self.NON_STATIONARY_STR + and module_info[self.POST_OBS_DATA_DIST_KEY] == self.STATIONARY_STR + ): + quantization_reasoning = ( + quantization_reasoning + dynamic_per_tensor_string + dynamic_per_tensor_reasoning_string + ) + + # format the overall suggestion string with the specific inputs + module_suggestion_string = suggestion_string_template.format( + module_fqn, quantization_type, quantization_reasoning + ) + + # append to overall suggestion + dynamic_vs_static_string += module_suggestion_string + + if not modules_added: + dynamic_vs_static_string += "No applicable layers for suggestions. Only linear and conv are valid.\n" + + # return the string as well as the dictionary of information + return (dynamic_vs_static_string, module_dynamic_static_info) + + +class InputWeightEqualizationDetector(DetectorBase): + r""" + Determines whether input-weight equalization can help improve quantization for certain modules. + + Specifically, this list of modules includes: + linear + conv + + Determines whether input-weight equalization is recommended based on the comp stat: + s_c = sqrt(w_c/W)/sqrt(i_c/I) + where: + w_c is range of weight for channel c, W is range of weight over all channels + i_c is range of input for channel c, I is range of input over all channels + + if s_c >= threshold or <= 1 / threshold, recommends input-weight equalization + + Args: + ratio_threshold (float): The threshold for s_c to determine if input-weight equalization is suggested + Should be between 0 and 1 (both non-inclusive) + ch_axis (int, optional): The channel axis being observed to determine input weight equalization + Default: 1 + + * :attr:`ratio_threshold`: The threshold for s_c to determine if input-weight equalization is suggested + Should be between 0 and 1 + + * :attr:`ch_axis`: The channel axis being observed to determine input weight equalization + + * :attr:`SUPPORTED_MODULES`: This specifies the modules that are supported for input-weight equalization + + * :attr:`DEFAULT_PRE_OBSERVER_NAME`: The name of the pre-observer to be inserted for this detector + """ + + SUPPORTED_MODULES: Set[Callable] = {nn.Linear, + nn.Conv1d, + nn.Conv2d, + nn.Conv3d, + nnqat.Linear, + nnqat.Conv1d, + nnqat.Conv2d, + nnqat.Conv3d} + + # names for the pre and post observers that are inserted + DEFAULT_PRE_OBSERVER_NAME: str = "model_report_pre_observer" + + # weight / activation prefix for each of the below info + WEIGHT_PREFIX = "weight_" + ACTIVATION_PREFIX = "input_activation_" + + # string names for keys of info dictionaries + PER_CHANNEL_MAX_KEY = "per_channel_max" + PER_CHANNEL_MIN_KEY = "per_channel_min" + GLOBAL_MAX_KEY = "global_max" + GLOBAL_MIN_KEY = "global_min" + + # keys for return dict of recommendations + RECOMMENDED_KEY = "input_weight_equalization_recommended" + COMP_METRIC_KEY = "input_weight_channel_comparison_metrics" + THRESHOLD_KEY = "input_weight_threshold" + CHANNEL_KEY = "input_weight_channel_axis" + + # default weight and info strings + WEIGHT_STR = "weight" + INPUT_STR = "input" + + # default for what ratio we recommend input weight + DEFAULT_RECOMMEND_INPUT_WEIGHT_CHANNEL_RATIO = 0.4 + + def __init__(self, ratio_threshold: float, ch_axis: int = 1): + # ensure passed in inputs are valid + if ratio_threshold <= 0 or ratio_threshold >= 1: + raise ValueError("Make sure threshold is > 0 and < 1") + + # initialize attributes based on args + self.ratio_threshold: float = ratio_threshold + self.ch_axis: int = ch_axis + + def _is_supported(self, module: nn.Module, insert: bool = False) -> bool: + r"""Returns whether the given module is supported for observers + + Args + module: The module to check and ensure is supported + insert: True if this is check for observer insertion, false if for report gen + + Returns True if the module is supported by observer, False otherwise + """ + # check to see if module is of a supported type + is_supported_type = sum([type(module) is x for x in self.SUPPORTED_MODULES]) > 0 + + # this is check for observer insertion + if insert: + return is_supported_type + else: + # this is for report gen and we also need to check if it contains observers + has_obs = hasattr(module, self.DEFAULT_PRE_OBSERVER_NAME) + return is_supported_type and has_obs + + def get_qconfig_info(self, model) -> Dict[str, DetectorQConfigInfo]: + r""" Returns the DetectorQConfigInfo for each module_fqn relevant + Args + model (nn.Module or subclass): model to find observer insertion points + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to: + A DetectorQConfigInfo with the information to generate a QConfig for a specific module + """ + # run the helper function to populate the dictionary + # find the range of inputs + input_values: Dict[str, Dict] = self._extract_input_info(model) + + # find the range of weights + weight_values: Dict[str, Dict] = self._extract_weight_info(model) + + # calculate per_channel comparison statistic s_c + comp_stats: Dict[str, torch.Tensor] = self._generate_comparison_values(input_values, weight_values) + + # generate the return dictionary + input_weight_equalization_info: Dict[str, Dict] = self._generate_dict_info(input_values, weight_values, comp_stats) + + # we actually have a qconfig info object we are populating + module_fqn_to_detector_qconfig_info = {} + + for module_fqn in input_weight_equalization_info: + # create a detector info instance + detector_qconfig_info = DetectorQConfigInfo(module_fqn) + + # see if per channel quantization is supported + input_weight_recommended: bool = input_weight_equalization_info[module_fqn][self.RECOMMENDED_KEY] + detector_qconfig_info.is_equalization_recommended = input_weight_recommended + module_fqn_to_detector_qconfig_info[module_fqn] = detector_qconfig_info + + return module_fqn_to_detector_qconfig_info + + def determine_observer_insert_points(self, prepared_fx_model: GraphModule) -> Dict[str, Dict[str, Any]]: + r"""Determines where observers need to be inserted for the Input Weight Equalization Detector. + For this detector, we want to place observers in front of supported layers. + + Currently inserts observers for: + linear layers + conv layers + + Args: + prepared_fx_model (GraphModule): The prepared Fx GraphModule + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to a Dict with: + key "target_node" -> the node we are trying to observe with this observer (torch.fx.node.Node) + key "observer_to_insert" -> the observer we wish to insert (ObserverBase) + key "is_post_observer" -> True if this is meant to be a post-observer for target_node, False if pre-observer + key "observer_args" -> The arguments that are meant to be passed into the observer + """ + + # observer for this detector is ModelReportObserver + obs_ctr = ModelReportObserver + + # return dict + obs_fqn_to_info: Dict[str, Dict[str, Any]] = {} + + for fqn, module in prepared_fx_model.named_modules(): + # check to see if module is of a supported type + if self._is_supported(module, insert=True): + # if it's a supported type, we want to get node and add observer insert locations + targeted_node = self._get_targeting_node(prepared_fx_model, fqn) + + # add entry for pre-observer + pre_obs_fqn = fqn + "." + self.DEFAULT_PRE_OBSERVER_NAME + + obs_fqn_to_info[pre_obs_fqn] = { + DETECTOR_TARGET_NODE_KEY: targeted_node, + DETECTOR_OBS_TO_INSERT_KEY: obs_ctr(ch_axis=self.ch_axis), + DETECTOR_IS_POST_OBS_KEY: False, + DETECTOR_OBS_ARGS_KEY: targeted_node.args, + } + + return obs_fqn_to_info + + def get_detector_name(self) -> str: + r"""Returns the name of this detector""" + return "input_weight_equalization_detector" + + def _extract_input_info(self, model: GraphModule) -> Dict[str, Dict]: + r""" + Takes in a calibrated GraphModule and then finds the relevant observers. + It then extracts the input information for each observer returns it + + Args + model (GraphModule): The prepared and calibrated GraphModule with inserted ModelReportObservers + + Returns a dict mapping relevant module fqns (str) to a dict with keys: + "input_activation_per_channel_max" : maps to the per_channel max values + "input_activation_per_channel_min" : maps to the per_channel min values + "input_activation_global_max" : maps to the global max recorded + "input_activation_global_min" : maps to the global min recorded + """ + + # return dictionary mapping observer fqns to desired info + input_info: Dict[str, Dict] = {} + + for fqn, module in model.named_modules(): + # if module is supported and it has a pre-observer + if self._is_supported(module): + # get pre observer for the module + pre_obs = getattr(module, self.DEFAULT_PRE_OBSERVER_NAME) + + input_info[fqn] = { + self.ACTIVATION_PREFIX + self.PER_CHANNEL_MAX_KEY: pre_obs.max_val, + self.ACTIVATION_PREFIX + self.PER_CHANNEL_MIN_KEY: pre_obs.min_val, + self.ACTIVATION_PREFIX + self.GLOBAL_MAX_KEY: max(pre_obs.max_val), + self.ACTIVATION_PREFIX + self.GLOBAL_MIN_KEY: min(pre_obs.min_val), + } + + return input_info + + def _extract_weight_info(self, model: GraphModule) -> Dict[str, Dict]: + r""" + Takes in a calibrated GraphModule and then finds the relevant observers. + It then extracts the weight information for each layer an observer is attached to. + + Args + model (GraphModule): The prepared and calibrated GraphModule with inserted ModelReportObservers + + Returns a dict mapping module fqns (str) to a dict with keys: + "per_channel_max" : maps to the per_channel max values + "per_channel_min" : maps to the per_channel min values + "global_max" : maps to the global max recorded + "global_min" : maps to the global min recorded + """ + # return dictionary mapping observer fqns to desired info + weight_info: Dict[str, Dict] = {} + + for fqn, module in model.named_modules(): + # if module is supported and it has a pre-observer + if self._is_supported(module): + # we don't need actual observer, just the module weights + # calculate min and max vals + device = module.weight.device + min_val: torch.Tensor = torch.tensor([float('inf')], device=device) + max_val: torch.Tensor = torch.tensor([float('-inf')], device=device) + x_copy = module.weight + x_dim = x_copy.size() + + new_axis_list = [i for i in range(len(x_dim))] # noqa: C416 + new_axis_list[self.ch_axis] = 0 + new_axis_list[0] = self.ch_axis + y = x_copy.permute(new_axis_list) + + # Need to match dtype of min/max because the updates to buffers + # are done in place and types need to match for comparisons + y = y.to(min_val.dtype) + y = torch.flatten(y, start_dim=1) + if min_val.numel() == 0 or max_val.numel() == 0: + min_val, max_val = torch.aminmax(y, dim=1) + else: + min_val_cur, max_val_cur = torch.aminmax(y, dim=1) + min_val = torch.min(min_val_cur, min_val) + max_val = torch.max(max_val_cur, max_val) + + weight_info[fqn] = { + self.WEIGHT_PREFIX + self.PER_CHANNEL_MAX_KEY: max_val, + self.WEIGHT_PREFIX + self.PER_CHANNEL_MIN_KEY: min_val, + self.WEIGHT_PREFIX + self.GLOBAL_MAX_KEY: max(max_val), + self.WEIGHT_PREFIX + self.GLOBAL_MIN_KEY: min(min_val), + } + + return weight_info + + def _calculate_range_ratio(self, info_dict: Dict, info_str: str, module_fqn: str) -> torch.Tensor: + r""" + Takes in an info dict and calculates the s_c matrix. + + Args: + info_dict (dict): A dictionary of either input or weight range info + info_str (str): A str describing whether currently looking at weight or input info + Either "weight" or "input" + module_fqn (str): The fqn of the module we are looking at + + Returns a tensor of values, where each value is the s_c stat for a different channel + """ + # calculate the ratios of the info + # get the prefix str + prefix_str = self.ACTIVATION_PREFIX if info_str == self.INPUT_STR else self.WEIGHT_PREFIX + + per_channel_range = info_dict[prefix_str + self.PER_CHANNEL_MAX_KEY] - info_dict[prefix_str + self.PER_CHANNEL_MIN_KEY] + global_range = info_dict[prefix_str + self.GLOBAL_MAX_KEY] - info_dict[prefix_str + self.GLOBAL_MIN_KEY] + + if global_range == 0: + range_zero_explanation = "We recommend removing this channel as it doesn't provide any useful information." + raise ValueError( + "The range of the {} data for module {} is 0, which means you have a constant value channel. {}".format( + info_str, module_fqn, range_zero_explanation + ) + ) + + ratio = per_channel_range / global_range + + return ratio + + def _generate_comparison_values(self, input_info: Dict, weight_info: Dict) -> Dict[str, torch.Tensor]: + r""" + Takes in the information on the min and max values of the inputs and weights and: + Calculates the comp stat for each channel: s_c = sqrt(w_c/W)/sqrt(i_c/I) + + Args: + input_info (dict): A dict mapping each observer to input range information + weight_info (dict): A dict mapping each observer to weight range information + + Returns a dict mapping relevant observer fqns (str) to a 1-D tensor. + Each value is a different s_c value for a different channel + """ + # create return dictionary for each observer + module_fqn_to_channel: Dict[str, torch.Tensor] = {} + + # for each module (both passed in dicts should have same keys) + for module_fqn in input_info: + + # raise error if not in weight info + if module_fqn not in weight_info: + raise KeyError(f"Unable to find weight range stats for module {module_fqn}") + + # calculate the ratios of the weight info and input info + weight_ratio = self._calculate_range_ratio(weight_info[module_fqn], self.WEIGHT_STR, module_fqn) + input_ratio = self._calculate_range_ratio(input_info[module_fqn], self.INPUT_STR, module_fqn) + + # if mismatched size, because of grouping, we want to replicate weight enough times + weight_channels = len(weight_ratio) + input_channels = len(input_ratio) + if weight_channels != input_channels: + # we try to replicate + assert input_channels % weight_channels == 0, "input channels should be divisible by weight channels." + # get replication factor + rep_factor: int = input_channels // weight_channels + + # weight ratio is (n,), input ratio is (k,), we just repeat weight ratio k // n + weight_ratio = weight_ratio.repeat(rep_factor) + + # calculate the s metric per channel + s = torch.sqrt(weight_ratio) / torch.sqrt(input_ratio) + module_fqn_to_channel[module_fqn] = s + + # return compiled observer ratios + return module_fqn_to_channel + + def _generate_dict_info(self, input_info: Dict, weight_info: Dict, comp_stats: Dict) -> Dict[str, Dict]: + r""" + Helper function for generate_detector_report that does the generation of the dictionary. + This process is done as specified in generate_detector_report documentation + + Args: + input_info (dict): A dict mapping each module to input range information + weight_info (dict): A dict mapping each module to weight range information + comp_stats (dict): A dict mapping each module to its corresponding comp stat + + Returns a dictionary mapping each module with relevant ModelReportObservers around them to: + whether input weight equalization is recommended + their s_c metric compared to the threshold + the threshold used to make the recommendation + the channel used for recording data + the input channel range info + the weight channel range info + """ + # store modules input weight equalization info + input_weight_equalization_info: Dict[str, Dict] = {} + + # for each module we add separate set of suggestions + for module_fqn in input_info: + + # get relevant info for this module + mod_input_info: Dict = input_info[module_fqn] + mod_weight_info: Dict = weight_info[module_fqn] + mod_comp_stat: Dict = comp_stats[module_fqn] + + # decide if each channel should have input weight equalization or not + channel_rec_vals: list = [] + + for val in mod_comp_stat: + float_rep: float = val.item() + + # decide if recommending input weight equalization + recommended: bool = float_rep >= self.ratio_threshold and float_rep <= 1 / self.ratio_threshold + channel_rec_vals.append(recommended) + + # build the return dict input + # also unpack input and weight dicts into it + input_weight_equalization_info[module_fqn] = { + self.RECOMMENDED_KEY: channel_rec_vals, + self.COMP_METRIC_KEY: mod_comp_stat, + self.THRESHOLD_KEY: self.ratio_threshold, + self.CHANNEL_KEY: self.ch_axis, + **mod_input_info, + **mod_weight_info, + } + + # return our compiled info for each module + return input_weight_equalization_info + + def generate_detector_report(self, model: GraphModule) -> Tuple[str, Dict[str, Any]]: + r""" + Determines whether input weight equalization is appropriate for a given module. + + Takes advantage of the ModelReport Observer which records per channel information of input range + It then uses the passed in weight info inconjunction to compute the desired ratio + Finally, it gives suggestions based on this information for each module of interest + + Args: + model (GraphModule): The prepared and calibrated GraphModule with inserted ModelReportObservers + + Returns a tuple with two elements: + String report of of whether input weight equalization is recommended for certain modules + Dictionary mapping modules of interest to: + whether input weight equalization is recommended + their s_c metric compared to the threshold + the threshold used to make the recommendation + the channel used for recording data + the input channel range info + the weight channel range info + """ + + # find the range of inputs + input_values: Dict[str, Dict] = self._extract_input_info(model) + + # find the range of weights + weight_values: Dict[str, Dict] = self._extract_weight_info(model) + + # calculate per_channel comparison statistic s_c + comp_stats: Dict[str, torch.Tensor] = self._generate_comparison_values(input_values, weight_values) + + # generate the return dictionary + input_weight_equalization_info: Dict[str, Dict] = self._generate_dict_info(input_values, weight_values, comp_stats) + + # now we can generate report based on this information + input_weight_string = "Input-Weight Equalization suggestions: \n" + + # some strings to be formatted depending on module we are adding + module_suggestion_str = "For Module {} looked at with axis {}: \n" + channel_suggestion_str = "\tWe suggest {} input weight equalization because {}\n" + use_str = "to use" + no_use_str = "to not use" + input_weight_benefit_str = "{}/{} channels would benefit and we expect significant reduction in quantization error." + input_weight_non_benefit_reasoning = "{}/{} channels benefitting from input-weight equalization being applied." + input_weight_non_benefit_str = "we don't expect much improvement from input-weight equalization based on {}" + + # added module check + added_module: bool = False + + # compile the suggestion string + for module_fqn in input_weight_equalization_info: + # we added at least 1 module + added_module = True + # add the module level description + input_weight_string += module_suggestion_str.format(module_fqn, self.ch_axis) + + mod_info: Dict[str, Any] = input_weight_equalization_info[module_fqn] + + # gather info on how many channels would benefit from input weight and + recommendation_per_channel: torch.Tensor = mod_info[self.RECOMMENDED_KEY] + num_recs = sum(recommendation_per_channel) + + if num_recs / len(recommendation_per_channel) >= self.DEFAULT_RECOMMEND_INPUT_WEIGHT_CHANNEL_RATIO: + input_benefit_formatted = input_weight_benefit_str.format(num_recs, len(recommendation_per_channel)) + channel_str = channel_suggestion_str.format(use_str, input_benefit_formatted) + input_weight_string += channel_str + else: + non_benefit_reason_formatted = input_weight_non_benefit_reasoning.format(num_recs, len(recommendation_per_channel)) + non_benefit_str = input_weight_non_benefit_str.format(non_benefit_reason_formatted) + channel_str = channel_suggestion_str.format(no_use_str, non_benefit_str) + input_weight_string += channel_str + + # if no modules looked at, amend return string + if not added_module: + input_weight_string += "No applicable layers for suggestions. Only linear and conv valid.\n" + + # return a tuple with the string explanation and the compiled dict info + return (input_weight_string, input_weight_equalization_info) + + +class OutlierDetector(DetectorBase): + r""" + Determines whether there are significant outliers in activation data around a certain layer. + + This is ideally used in conjunction with information on stationary vs. non-stationary distribution: + If the data is stationary, and there are significant outliers, then we want to flag them + We want to do this on a per channel basis for detecting outliers + + Determines whether activation data is flagged as outlier based on if data is stationary and: + p_r = avg(100th percentile / "reference_percentile"th percentile) + where: + p_r is average percentile ratio across all batches in the epoch + reference_percentile is a percentile values between 0 and 100 exclusive + + if p_r is above some threshold, then we consider the activations to have significant outliers + + Args: + ratio_threshold (float, optional): The threshold for p_r to determine if there are outliers in activations + Should be >= 1 + Default: 3.5 + reference_percentile (float, optional): The denominator to find the relative scale of the 100th percentile + Should be between 0 and 1 + Default: 0.975 + fraction_batches_used_threshold (float, optional): Threshold of fraction of batches per channel to determine outlier + If fraction is below this, we deem number of samples used to calculate outliers as insignificant and alert user + regardless of whether we detected outliers or not in channel to take a closer look at channel results + Should be between 0 and 1 + Default: 0.95 + ch_axis (int, optional): The channel axis being observed to determine input weight equalization + Default: 1 + + * :attr:`ratio_threshold`: The threshold for p_r to determine if there are outliers in activations + The p_r value (average ratio of 100th percentile/reference_percentile) is compared to ratio_threshold + If it is significantly greater, then we consider it an outlier + This threshold was calculated based on the ratio of the percentiles in a normal distribution + The calculations behind value choice: https://drive.google.com/file/d/1N2wdtXWI-kOH8S7HH4-PYB_NmqzZil4p/view?usp=sharing + + * :attr:`reference_percentile`: The denominator of the top fraction to find the relative scale of the 100th percentile + Should be between 0 and 1 + The calculations behind value choice: https://drive.google.com/file/d/1N2wdtXWI-kOH8S7HH4-PYB_NmqzZil4p/view?usp=sharing + + * :attr:`fraction_batches_used_threshold`: The fraction of batches to determine outliers for each channel should be above this + Some batches may not be used because of 0-based errors, so this is to ensure a good amount of the total batches are used + Should be between 0 and 1 + + * :attr:`ch_axis`: The channel axis being observed to determine outliers + + * :attr:`DEFAULT_PRE_OBSERVER_NAME`: The name of the pre-observer to be inserted for this detector + """ + + # names for the pre observers that are inserted + DEFAULT_PRE_OBSERVER_NAME: str = "model_report_pre_observer" + + # pre activation prefix + INPUT_ACTIVATION_PREFIX = "input_activation_" + + # names for dict keys + OUTLIER_KEY = "outliers_detected" + NUM_BATCHES_KEY = "outlier_detection_batches_used" + IS_SUFFICIENT_BATCHES_KEY = "outlier_detection_is_sufficient_batches" + COMP_METRIC_KEY = "outlier_detection_percentile_ratios" + RATIO_THRES_KEY = "outlier_detection_ratio_threshold" + REF_PERCENTILE_KEY = "outlier_detection_reference_percentile" + CHANNEL_AXIS_KEY = "outlier_detection_channel_axis" + MAX_VALS_KEY = INPUT_ACTIVATION_PREFIX + "per_channel_max" + CONSTANT_COUNTS_KEY = "constant_batch_counts" + + def __init__( + self, + ratio_threshold: float = 3.5, + reference_percentile: float = 0.975, + fraction_batches_used_threshold: float = 0.95, + ch_axis: int = 1, + ): + # initialize the variables of interest + self.ratio_threshold = ratio_threshold + + # make sure passed in percentile is valid + assert reference_percentile >= 0 and reference_percentile <= 1 + assert fraction_batches_used_threshold >= 0 and fraction_batches_used_threshold <= 1 + self.reference_percentile = reference_percentile + self.fraction_batches_used_threshold = fraction_batches_used_threshold + self.ch_axis = ch_axis + + def get_detector_name(self) -> str: + r"""Returns the name of this detector""" + return "outlier_detector" + + def _supports_insertion(self, module: nn.Module) -> bool: + r"""Returns whether the given module is supported for observers insertion + + Any module that doesn't have children and isn't an observer itself is supported + + Args + module: The module to check and ensure is supported + + Returns True if the module is supported by observer, False otherwise + """ + # case for insertion of module + # check if the module has any children and isn't observer + num_children = len(list(module.children())) + return num_children == 0 and not _is_activation_post_process(module) + + def get_qconfig_info(self, model) -> Dict[str, DetectorQConfigInfo]: + r""" Returns the DetectorQConfigInfo for each module_fqn relevant + Args + model (nn.Module or subclass): model to find observer insertion points + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to: + A DetectorQConfigInfo with the information to generate a QConfig for a specific module + """ + # currently doesn't do anything for outlier detector + return {} + + def _supports_report_gen(self, module: nn.Module) -> bool: + r"""Returns whether the given module is supported for report generation + + Any module that has a model report pre-observer is supported + + Args + module: The module to check and ensure is supported + + Returns True if the module is supported by observer, False otherwise + """ + return hasattr(module, self.DEFAULT_PRE_OBSERVER_NAME) + + def determine_observer_insert_points(self, prepared_fx_model: GraphModule) -> Dict[str, Dict[str, Any]]: + r""" Determines where observers need to be inserted for the Outlier Detector. + + For this detector, we want to place observers in front of supported layers. + + Currently inserts observers for: + all layers that do not have children (leaf level layers) + + Args: + prepared_fx_model (GraphModule): The prepared Fx GraphModule + + Returns a Dict mapping from unique observer fqns (where we want to insert them) to a Dict with: + key "target_node" -> the node we are trying to observe with this observer (torch.fx.node.Node) + key "observer_to_insert" -> the observer we wish to insert (ObserverBase) + key "is_post_observer" -> True if this is meant to be a post-observer for target_node, False if pre-observer + key "observer_args" -> The arguments that are meant to be passed into the observer + """ + # observer for this detector is ModelReportObserver + obs_ctr = ModelReportObserver + + # return dict + obs_fqn_to_info: Dict[str, Dict[str, Any]] = {} + + for fqn, module in prepared_fx_model.named_modules(): + # check to see if module is of a supported type + if self._supports_insertion(module): + # if it's a supported type, we want to get node and add observer insert locations + targeted_node = self._get_targeting_node(prepared_fx_model, fqn) + + # add entry for pre-observer + pre_obs_fqn = fqn + "." + self.DEFAULT_PRE_OBSERVER_NAME + + obs_fqn_to_info[pre_obs_fqn] = { + DETECTOR_TARGET_NODE_KEY: targeted_node, + DETECTOR_OBS_TO_INSERT_KEY: obs_ctr(ch_axis=self.ch_axis, comp_percentile=self.reference_percentile), + DETECTOR_IS_POST_OBS_KEY: False, + DETECTOR_OBS_ARGS_KEY: targeted_node.args, + } + + return obs_fqn_to_info + + def _calculate_outlier_info( + self, + percentile_ratios: torch.Tensor, + counted_batches: torch.Tensor, + total_batches: int, + ) -> Dict[str, List[bool]]: + r""" + Gives info on whether the percentile ratios calculated would be considered outliers + Also gives information on whether the collected data is statistically significant to make this claim + + Args: + percentile_ratios (torch.Tensor): The average percentile_ratios per channel calculated by the observer + counted_batches (torch.Tensor): The number of batches used for average calculation per tensor + total_batches (int): The total number of batches that passed through observer in this epoch + + Returns a dictionary mapping: + "outliers_detected" : list of bools per channel that are true if it is considered an outlier + "is_sufficient_batches": if o_r was >= fraction_batches_used_threshold: + where o_r = counted_batches / total_batches + """ + outlier_dict: Dict[str, List[bool]] = {self.OUTLIER_KEY: [], self.IS_SUFFICIENT_BATCHES_KEY: []} + + # get both as flattened lists for easy mapping + ratios_list: List = percentile_ratios.tolist() + num_batches_list: List = counted_batches.tolist() + + # calculate whether channels were statistically significant + significant_size = [ + batch_size / total_batches >= self.fraction_batches_used_threshold for batch_size in num_batches_list + ] + outlier_dict[self.IS_SUFFICIENT_BATCHES_KEY] = significant_size + + # calculate for each channel whether it's an outlier or not based on ratio + outlier_detected = [ratio > self.ratio_threshold for ratio in ratios_list] + outlier_dict[self.OUTLIER_KEY] = outlier_detected + + # return the dictionary with the two lists + return outlier_dict + + def _generate_info_dict(self, model: GraphModule) -> Dict[str, Dict]: + r""" + Helper function for generate_detector_report that does the generation of the dictionary. + This process is done as specified in generate_detector_report documentation + + Args: + model (GraphModule): The prepared and calibrated GraphModule with inserted ModelReportObservers + + Returns a dict mapping relevant module fqns to: + whether there were outliers found in activation before + the number of batches used for each channel + whether fraction of applicable batches used is above fraction_batches_used_threshold + their p_r metric compared to the threshold + the threshold used to make the recommendation + the reference_percentile used to make the recommendation + the channel axis used to determine individual channels + the constant batch counts per channel + the per channel max values + """ + # return dictionary mapping observer fqns to desired info + info_dict: Dict[str, Dict] = {} + + for fqn, module in model.named_modules(): + # if module is supported and it has a pre-observer + if self._supports_report_gen(module): + # get pre observer for the module + pre_obs: ModelReportObserver = getattr(module, self.DEFAULT_PRE_OBSERVER_NAME) + + # get the number of batches and calculated ratio thresholds + num_batches: torch.Tensor = pre_obs.percentile_batches_tracked + average_ratios: torch.Tensor = pre_obs.average_percentile_ratio + channel_batch_cnts: torch.Tensor = pre_obs.constant_channels + total_batches: int = pre_obs.num_batches_tracked + + # also get the max values + max_vals: torch.Tensor = pre_obs.max_val + + # we have to specifically modify how we are recording negative ratio for pre-relu layers + for index, ratio_val in enumerate(average_ratios): + # check if we have a negative ratio + # a ratio might be negative if we have a situation where the 100th percentile is + # > 0 while the nth percentile is < 0, in which case this would not be detected + # as an outlier. Since we care more about magnitude, we make it positive. + if ratio_val.item() < 0: + # first make it positive + average_ratios[index] = -ratio_val + + if ratio_val.item() < 1: + # if it's less than 1 we have the flip it as well + average_ratios[index] = 1 / ratio_val + + outlier_calcs = self._calculate_outlier_info(average_ratios, num_batches, total_batches) + + # calculate whether ratios were outliers + info_dict[fqn] = { + self.CHANNEL_AXIS_KEY: self.ch_axis, + self.REF_PERCENTILE_KEY: self.reference_percentile, + self.RATIO_THRES_KEY: self.ratio_threshold, + self.COMP_METRIC_KEY: average_ratios, + self.NUM_BATCHES_KEY: num_batches, + self.OUTLIER_KEY: outlier_calcs[self.OUTLIER_KEY], + self.IS_SUFFICIENT_BATCHES_KEY: outlier_calcs[self.IS_SUFFICIENT_BATCHES_KEY], + self.CONSTANT_COUNTS_KEY: channel_batch_cnts, + self.MAX_VALS_KEY: max_vals + } + + return info_dict + + def generate_detector_report(self, model: GraphModule) -> Tuple[str, Dict[str, Any]]: + r""" + Determines whether input weight equalization is appropriate for a given module. + + Takes advantage of the ModelReport Observer which records the relevant percentile information + + Args: + model (GraphModule): The prepared and calibrated GraphModule with inserted ModelReportObservers + + Returns a tuple with two elements: + String report of of whether there are outliers in the activations around certain modules + Dictionary mapping modules of interest to: + whether there were outliers found in activation before + the number of batches used for each channel + whether fraction of applicable batches used is above fraction_batches_used_threshold + their p_r metric compared to the threshold + the threshold used to make the recommendation + the reference_percentile used to make the recommendation + the channel axis used to determine individual channels + the constant batch counts per channel + the per channel max values + """ + # generate the information dictionary of outlier information + info_dict = self._generate_info_dict(model) + + # now we can generate report based on this information + outlier_string = "Outlier detection report: \n" + + # added module check + added_module: bool = False + + # some strings to be formatted depending on module we are adding + module_suggestion_str = "For Module {} looked at with axis {}: \n" + channel_suggestion_str = "\tFor channel {}, we found outliers in the preceding activation data with {}.\n" + channel_max_value_str = "a max value across all batches of {}" + note_string = "Note: outlier detection is only reliable for {}. We recommend {} to ensure the most accurate results." + note_distribution = "stationary distributions" + note_rec = "running the static vs. dynamic detector to ensure activation data before modules above is stationary" + + # suggestion for constant batch check since that can make it no outliers + constant_str = "\tFor channel {}, we found {} constant value batches. {}\n" + constant_suggestion = "We recommend taking a look at the dict and data to see how frequent this occurred and why." + + # compile the suggestion string + for module_fqn in info_dict: + # get module specific info + mod_info: Dict[str, Any] = info_dict[module_fqn] + # check to see if we already added high level model desc + added_model_desc = False + # look at each individual channel and add a suggestion + for index, outlier_detected in enumerate(mod_info[self.OUTLIER_KEY]): + if outlier_detected: + # we found at least 1 outlier + if not added_model_desc: + # add the module level description + outlier_string += module_suggestion_str.format(module_fqn, self.ch_axis) + added_model_desc = True + + # we mark that we found at least one outlier + added_module = True + max_value_found_str = channel_max_value_str.format(mod_info[self.MAX_VALS_KEY][index]) + channel_str = channel_suggestion_str.format(index, max_value_found_str) + outlier_string += channel_str + + # also check if we found constant batch + if mod_info[self.CONSTANT_COUNTS_KEY][index] != 0: + # make sure we add a module level highlight. + if not added_model_desc: + # add the module level description + outlier_string += module_suggestion_str.format(module_fqn, self.ch_axis) + added_model_desc = True + + constant_values_for_channel = mod_info[self.CONSTANT_COUNTS_KEY][index] + formatted_str = constant_str.format(index, constant_values_for_channel, constant_suggestion) + outlier_string += formatted_str + # we also added at least one thing to description + added_module = True + + + # if found outlier, give suggestion, else give default response + if added_module: + # compose the note string + note_composed = note_string.format(note_distribution, note_rec) + outlier_string += note_composed + else: + outlier_string += "There were no outliers found in the activations.\n" + + return (outlier_string, info_dict) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/fuse.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/fuse.py new file mode 100644 index 0000000000000000000000000000000000000000..91b876997d10910e5b411225c2654857eab07f2b --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/fuse.py @@ -0,0 +1,161 @@ +from torch.fx import ( + GraphModule, + Node, + map_arg +) +from torch.fx.graph import Graph +from .match_utils import ( + _is_match, + MatchAllNode, +) +from .pattern_utils import ( + _sorted_patterns_dict, +) + +from ..backend_config import ( + BackendConfig, + get_native_backend_config, +) +from ..backend_config.utils import ( + get_fuser_method_mapping, + get_fusion_pattern_to_root_node_getter, + get_fusion_pattern_to_extra_inputs_getter, +) + +from .custom_config import FuseCustomConfig + +from .fuse_handler import ( + _get_fusion_pattern_to_fuse_handler_cls, + FuseHandler, +) + +from typing import Any, Callable, Dict, List, Tuple, Union +import warnings + +from torch.ao.quantization.utils import Pattern, NodePattern + + +__all__ = [ + "fuse", + # TODO: We should make this private in the future + # This is currently needed for test_public_bindings for some reason + "FuseHandler", +] + + +def fuse( + model: GraphModule, + is_qat: bool, + fuse_custom_config: Union[FuseCustomConfig, Dict[str, Any], None] = None, + backend_config: Union[BackendConfig, Dict[str, Any], None] = None, +) -> GraphModule: + if fuse_custom_config is None: + fuse_custom_config = FuseCustomConfig() + + if isinstance(fuse_custom_config, Dict): + warnings.warn( + "Passing a fuse_custom_config_dict to fuse is deprecated and will not be supported " + "in a future version. Please pass in a FuseCustomConfig instead.") + fuse_custom_config = FuseCustomConfig.from_dict(fuse_custom_config) + + if isinstance(backend_config, Dict): + warnings.warn( + "Passing a backend_config_dict to prepare is deprecated and will not be supported " + "in a future version. Please pass in a BackendConfig instead.") + backend_config = BackendConfig.from_dict(backend_config) + + named_modules = dict(model.named_modules()) + + if backend_config is None: + backend_config = get_native_backend_config() + + fusion_pattern_to_fuse_handler_cls = _sorted_patterns_dict(_get_fusion_pattern_to_fuse_handler_cls(backend_config)) + fuser_method_mapping = get_fuser_method_mapping(backend_config) + fusion_pattern_to_root_node_getter = get_fusion_pattern_to_root_node_getter(backend_config) + fusion_pattern_to_extra_inputs_getter = get_fusion_pattern_to_extra_inputs_getter(backend_config) + + # find fusion + fusion_pairs = _find_matches( + model, model.graph, fusion_pattern_to_fuse_handler_cls) + # TODO: change this to inplace changes to graph, since we no longer construct + # new GraphModule anymore + fused_graph = Graph() + env: Dict[Any, Any] = {} + + def load_arg(a): + return map_arg(a, lambda node: env[node.name]) + + def default_root_node_getter(node_pattern): + while not isinstance(node_pattern[-1], Node): + node_pattern = node_pattern[-1] + return node_pattern[-1] + + for node in model.graph.nodes: + maybe_last_node, pattern, matched_node_pattern, obj, node_to_subpattern = \ + fusion_pairs.get(node.name, (None, None, None, None, None)) + # get the corresponding subpattern for the current node + if node_to_subpattern is not None: + node_subpattern = node_to_subpattern.get(node, None) + else: + node_subpattern = None + if maybe_last_node is node: + assert obj is not None + root_node_getter = fusion_pattern_to_root_node_getter.get(pattern, default_root_node_getter) + root_node = root_node_getter(matched_node_pattern) # type: ignore[index] + extra_inputs_getter = fusion_pattern_to_extra_inputs_getter.get(pattern, None) + extra_inputs = [] + if extra_inputs_getter is not None: + extra_inputs = extra_inputs_getter(matched_node_pattern) + # TODO: add validation that root_node is a module and has the same type + # as the root_module in the configuration + env[node.name] = obj.fuse( + load_arg, named_modules, fused_graph, root_node, extra_inputs, matched_node_pattern, # type: ignore[arg-type] + fuse_custom_config, fuser_method_mapping, is_qat) + elif maybe_last_node is None or node_subpattern is MatchAllNode: + env[node.name] = fused_graph.node_copy(node, load_arg) + # node matched in patterns and is not root is removed here + + model = GraphModule(model, fused_graph) + return model + +def _find_matches( + root: GraphModule, + graph: Graph, + pattern_to_fuse_handler_cls: Dict[Pattern, Callable], +) -> Dict[str, Tuple[Node, Pattern, NodePattern, FuseHandler, Dict[Node, Any]]]: + modules = dict(root.named_modules()) + # node name -> (root_node, match_value) + match_map : Dict[ + str, Tuple[Node, Pattern, NodePattern, FuseHandler, Dict[Node, Any]]] = {} + # a map from node to the matched subpattern + node_to_subpattern: Dict[Node, Any] = {} + + # TODO: dedup with quantization matching function in match_utils.py + def apply_match(pattern, node, match, matched_node_pattern, node_to_subpattern): + if isinstance(pattern, tuple): + s, *args = pattern + current_node_pattern: List[Node] = [] + apply_match(s, node, match, current_node_pattern, node_to_subpattern) + for subpattern, arg in zip(args, node.args): + apply_match(subpattern, arg, match, current_node_pattern, node_to_subpattern) + matched_node_pattern.append(tuple(current_node_pattern)) + else: + # the first pattern matches will take precedence + if node.name not in match_map: + matched_node_pattern.append(node) + # MatchAllNode here is actually MatchAllInputNode which should not + # be added to match_map + if pattern is not MatchAllNode: + node_to_subpattern[node] = pattern + root_node, pattern, handler = match + match_map[node.name] = (root_node, pattern, matched_node_pattern, handler, node_to_subpattern) + + for node in reversed(graph.nodes): + if node.name not in match_map: + for pattern, fuse_handler_cls in pattern_to_fuse_handler_cls.items(): + matched_node_pattern: List[Node] = [] + if _is_match(modules, node, pattern): + apply_match(pattern, node, (node, pattern, fuse_handler_cls(node)), matched_node_pattern, node_to_subpattern) + break + + return match_map diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/graph_module.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/graph_module.py new file mode 100644 index 0000000000000000000000000000000000000000..cc9187285ae6313b07e03fe47e0eaec8ca4a265b --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/graph_module.py @@ -0,0 +1,119 @@ +import torch +import copy +from torch.fx import GraphModule +from torch.fx.graph import Graph +from typing import Union, Dict, Any, Set + +__all__ = [ + "FusedGraphModule", + "ObservedGraphModule", + "ObservedStandaloneGraphModule", + "QuantizedGraphModule", +] + +class FusedGraphModule(GraphModule): + def __init__(self, root: Union[torch.nn.Module, Dict[str, Any]], graph: Graph, preserved_attr_names: Set[str]): + self.preserved_attr_names = preserved_attr_names + preserved_attrs = {attr: getattr(root, attr) for attr in self.preserved_attr_names if hasattr(root, attr)} + super().__init__(root, graph) + for attr in preserved_attrs: + setattr(self, attr, preserved_attrs[attr]) + + # GraphModule does not copy attributes which are not in the __dict__ + # of vanilla nn.Module. So, we override __deepcopy__ in order + # to copy the quantization specific attributes correctly. + def __deepcopy__(self, memo): + fake_mod = torch.nn.Module() + fake_mod.__dict__ = copy.deepcopy(self.__dict__) + return FusedGraphModule(fake_mod, copy.deepcopy(self.graph), copy.deepcopy(self.preserved_attr_names)) + +class ObservedGraphModule(GraphModule): + + def __init__(self, root: Union[torch.nn.Module, Dict[str, Any]], graph: Graph, preserved_attr_names: Set[str]): + self.preserved_attr_names = { + '_activation_post_process_map', + '_activation_post_process_indexes', + '_patterns', + '_node_name_to_qconfig', + '_prepare_custom_config', + '_equalization_node_name_to_qconfig', + '_node_name_to_scope', + '_qconfig_mapping', + '_is_qat', + '_observed_node_names'}.union(preserved_attr_names) + preserved_attrs = {attr: getattr(root, attr) for attr in self.preserved_attr_names if hasattr(root, attr)} + super().__init__(root, graph) + for attr in preserved_attrs: + setattr(self, attr, preserved_attrs[attr]) + + # GraphModule does not copy attributes which are not in the __dict__ + # of vanilla nn.Module. So, we override __deepcopy__ in order + # to copy the quantization specific attributes correctly. + def __deepcopy__(self, memo): + fake_mod = torch.nn.Module() + fake_mod.__dict__ = copy.deepcopy(self.__dict__) + return ObservedGraphModule(fake_mod, copy.deepcopy(self.graph), copy.deepcopy(self.preserved_attr_names)) + +def _is_observed_module(module: Any) -> bool: + return hasattr(module, "meta") and "_observed_graph_module_attrs" in module.meta + +def _get_observed_graph_module_attr(model: Union[torch.nn.Module, GraphModule], attr_name: str) -> Any: + if hasattr(model, "meta") and "_observed_graph_module_attrs" in model.meta: # type: ignore[operator, index] + return getattr(model.meta["_observed_graph_module_attrs"], attr_name) # type: ignore[index] + return None + +class ObservedStandaloneGraphModule(ObservedGraphModule): + def __init__(self, root: Union[torch.nn.Module, Dict[str, Any]], graph: Graph, preserved_attr_names: Set[str]): + preserved_attr_names = preserved_attr_names.union({ + "_standalone_module_input_quantized_idxs", + "_standalone_module_output_quantized_idxs"}) + super().__init__(root, graph, preserved_attr_names) + + def __deepcopy__(self, memo): + fake_mod = torch.nn.Module() + fake_mod.__dict__ = copy.deepcopy(self.__dict__) + return ObservedStandaloneGraphModule(fake_mod, copy.deepcopy(self.graph), copy.deepcopy(self.preserved_attr_names)) + +def _is_observed_standalone_module(module: Any) -> bool: + return _is_observed_module(module) and module.meta["_observed_graph_module_attrs"].is_observed_standalone_module + +def _save_packed_weight(self, destination, prefix, keep_vars): + for attr_name in dir(self): + if "_packed_weight" in attr_name and \ + isinstance(getattr(self, attr_name), torch._C.ScriptObject): # type: ignore[attr-defined] + packed_weight = getattr(self, attr_name) + destination[prefix + attr_name] = packed_weight + +class QuantizedGraphModule(GraphModule): + """ This class is created to make sure PackedParams + (e.g. LinearPackedParams, Conv2dPackedParams) to appear in state_dict + so that we can serialize and deserialize quantized graph module with + torch.save(m.state_dict()) and m.load_state_dict(state_dict) + """ + def __init__(self, root: Union[torch.nn.Module, Dict[str, Any]], graph: Graph, preserved_attr_names: Set[str]): + self.preserved_attr_names = preserved_attr_names + preserved_attrs = {attr: getattr(root, attr) for attr in self.preserved_attr_names if hasattr(root, attr)} + super().__init__(root, graph) + for attr in preserved_attrs: + setattr(self, attr, preserved_attrs[attr]) + self._register_state_dict_hook(_save_packed_weight) + + def _load_from_state_dict(self, state_dict, prefix, local_metadata, strict, + missing_keys, unexpected_keys, error_msgs): + attrs_to_pop = [] + for attr_name in state_dict: + if attr_name.startswith("_packed_weight") and isinstance(state_dict[attr_name], torch._C.ScriptObject): # type: ignore[attr-defined] # noqa: B950 + setattr(self, attr_name, state_dict[attr_name]) + attrs_to_pop.append(attr_name) + + # pop the packed param attributesn + for attr_name in attrs_to_pop: + state_dict.pop(attr_name) + + super()._load_from_state_dict(state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) + + + def __deepcopy__(self, memo): + fake_mod = torch.nn.Module() + fake_mod.__dict__ = copy.deepcopy(self.__dict__) + return QuantizedGraphModule(fake_mod, copy.deepcopy(self.graph), copy.deepcopy(self.preserved_attr_names)) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/lower_to_fbgemm.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/lower_to_fbgemm.py new file mode 100644 index 0000000000000000000000000000000000000000..ef58652b1adda0dc135fbef21afe789d6f538eda --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/lower_to_fbgemm.py @@ -0,0 +1,16 @@ +from ._lower_to_native_backend import _lower_to_native_backend +from ..qconfig import QConfigAny +from torch.fx import GraphModule +from typing import Dict, Tuple + +__all__ = ['lower_to_fbgemm'] + +def lower_to_fbgemm( + model: GraphModule, + qconfig_map: Dict[str, QConfigAny], + node_name_to_scope: Dict[str, Tuple[str, type]] +) -> GraphModule: + """ Lower a quantized reference model (with reference quantized operator patterns) + to fbgemm + """ + return _lower_to_native_backend(model, qconfig_map, node_name_to_scope) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/lstm_utils.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/lstm_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..9f163a1869ac1dc12ed2dca4a59a698482afc2f2 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/lstm_utils.py @@ -0,0 +1,183 @@ +import copy +import operator +import torch +from typing import Any, Callable, Optional, Tuple +from torch.ao.quantization import ( + default_weight_observer, + default_weight_fake_quant, + FakeQuantizeBase, + QConfig, + QConfigMapping, +) +from torch.ao.quantization.backend_config import BackendConfig +from torch.ao.quantization.observer import _PartialWrapper +from torch.ao.quantization.quantize_fx import ( + convert_to_reference_fx, + prepare_fx, +) + +# TODO: move all LSTM util functions from fx/utils.py to this file +def _get_lstm_with_individually_observed_parts( + float_lstm: torch.nn.LSTM, + example_inputs: Tuple[Any, ...], + backend_config: Optional[BackendConfig] = None, + linear_output_obs_ctr: Optional[_PartialWrapper] = None, + sigmoid_obs_ctr: Optional[_PartialWrapper] = None, + tanh_obs_ctr: Optional[_PartialWrapper] = None, + cell_state_obs_ctr: Optional[_PartialWrapper] = None, + hidden_state_obs_ctr: Optional[_PartialWrapper] = None, +) -> torch.ao.nn.quantizable.LSTM: + """ + Return an observed `torch.ao.nn.quantizable.LSTM` created from a `torch.nn.LSTM` + with specific observers or fake quantizes assigned to the inner ops or submodules. + + In both eager and FX graph mode quantization, `torch.ao.nn.quantizable.LSTM` is + used as an observed custom module, which is responsible for inserting its own + observers. By default, all inner ops inherit the parent custom module's QConfig. + Users who wish to override this behavior may extend `torch.ao.nn.quantizable.LSTM` + and use this helper function to customize the observer insertion logic. + + This is meant to be used to convert a float module to an observed module in the + custom module flow. + + Args: + `float_lstm`: The float LSTM module + `example_inputs`: example inputs for the forward function of the LSTM module + `backend_config`: BackendConfig to use to observe the LSTM module + `linear_output_obs_ctr`: observer or fake quantize for linear outputs Wx + b, + where W is the weight matrix, b is the bias, and x is either the inputs + or the hidden state from the previous layer (if any) + `sigmoid_obs_ctr`: observer or fake quantize for sigmoid activations + `tanh_obs_ctr`: observer or fake quantize for tanh activations + `cell_state_obs_ctr`: observer or fake quantize for the cell state + `hidden_state_obs_ctr`: observer or fake quantize for the hidden state and + the output + + Return: + A `torch.ao.nn.quantizable.LSTM` with the specified observers or fake quantizes + assigned to the inner ops. + """ + def make_qconfig(obs_ctr: _PartialWrapper) -> QConfig: + """ + Make a QConfig with fixed qparams observers or fake quantizes. + """ + if isinstance(obs_ctr(), FakeQuantizeBase): + weight = default_weight_fake_quant + else: + weight = default_weight_observer + return QConfig(activation=obs_ctr, weight=weight) + + quantizable_lstm = torch.ao.nn.quantizable.LSTM( + float_lstm.input_size, float_lstm.hidden_size, float_lstm.num_layers, float_lstm.bias, + float_lstm.batch_first, float_lstm.dropout, float_lstm.bidirectional) + quantizable_lstm.qconfig = float_lstm.qconfig + + for idx in range(float_lstm.num_layers): + quantizable_lstm.layers[idx] = torch.ao.nn.quantizable.modules.rnn._LSTMLayer.from_float(float_lstm, + idx, + float_lstm.qconfig, + batch_first=False) + + # Build QConfigMapping for the LSTM cell + # Note: FloatFunctional qconfigs will be configured separately below + cell_qm = QConfigMapping().set_global(float_lstm.qconfig) # type: ignore[arg-type] + if sigmoid_obs_ctr is not None: + cell_qm.set_module_name("input_gate", make_qconfig(sigmoid_obs_ctr)) + cell_qm.set_module_name("forget_gate", make_qconfig(sigmoid_obs_ctr)) + cell_qm.set_module_name("output_gate", make_qconfig(sigmoid_obs_ctr)) + if tanh_obs_ctr is not None: + cell_qm.set_module_name("cell_gate", make_qconfig(tanh_obs_ctr)) + + # Insert observers into each LSTM cell + # TODO: maybe make this work for layer_bw as well + for layer in quantizable_lstm.layers: + cell = layer.layer_fw.cell + cell = prepare_fx(cell, cell_qm, example_inputs, backend_config=backend_config) + # HACK: Manually replace the activation_post_process following these ops. + # This is needed for FloatFunctional ops because there is currently no way + # to configure these ops in FX graph mode quantization today. This is because + # the FloatFunctional modules simply disappear from the graph after tracing. + # In the future, we should rewrite quantizable LSTM without FloatFunctionals. + op_index_to_activation_post_process_ctr = { + (torch.add, 0): linear_output_obs_ctr, # gates.add + (torch.mul, 0): cell_state_obs_ctr, # fgate_cx.mul + (torch.mul, 1): cell_state_obs_ctr, # igate_cgate.mul + (torch.add, 1): cell_state_obs_ctr, # fgate_cx_igate_cgate.add + (torch.mul, 2): hidden_state_obs_ctr, # ogate_cy.mul + } + add_count = 0 + mul_count = 0 + for node in cell.graph.nodes: + op_index: Optional[Tuple[Callable, int]] = None # e.g. (torch.add, 1) + if node.target == torch.add: + op_index = (torch.add, add_count) + add_count += 1 + elif node.target == torch.mul: + op_index = (torch.mul, mul_count) + mul_count += 1 + else: + # Neither torch.add nor torch.mul + continue + if op_index not in op_index_to_activation_post_process_ctr: + continue + assert len(node.users) == 1 + activation_post_process_name = next(iter(node.users.keys())).name + activation_post_process_ctr = op_index_to_activation_post_process_ctr[op_index] + if activation_post_process_ctr is not None: + setattr(cell, activation_post_process_name, activation_post_process_ctr()) + layer.layer_fw.cell = cell + return quantizable_lstm + +def _get_reference_quantized_lstm_module( + observed_lstm: torch.ao.nn.quantizable.LSTM, + backend_config: Optional[BackendConfig] = None, +) -> torch.ao.nn.quantized.LSTM: + """ + Return a `torch.ao.nn.quantized.LSTM` created from a `torch.ao.nn.quantizable.LSTM` + with observers or fake quantizes inserted through `prepare_fx`, e.g. from + `_get_lstm_with_individually_observed_parts`. + + This is meant to be used to convert an observed module to a quantized module in the + custom module flow. + + Args: + `observed_lstm`: a `torch.ao.nn.quantizable.LSTM` observed through `prepare_fx` + `backend_config`: BackendConfig to use to produce the reference quantized model + + Return: + A reference `torch.ao.nn.quantized.LSTM` module. + """ + quantized_lstm = torch.ao.nn.quantized.LSTM( + observed_lstm.input_size, observed_lstm.hidden_size, observed_lstm.num_layers, + observed_lstm.bias, observed_lstm.batch_first, observed_lstm.dropout, + observed_lstm.bidirectional) + + for i, layer in enumerate(quantized_lstm.layers): + cell = copy.deepcopy(observed_lstm.layers.get_submodule(str(i)).layer_fw.cell) # type: ignore[union-attr] + cell = convert_to_reference_fx(cell, backend_config=backend_config) # type: ignore[arg-type] + assert isinstance(cell, torch.fx.GraphModule) + # HACK: Manually remove input quantize nodes and output dequantize nodes, + # since custom modules expect quint8 inputs and outputs for now. Note that + # this functionality is supposedly handled through PrepareCustomConfig's + # `set_input_quantized_indexes` and `set_output_quantized_indexes`, but that + # API doesn't currently handle tuple inputs and outputs, so we have to do + # this manually for now. In the future we should (1) relax the restriction + # on custom module input/output dtypes, and (2) expand support for complex + # input/output structures. + for node in cell.graph.nodes: + if node.target == torch.quantize_per_tensor: + arg = node.args[0] + # Remove quantize(x), quantize(hidden[0]), and quantize(hidden[1]) + if arg.target == "x" or (arg.target == operator.getitem and arg.args[0].target == "hidden"): + with cell.graph.inserting_before(node): + node.replace_all_uses_with(arg) + cell.graph.erase_node(node) + if node.target == "output": + # Remove all dequantize nodes in the output tuple + for arg in node.args[0]: + with cell.graph.inserting_before(node): + node.replace_input_with(arg, arg.args[0]) + cell.graph.eliminate_dead_code() + cell.recompile() + layer.layer_fw.cell = cell + return quantized_lstm diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/match_utils.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/match_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..cf287db8c5245453afc795565f130ed64080674d --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/match_utils.py @@ -0,0 +1,237 @@ +import sys +import torch +from torch.fx.graph import ( + Graph, + Node, +) +from torch.ao.quantization.utils import Pattern +from .quantize_handler import ( + QuantizeHandler, +) +from ..qconfig import ( + QConfigAny, +) +from ..utils import ( + MatchAllNode +) +from .graph_module import ( + _is_observed_standalone_module, +) +from torch.nn.utils.parametrize import type_before_parametrizations +from typing import Any, Dict, List, Callable, Optional, Tuple, Type, Set, Iterable + + +__all__: List[str] = [] + +# TODO(future PR): the 1st argument is typed as `List[Node]`, but a better type +# would be a recursive `List[Union[Node, Tuple[Union[Node, ...]]]]` +_MatchResult = Tuple[Node, List[Node], Optional[Pattern], QuantizeHandler] + +_MatchResultWithQConfig = Tuple[Node, List[Node], Optional[Pattern], QuantizeHandler, + QConfigAny] + +# Note: The order of patterns is important! match function will take whatever is matched first, so we'll +# need to put the fusion patterns before single patterns. For example, add_relu should be registered come before relu. +# decorators are applied in the reverse order we see. Also when we match the nodes in the graph with these patterns, +# we'll start from the last node of the graph and traverse back. +def _is_match(modules, node, pattern, max_uses=sys.maxsize): + """ Matches a node in fx against a pattern + """ + if isinstance(pattern, tuple): + self_match, *arg_matches = pattern + if self_match is getattr: + assert len(pattern) == 2, 'Expecting getattr pattern to have two elements' + arg_matches = [] + else: + self_match = pattern + arg_matches = [] + + if isinstance(self_match, type) and issubclass(self_match, MatchAllNode): + return True + + if node == pattern: + return True + + if not isinstance(node, Node) or len(node.users) > max_uses: + return False + + if isinstance(self_match, type) and issubclass(self_match, torch.nn.Module): + if node.op != 'call_module': + return False + if not type_before_parametrizations(modules[node.target]) == self_match: + return False + elif callable(self_match): + if node.op != 'call_function' or node.target is not self_match: + return False + elif node.target is getattr: + if node.args[1] != pattern[1]: + return False + elif isinstance(self_match, str): + if node.op != 'call_method' or node.target != self_match: + return False + elif node.target != self_match: + return False + + if not arg_matches: + return True + + if len(arg_matches) != len(node.args): + return False + + return all(_is_match(modules, node, arg_match, max_uses=1) for node, arg_match in zip(node.args, arg_matches)) + +def _find_matches( + graph: Graph, + modules: Dict[str, torch.nn.Module], + patterns: Dict[Pattern, QuantizeHandler], + root_node_getter_mapping: Dict[Pattern, Callable], + standalone_module_names: Optional[List[str]] = None, + standalone_module_classes: Optional[List[Type]] = None, + custom_module_classes: Optional[List[Any]] = None) -> Dict[str, _MatchResult]: + """ + Matches the nodes in the input graph to quantization patterns, and + outputs the information needed to quantize them in future steps. + + Inputs: + - graph: an fx.Graph object + - modules: a mapping of fully qualified module name to instance, + for example, {'foo': ModuleFoo, ...} + - patterns: a mapping from a tuple of nodes in reverse order to + uninitialized QuantizeHandler subclass. + + Outputs a map of + node_name -> + (node, matched_values, matched_pattern, QuantizeHandler instance, + qconfig) + + For example, { + 'relu_1': (relu_1, [relu_1], torch.nn.functional.relu, + , QConfig(...)), + ... + } + """ + if custom_module_classes is None: + custom_module_classes = [] + + if standalone_module_classes is None: + standalone_module_classes = [] + + if standalone_module_names is None: + standalone_module_names = [] + + match_map: Dict[str, _MatchResult] = {} + all_matched : Set[str] = set() + + def _recursive_record_node_in_match_map( + last_node, + match_map, + node_pattern, + matched_node_pattern, + pattern, + match_value): + if isinstance(node_pattern, Node): + match_map[node_pattern.name] = ( + last_node, matched_node_pattern, pattern, match_value) + elif not isinstance(node_pattern, Iterable): + return + else: + for n in node_pattern: + _recursive_record_node_in_match_map(last_node, match_map, n, matched_node_pattern, pattern, match_value) + + # TODO: 1. merge with fuse matcher 2. document the code + def record_match( + pattern, + node, + last_node, + matched_node_pattern, + match_map): + if isinstance(pattern, tuple): + s, *args = pattern + is_single_arg = len(args) == 1 + current_node_pattern: List[Node] = [] + record_match( + s, + node, + last_node, + matched_node_pattern, + match_map) + if pattern[0] is not getattr: + for subpattern, arg in zip(args, node.args): + record_match( + subpattern, + arg, + node, + current_node_pattern, + match_map) + if len(current_node_pattern) > 1: + # current_node_pattern is the node pattern we get from matching + # the subpattern with arguments of the node + # we use is_single_arg to recover the original structure of the pattern + # if the original pattern has a single argument, we will have + # (original_op, (original_arg, ...)) + # otherwise, we'll have a list of arguments + # (original_op, arg0, arg1, arg2, ...) + if is_single_arg: + matched_node_pattern.append(tuple(current_node_pattern)) + else: + matched_node_pattern.extend(list(current_node_pattern)) + else: + matched_node_pattern.append(current_node_pattern[0]) + else: + matched_node_pattern.append(node) + + for node in reversed(graph.nodes): + if node.name not in match_map and node.name not in all_matched: + for pattern, quantize_handler_cls in patterns.items(): + root_node_getter = root_node_getter_mapping.get(pattern, None) + if _is_match(modules, node, pattern) and node.name not in match_map: + matched_node_pattern: List[Node] = [] + record_match( + pattern, + node, + node, + matched_node_pattern, + match_map) + quantize_handler = quantize_handler_cls( # type: ignore[operator] + matched_node_pattern, + modules, + root_node_getter) + last_node = node + # record the match for all nodes in the pattern + _recursive_record_node_in_match_map( + last_node, + match_map, + # we need to record all nodes in the matched pattern in the match_map + matched_node_pattern, + # this is a part of the value corresponding to the node + matched_node_pattern, + pattern, + quantize_handler) + break + + # add custom module instances to the match result + assert modules is not None + for node in graph.nodes: + if node.op == 'call_module' and \ + type(modules[node.target]) in custom_module_classes: + match_map[node.name] = ( + node, node, None, QuantizeHandler(node, modules, is_custom_module=True)) + + def is_standalone_module(node_target: str, modules: Dict[str, torch.nn.Module]): + assert modules is not None + return ( + node_target in standalone_module_names or # type: ignore[operator] + type(modules[node_target]) in standalone_module_classes # type: ignore[operator] + ) + + # add standalone modules to the match + for node in graph.nodes: + if node.op == 'call_module' and \ + (is_standalone_module(node.target, modules) or + _is_observed_standalone_module(modules[node.target])): + # add node to matched nodes + match_map[node.name] = ( + node, node, None, + QuantizeHandler(node, modules, is_standalone_module=True)) + + return match_map diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/qconfig_mapping_utils.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/qconfig_mapping_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..0b906a1777de0168511190a4d4d6ec4442a36b99 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/qconfig_mapping_utils.py @@ -0,0 +1,343 @@ +import torch +import re +from collections import defaultdict, OrderedDict +from typing import Callable, Any, Dict, Tuple, Set, List, Union +from torch.ao.quantization import QConfig +from torch.ao.quantization.qconfig import _add_module_to_qconfig_obs_ctr, QConfigAny, qconfig_equals +from torch.ao.quantization.observer import ( + _is_activation_post_process, +) +from torch.ao.quantization.backend_config import ( + BackendConfig, + DTypeConfig, +) +from torch.ao.quantization.backend_config.utils import ( + get_module_to_qat_module, +) + +from torch.fx import ( + GraphModule, +) +from torch.fx.graph import ( + Graph, +) +from torch.ao.nn.intrinsic import _FusedModule + +from ..utils import ( + _parent_name, + get_qconfig_dtypes, +) +from ..qconfig_mapping import ( + _OBJECT_TYPE_DICT_KEY, + _MODULE_NAME_DICT_KEY, + _MODULE_NAME_REGEX_DICT_KEY, + QConfigMapping, +) + +__all__: List[str] = [] + + + +def _maybe_adjust_qconfig_for_module_name_object_type_order( + qconfig_mapping: QConfigMapping, + cur_module_path: str, + cur_object_type: Callable, + cur_object_type_idx: int, + fallback_qconfig: QConfigAny, +) -> QConfigAny: + for (module_name, object_type, index), qconfig in qconfig_mapping.module_name_object_type_order_qconfigs.items(): + if ( + (module_name == cur_module_path) and + (object_type == cur_object_type) and + (index == cur_object_type_idx) + ): + return qconfig + return fallback_qconfig + + +def _update_qconfig_for_fusion(model: GraphModule, qconfig_mapping: QConfigMapping): + """ + Update the QConfigMapping to account for fused modules such as LinearReLU. + This assumes the QConfigMapping's attributes have already been converted to OrderedDicts. + """ + object_type_dict = qconfig_mapping.object_type_qconfigs + if len(object_type_dict) == 0: + return qconfig_mapping + + modules = dict(model.named_modules()) + + for node in model.graph.nodes: + if node.op == 'call_module' and node.target in modules: + maybe_fused_module = modules[str(node.target)] + if not isinstance(maybe_fused_module, _FusedModule): + continue + + ops = list(maybe_fused_module._modules.values()) + fused_qconfig = object_type_dict.get(type(ops[0]), None) + + # Raise an error if the modules in the fused module have + # different qconfigs specified in the qconfig_dict + # TODO: currently it only works for modules, + # need to make this work for torch.nn.functional.relu + # TODO: currently it only works for object_type configurations, + # ideally it should work for different types of configurations, + # maybe we want to redesign this part + for op in ops[1:]: + if not qconfig_equals(object_type_dict.get(type(op), None), fused_qconfig): + raise LookupError( + "During fusion, we need to specify the same " + + f"qconfigs for all module types in {type(maybe_fused_module)} " + + f"offending type: {type(op)}") + + if fused_qconfig is not None: + object_type_dict[type(maybe_fused_module)] = fused_qconfig + +def _generate_node_name_to_qconfig( + root: torch.nn.Module, + modules: Dict[str, torch.nn.Module], + input_graph: Graph, + qconfig_mapping: QConfigMapping, + node_name_to_scope: Dict[str, Tuple[str, type]]) -> Dict[str, QConfigAny]: + global_qconfig = qconfig_mapping.global_qconfig + node_name_to_qconfig = {} + + # example: + # + # {'foo.bar': {F.linear: 0, F.conv2d: 1, ...}, ...} + # + # meaning in submodule 'foo.bar', we have seen 0 F.linear and + # 1 F.conv2d invocations so far. + submodule_to_object_type_to_cur_idx: Dict[str, Dict[Callable, int]] = \ + defaultdict(lambda: defaultdict(int)) + for node in input_graph.nodes: + qconfig = None + if node.op == "get_attr": + module_name, _ = _parent_name(node.target) + qconfig = _maybe_adjust_qconfig_for_module_type_or_name( + qconfig_mapping, type(modules[module_name]), module_name, global_qconfig) + qconfig_with_device_check = _add_module_to_qconfig_obs_ctr(qconfig, modules.get(node.target, None)) + elif node.op == "call_function": + # precedence: module_name_qconfig + # > function_qconfig > global_qconfig + # module_name takes precedence over function qconfig + function_qconfig = _get_object_type_qconfig( + qconfig_mapping, node.target, global_qconfig) + module_path, module_type = node_name_to_scope[node.name] + qconfig = _maybe_adjust_qconfig_for_module_type_or_name( + qconfig_mapping, module_type, module_path, function_qconfig) + + cur_object_type_idx = \ + submodule_to_object_type_to_cur_idx[module_path][node.target] + submodule_to_object_type_to_cur_idx[module_path][node.target] += 1 + qconfig = _maybe_adjust_qconfig_for_module_name_object_type_order( + qconfig_mapping, module_path, node.target, cur_object_type_idx, qconfig) + qconfig_with_device_check = _add_module_to_qconfig_obs_ctr(qconfig, modules.get(node.target, None)) + + elif node.op == "call_method": + module_path, module_type = node_name_to_scope[node.name] + # first use node.target (string) to get the qconfig + # this is to support configs like + # "object_type": [("reshape", qconfig)] + qconfig = _maybe_adjust_qconfig_for_module_type_or_name( + qconfig_mapping, node.target, module_path, global_qconfig) + # if there is no special config for the method, we'll fall back to the + # config for the module that contains the call_method node + qconfig = _maybe_adjust_qconfig_for_module_type_or_name( + qconfig_mapping, module_type, module_path, qconfig) + # currently call_method does not support modifying qconfig + # by order, we can add this later if it is needed. + qconfig_with_device_check = _add_module_to_qconfig_obs_ctr(qconfig, modules.get(node.target, None)) + + elif node.op == 'call_module': + # if the node is an observer, just continue - don't add it to the qconfig_map + if _is_activation_post_process(modules[node.target]): + continue + qconfig = _maybe_adjust_qconfig_for_module_type_or_name( + qconfig_mapping, type(modules[node.target]), node.target, global_qconfig) + + module_path, module_type = node_name_to_scope[node.name] + # Note: for call_module, the module_path is the current module's name. + # to meaningfully count invocations, we need to count them in the parent + # module. + parent_name, _ = _parent_name(module_path) + cur_object_type_idx = \ + submodule_to_object_type_to_cur_idx[parent_name][module_type] + submodule_to_object_type_to_cur_idx[parent_name][module_type] += 1 + qconfig = _maybe_adjust_qconfig_for_module_name_object_type_order( + qconfig_mapping, parent_name, module_type, cur_object_type_idx, + qconfig) + qconfig_with_device_check = _add_module_to_qconfig_obs_ctr(qconfig, modules.get(node.target, None)) + + # regex is not supported eager mode propagate_qconfig_, we'll + # need to set the qconfig explicitly here in case regex + # is used + modules[node.target].qconfig = qconfig_with_device_check + else: + qconfig_with_device_check = None + + node_name_to_qconfig[node.name] = qconfig_with_device_check + return node_name_to_qconfig + + +def _check_is_valid_config_dict(config_dict: Any, allowed_keys: Set[str], dict_name: str) -> None: + r""" Checks if the given config_dict has the correct keys + + Args: + `config_dict`: dictionary whose keys we want to check + """ + + for k in config_dict.keys(): + if k not in allowed_keys: + raise ValueError( + 'Expected ' + dict_name + ' to have the following keys: ' + + str(allowed_keys) + '. But found \'' + k + + '\' instead.') + + +def _compare_prepare_convert_qconfig_mappings( + prepare_qconfig_mapping: QConfigMapping, + convert_qconfig_mapping: QConfigMapping): + r""" Compare the qconfig_mapping passed in convert to the one from prepare and check the values + + Args: + `prepare_qconfig_mapping`: configuration for prepare quantization step + `convert_qconfig_mapping`: configuration for convert quantization step + """ + assert qconfig_equals(prepare_qconfig_mapping.global_qconfig, convert_qconfig_mapping.global_qconfig), \ + "Expected global qconfigs to be the same in the prepare and convert quantization configs" + prepare_dicts: List[OrderedDict] = [ + prepare_qconfig_mapping.object_type_qconfigs, + prepare_qconfig_mapping.module_name_qconfigs, + prepare_qconfig_mapping.module_name_regex_qconfigs, + ] + convert_dicts: List[OrderedDict] = [ + convert_qconfig_mapping.object_type_qconfigs, + convert_qconfig_mapping.module_name_qconfigs, + convert_qconfig_mapping.module_name_regex_qconfigs, + ] + dict_names = [_OBJECT_TYPE_DICT_KEY, _MODULE_NAME_DICT_KEY, _MODULE_NAME_REGEX_DICT_KEY] + for i in range(len(prepare_dicts)): + for name in prepare_dicts[i].keys(): + assert name in convert_dicts[i], f"Missing key {dict_names[i]} {name} in convert QConfigMapping \ + when it was present in prepare" + assert convert_dicts[i][name] is None \ + or qconfig_equals(prepare_dicts[i][name], convert_dicts[i][name]), \ + f"Expected convert QConfigMapping to have the same qconfig as prepare for key {dict_names[i]} {name}; \ + prepare: {prepare_dicts[i][name]}; convert: {convert_dicts[i][name]}" + +def _is_qconfig_supported_by_dtype_configs(qconfig: QConfig, dtype_configs: List[DTypeConfig]): + for dtype_config in dtype_configs: + is_dynamic = dtype_config.is_dynamic + if is_dynamic is None: + is_dynamic = False + input_dtype = dtype_config.input_dtype or torch.float + weight_dtype = dtype_config.weight_dtype or torch.float + bias_dtype = dtype_config.bias_dtype or torch.float + output_dtype = dtype_config.output_dtype or torch.float + qconfig_activation_dtype, qconfig_weight_dtype, qconfig_input_act_is_dynamic = \ + get_qconfig_dtypes(qconfig) + qconfig_bias_dtype = torch.float16 \ + if ( + qconfig_activation_dtype == torch.float16 + and qconfig_weight_dtype == torch.float16 + and not is_dynamic + ) else torch.float + + if is_dynamic: + is_match = qconfig_input_act_is_dynamic and \ + input_dtype == qconfig_activation_dtype and \ + output_dtype == torch.float and \ + weight_dtype == qconfig_weight_dtype + else: + is_match = input_dtype == qconfig_activation_dtype and \ + output_dtype == qconfig_activation_dtype and \ + weight_dtype == qconfig_weight_dtype and \ + bias_dtype == qconfig_bias_dtype + if is_match: + return True + return False + +def _get_object_type_qconfig( + qconfig_mapping: QConfigMapping, + object_type: Union[Callable, str], + fallback_qconfig: QConfigAny) -> QConfigAny: + return qconfig_mapping.object_type_qconfigs.get(object_type, fallback_qconfig) + + +def _get_module_name_regex_qconfig(qconfig_mapping, module_name, fallback_qconfig): + for regex_pattern, qconfig in qconfig_mapping.module_name_regex_qconfigs.items(): + if re.match(regex_pattern, module_name): + # first match wins + return qconfig + return fallback_qconfig + + +def _get_module_name_qconfig(qconfig_mapping, module_name, fallback_qconfig): + if module_name == '': + # module name qconfig not found + return fallback_qconfig + if module_name in qconfig_mapping.module_name_qconfigs: + return qconfig_mapping.module_name_qconfigs[module_name] + else: + parent, _ = _parent_name(module_name) + return _get_module_name_qconfig(qconfig_mapping, parent, fallback_qconfig) + + +def _maybe_adjust_qconfig_for_module_type_or_name(qconfig_mapping, module_type, module_name, global_qconfig): + # get qconfig for module_name, + # fallback to module_name_regex_qconfig, module_type_qconfig, + # global_qconfig if necessary + module_type_qconfig = _get_object_type_qconfig( + qconfig_mapping, module_type, global_qconfig) + module_name_regex_qconfig = _get_module_name_regex_qconfig( + qconfig_mapping, module_name, module_type_qconfig) + module_name_qconfig = _get_module_name_qconfig( + qconfig_mapping, module_name, module_name_regex_qconfig) + return module_name_qconfig + + +def _get_flattened_qconfig_dict(qconfig_mapping: QConfigMapping) -> Dict[Union[Callable, str], QConfigAny]: + """ flatten the global, object_type and module_name qconfig + to the same qconfig_dict so that it can be used by + propagate_qconfig_ function. + "module_name_regex" is ignored for now since it's not supported + in propagate_qconfig_, but it can be fixed later. + + For example: + Input: { + "": qconfig, + "object_type": [ + (torch.add, qconfig) + ], + "module_name": [ + ("conv", qconfig) + ] + } + + Output: { + "": qconfig, + torch.add: qconfig, + "conv": qconfig + } + """ + flattened: Dict[Union[Callable, str], QConfigAny] = {"": qconfig_mapping.global_qconfig} + for obj, qconfig in qconfig_mapping.object_type_qconfigs.items(): + flattened[obj] = qconfig + for obj, qconfig in qconfig_mapping.module_name_qconfigs.items(): + flattened[obj] = qconfig + return flattened + + +def _update_qconfig_for_qat( + qconfig_mapping: QConfigMapping, + backend_config: BackendConfig): + """ + Update the qconfig_mapping to account for module swaps during QAT. + During QAT we perform a module swap on the nn.Module types to the corresponding nn.qat.modules types. + """ + module_to_qat_module_class = get_module_to_qat_module(backend_config) + object_type_dict = qconfig_mapping.object_type_qconfigs + new_object_type_dict = object_type_dict.copy() + for k, v in new_object_type_dict.items(): + if k in module_to_qat_module_class: + object_type_dict[module_to_qat_module_class[k]] = v diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/tracer.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/tracer.py new file mode 100644 index 0000000000000000000000000000000000000000..47f326caf7043f54866f860ab464c3434eb91a5d --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/tracer.py @@ -0,0 +1,45 @@ +import torch +from torch.fx._symbolic_trace import Tracer +from torch.fx.proxy import Scope +from torch.ao.nn.intrinsic import _FusedModule +from typing import List, Callable + +__all__ = [ + "QuantizationTracer", +] + +class ScopeContextManager(torch.fx.proxy.ScopeContextManager): + def __init__( + self, + scope: Scope, + current_module: torch.nn.Module, + current_module_path: str + ): + super().__init__(scope, Scope(current_module_path, type(current_module))) + + +class QuantizationTracer(Tracer): + def __init__( + self, skipped_module_names: List[str], skipped_module_classes: List[Callable] + ): + super().__init__() + self.skipped_module_names = skipped_module_names + self.skipped_module_classes = skipped_module_classes + # NB: initialized the module_type of top level module to None + # we are assuming people won't configure the model with the type of top level + # module here, since people can use "" for global config + # We can change this if there is a use case that configures + # qconfig using top level module type + self.scope = Scope("", None) + self.record_stack_traces = True + + def is_leaf_module(self, m: torch.nn.Module, module_qualified_name: str) -> bool: + return ( + ( + (m.__module__.startswith("torch.nn") or m.__module__.startswith("torch.ao.nn")) + and not isinstance(m, torch.nn.Sequential) + ) + or module_qualified_name in self.skipped_module_names + or type(m) in self.skipped_module_classes + or isinstance(m, _FusedModule) + ) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/utils.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/utils.py new file mode 100644 index 0000000000000000000000000000000000000000..38176c5907d940071d06bd8c4fe56dea771d1a5b --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/fx/utils.py @@ -0,0 +1,885 @@ +import copy +import torch +import torch.nn as nn +from torch.ao.quantization import ( + QConfigAny, + QuantType, +) +from torch.ao.quantization.backend_config import ( + DTypeWithConstraints, +) +from torch.ao.quantization.fake_quantize import ( + FakeQuantizeBase, + FixedQParamsFakeQuantize, +) +from torch.ao.quantization.observer import ( + FixedQParamsObserver, + ObserverBase, +) +from torch.ao.quantization.qconfig import ( + float16_static_qconfig, + float16_dynamic_qconfig, + qconfig_equals, +) +from torch.ao.quantization.stubs import DeQuantStub +from torch.ao.quantization.utils import ( + activation_is_statically_quantized, +) +from torch.ao.quantization.observer import _is_activation_post_process +from torch.ao.quantization.qconfig_mapping import QConfigMapping + +from torch.fx import GraphModule, map_arg + +from torch.fx.graph import ( + Graph, + Node, +) +from .custom_config import PrepareCustomConfig +# importing the lib so that the quantized_decomposed ops are registered +from ._decomposed import quantized_decomposed_lib # noqa: F401 + +from typing import Callable, Optional, List, Dict, Any, Set, Tuple, Union, Type +from dataclasses import dataclass +from collections import namedtuple +import operator +import warnings + +# TODO: revisit this list. Many helper methods shouldn't be public +__all__ = [ + "all_node_args_except_first", + "all_node_args_have_no_tensors", + "assert_and_get_unique_device", + "collect_producer_nodes", + "create_getattr_from_value", + "create_node_from_old_node_preserve_meta", + "EMPTY_ARG_DICT", + "get_custom_module_class_keys", + "get_linear_prepack_op_for_dtype", + "get_new_attr_name_with_prefix", + "get_non_observable_arg_indexes_and_types", + "get_qconv_prepack_op", + "get_skipped_module_name_and_classes", + "graph_module_from_producer_nodes", + "maybe_get_next_module", + "NodeInfo", + "node_arg_is_bias", + "node_arg_is_weight", + "NON_OBSERVABLE_ARG_DICT", + "NON_QUANTIZABLE_WEIGHT_OPS", + "return_arg_list", + "ObservedGraphModuleAttrs", +] + +NON_QUANTIZABLE_WEIGHT_OPS = {torch.nn.functional.layer_norm, torch.nn.functional.group_norm, torch.nn.functional.instance_norm} + +@dataclass +class ObservedGraphModuleAttrs: + node_name_to_qconfig: Dict[str, QConfigAny] + node_name_to_scope: Dict[str, Tuple[str, type]] + prepare_custom_config: PrepareCustomConfig + equalization_node_name_to_qconfig: Dict[str, Any] + qconfig_mapping: QConfigMapping + is_qat: bool + observed_node_names: Set[str] + is_observed_standalone_module: bool = False + standalone_module_input_quantized_idxs: Optional[List[int]] = None + standalone_module_output_quantized_idxs: Optional[List[int]] = None + +def node_arg_is_weight(node: Node, arg: Any) -> bool: + """Returns if node arg is weight""" + weight_index = None + if "target_dtype_info" in node.meta: + weight_index = node.meta["target_dtype_info"].get("weight_index", None) + if weight_index is not None and weight_index < len(node.args) and node.args[weight_index] is arg: + return True + return node.kwargs.get("weight") is arg + +def node_arg_is_bias(node: Node, arg: Any) -> bool: + """Returns if node arg is bias""" + bias_index = None + if "target_dtype_info" in node.meta: + bias_index = node.meta["target_dtype_info"].get("bias_index", None) + if bias_index is not None and bias_index < len(node.args) and node.args[bias_index] is arg: + return True + return node.kwargs.get("bias") is arg + +def get_custom_module_class_keys(custom_module_mapping: Dict[QuantType, Dict[Type, Type]]) -> List[Any]: + r""" Get all the unique custom module keys in the custom config dict + e.g. + Input: + { + QuantType.STATIC: { + CustomModule1: ObservedCustomModule + }, + QuantType.DYNAMIC: { + CustomModule2: DynamicObservedCustomModule + }, + QuantType.WEIGHT_ONLY: { + CustomModule3: WeightOnlyObservedCustomModule + }, + } + + Output: + # extract the keys across all inner STATIC, DYNAMIC, and WEIGHT_ONLY dicts + [CustomModule1, CustomModule2, CustomModule3] + """ + # using set to dedup + float_custom_module_classes : Set[Any] = set() + for quant_mode in [QuantType.STATIC, QuantType.DYNAMIC, QuantType.WEIGHT_ONLY]: + quant_mode_custom_module_config = custom_module_mapping.get(quant_mode, {}) + quant_mode_custom_module_classes = set(quant_mode_custom_module_config.keys()) + float_custom_module_classes |= quant_mode_custom_module_classes + return list(float_custom_module_classes) + +def get_linear_prepack_op_for_dtype(dtype): + if dtype == torch.float16: + return torch.ops.quantized.linear_prepack_fp16 + elif dtype == torch.qint8: + return torch.ops.quantized.linear_prepack + else: + raise Exception("can't get linear prepack op for dtype:", dtype) + +def get_qconv_prepack_op(conv_op: Callable) -> Callable: + prepack_ops = { + torch.nn.functional.conv1d: torch.ops.quantized.conv1d_prepack, + torch.nn.functional.conv2d: torch.ops.quantized.conv2d_prepack, + torch.nn.functional.conv3d: torch.ops.quantized.conv3d_prepack, + torch.nn.functional.conv_transpose1d: torch.ops.quantized.conv_transpose1d_prepack, + torch.nn.functional.conv_transpose2d: torch.ops.quantized.conv_transpose2d_prepack, + torch.nn.functional.conv_transpose3d: torch.ops.quantized.conv_transpose3d_prepack, + } + prepack_op = prepack_ops.get(conv_op, None) + assert prepack_op, f"Didn't find prepack op for {conv_op}" + return prepack_op + +# Returns a function that can get a new attribute name for module with given +# prefix, for example, +# >> get_new_observer_name = get_new_attr_name_with_prefix('_observer') +# >> new_name = get_new_observer_name(module) +# new_name will be an unused attribute name on module, e.g. `_observer_1` +def get_new_attr_name_with_prefix(prefix: str) -> Callable: + prefix = prefix.replace(".", "_") + + def get_new_attr_name(module: torch.nn.Module): + def get_attr_name(i: int): + return prefix + str(i) + i = 0 + attr_name = get_attr_name(i) + while hasattr(module, attr_name): + i += 1 + attr_name = get_attr_name(i) + return attr_name + return get_new_attr_name + +def collect_producer_nodes(node: Node) -> Optional[List[Node]]: + r''' Starting from a target node, trace back until we hit inpu or + getattr node. This is used to extract the chain of operators + starting from getattr to the target node, for example + def forward(self, x): + observed = self.observer(self.weight) + return F.linear(x, observed) + collect_producer_nodes(observed) will either return a list of nodes that + produces the observed node or None if we can't extract a self contained + graph without free variables(inputs of the forward function). + ''' + nodes = [node] + frontier = [node] + while frontier: + node = frontier.pop() + all_args = list(node.args) + list(node.kwargs.values()) + for arg in all_args: + if not isinstance(arg, Node): + continue + if arg.op == 'placeholder': + # hit input, can't fold in this case + return None + nodes.append(arg) + if not (arg.op == 'call_function' and arg.target == getattr): + frontier.append(arg) + return nodes + +def graph_module_from_producer_nodes( + root: GraphModule, producer_nodes: List[Node]) -> GraphModule: + r''' Construct a graph module from extracted producer nodes + from `collect_producer_nodes` function + Args: + root: the root module for the original graph + producer_nodes: a list of nodes we use to construct the graph + Return: + A graph module constructed from the producer nodes + ''' + assert len(producer_nodes) > 0, 'list of producer nodes can not be empty' + # since we traced back from node to getattr + producer_nodes.reverse() + graph = Graph() + env: Dict[Any, Any] = {} + + def load_arg(a): + return map_arg(a, lambda node: env[node]) + for producer_node in producer_nodes: + env[producer_node] = graph.node_copy(producer_node, load_arg) + graph.output(load_arg(producer_nodes[-1])) + graph_module = GraphModule(root, graph) + return graph_module + +def assert_and_get_unique_device(module: torch.nn.Module) -> Any: + """ + Returns the unique device for a module, or None if no device is found. + Throws an error if multiple devices are detected. + """ + devices = {p.device for p in module.parameters()} | \ + {p.device for p in module.buffers()} + """ + As a temp workaround for AIMP HHC publish we added CPU check.remove it later. T163614564 + """ + if {torch.device("cpu"), torch.device("meta")} == devices: + warnings.warn("Both 'meta' and 'cpu' are present in the list of devices. Module can have one device. We Select 'cpu'.") + devices = {torch.device("cpu")} + "" + assert len(devices) <= 1, ( + "prepare only works with cpu or single-device CUDA modules, " + f"but got devices {devices}" + ) + device = next(iter(devices)) if len(devices) > 0 else None + return device + +def create_getattr_from_value(module: torch.nn.Module, graph: Graph, prefix: str, value: Any) -> Node: + """ + Given a value of any type, creates a getattr node corresponding to the value and + registers the value as a buffer to the module. + """ + get_new_attr_name = get_new_attr_name_with_prefix(prefix) + attr_name = get_new_attr_name(module) + device = assert_and_get_unique_device(module) + new_value = value.clone().detach() if isinstance(value, torch.Tensor) \ + else torch.tensor(value, device=device) + module.register_buffer(attr_name, new_value) + # Create get_attr with value + attr_node = graph.create_node("get_attr", attr_name) + return attr_node + +def all_node_args_have_no_tensors(node: Node, modules: Dict[str, torch.nn.Module], cache: Dict[Node, bool]) -> bool: + """ + If we know for sure that all of this node's args have no + tensors (are primitives), return True. If we either + find a tensor or are not sure, return False. Note: this + function is not exact. + """ + if cache and node in cache: + return cache[node] + + result = False # will be overwritten + if not isinstance(node, Node): + result = True + elif node.op == 'placeholder': + result = False + elif node.op == 'call_module': + assert isinstance(node.target, str) + if _is_activation_post_process(modules[node.target]): + result = all_node_args_have_no_tensors(node.args[0], modules, cache) # type: ignore[arg-type] + elif node.op == 'call_module': + result = False + elif node.op == 'call_function' and node.target is operator.getitem: + result = all_node_args_have_no_tensors(node.args[0], modules, cache) # type: ignore[arg-type] + elif node.op == 'get_attr': + result = False + elif node.target is getattr and node.args[1] in ['ndim', 'shape']: + # x1 = x0.ndim + result = True + elif node.op == 'call_method' and node.target == 'size': + # x1 = x0.size(0) + result = True + else: + found_one_tensor = False + for arg in node.args: + if isinstance(arg, list): + for list_el in arg: + if isinstance(list_el, Node): + this_list_el_args_have_no_tensors = \ + all_node_args_have_no_tensors(list_el, modules, cache) + found_one_tensor = found_one_tensor or \ + (not this_list_el_args_have_no_tensors) + # If found_one_tensor is True, there is no point in + # recursing further as the end result will always + # be True. + # TODO(future PR): remove this entire function and + # change to dtype inference without recursion. + if found_one_tensor: + result = not found_one_tensor + if cache: + cache[node] = result + return result + elif isinstance(arg, int): + pass + else: + if isinstance(arg, Node): + this_arg_args_have_no_tensors = all_node_args_have_no_tensors(arg, modules, cache) + found_one_tensor = found_one_tensor or \ + (not this_arg_args_have_no_tensors) + # If found_one_tensor is True, there is no point in + # recursing further as the end result will always + # be True. + # TODO(future PR): remove this entire function and + # change to dtype inference without recursion. + if found_one_tensor: + result = not found_one_tensor + if cache: + cache[node] = result + return result + else: + found_one_tensor = True + result = not found_one_tensor + if cache: + cache[node] = result + return result + +def all_node_args_except_first(node: Node) -> List[int]: + """ + Returns all node arg indices after first + """ + return list(range(1, len(node.args))) + +def return_arg_list(arg_indices: List[int]) -> Callable[[Node], List[int]]: + """ + Constructs a function that takes a node as arg and returns the arg_indices + that are valid for node.args + """ + def arg_indices_func(node: Node) -> List[int]: + return [i for i in arg_indices if i < len(node.args)] + return arg_indices_func + +NodeInfo = namedtuple("NodeInfo", "op target") + +# this dict identifies which indices of a node are non tensors +# so that they can be propagated correctly since inserting observers +# for them would cause errors + +NON_OBSERVABLE_ARG_DICT: Dict[NodeInfo, Dict[Union[type, torch.dtype], Callable[[Node], List[int]]]] = { + NodeInfo("call_method", "masked_fill") : { + torch.bool: return_arg_list([1]), + float: return_arg_list([2]) + }, + NodeInfo("call_method", "permute") : { + int: all_node_args_except_first + }, + NodeInfo("call_method", "repeat") : { + int: all_node_args_except_first + }, + NodeInfo("call_method", "reshape") : { + int: all_node_args_except_first + }, + NodeInfo("call_method", "size") : { + int: return_arg_list([1]) + }, + NodeInfo("call_method", "transpose") : { + int: all_node_args_except_first + }, + NodeInfo("call_method", torch.transpose) : { + int: all_node_args_except_first + }, + NodeInfo("call_method", "unsqueeze") : { + int: return_arg_list([1]) + }, + NodeInfo("call_method", "unsqueeze_") : { + int: return_arg_list([1]) + }, + NodeInfo("call_method", torch.unsqueeze) : { + int: return_arg_list([1]) + }, + NodeInfo("call_method", "view") : { + int: all_node_args_except_first + }, +} + +EMPTY_ARG_DICT: Dict[Union[type, torch.dtype], Callable[[Node], List[int]]] = {} + +def get_non_observable_arg_indexes_and_types(node: Node) -> Dict[Union[type, torch.dtype], Callable[[Node], List[int]]]: + """ + Returns a dict with of non float tensor types as keys and values which correspond to a + function to retrieve the list (which takes the node as an argument) + """ + info = NodeInfo(node.op, node.target) + + return NON_OBSERVABLE_ARG_DICT.get(info, EMPTY_ARG_DICT) + +def maybe_get_next_module( + node: Node, + modules: Dict[str, nn.Module], + target_module_type: Optional[Type[nn.Module]] = None, + target_functional_type: Any = None, +) -> Optional[Node]: + """ Gets the next module that matches what is needed in + is_target_module_type if it exists + + Args: + node: The node whose users we want to look at + target_module_type: Module type that we want to check + target_functional_type: Functional type that we want to check + """ + + for user in node.users.keys(): + if user.op == 'call_module' and target_module_type is not None and \ + isinstance(modules[str(user.target)], target_module_type): + return user + elif (user.op == 'call_function' and target_functional_type is not None and + user.target == target_functional_type): + return user + + return None + +def create_node_from_old_node_preserve_meta( + quantized_graph: Graph, + create_node_args: Tuple[Any, ...], + old_node: Node, +) -> Node: + """ + Creates `new_node` and copies the necessary metadata to it from `old_node`. + """ + new_node = quantized_graph.create_node(*create_node_args) + new_node.stack_trace = old_node.stack_trace + return new_node + +def get_skipped_module_name_and_classes( + prepare_custom_config: PrepareCustomConfig, + is_standalone_module: bool) -> Tuple[List[str], List[Type[Any]]]: + skipped_module_names = copy.copy(prepare_custom_config.non_traceable_module_names) + skipped_module_classes = copy.copy(prepare_custom_config.non_traceable_module_classes) + if not is_standalone_module: + # standalone module and custom module config are applied in top level module + skipped_module_names += list(prepare_custom_config.standalone_module_names.keys()) + skipped_module_classes += list(prepare_custom_config.standalone_module_classes.keys()) + skipped_module_classes += get_custom_module_class_keys(prepare_custom_config.float_to_observed_mapping) + + return skipped_module_names, skipped_module_classes + +def _is_custom_module_lstm( + node: Node, + named_modules: Dict[str, torch.nn.Module], + qconfig: QConfigAny = None, + # QuantizeHandler, but we cannot include the type here due to circular imports + qhandler: Optional[Any] = None, +) -> bool: + """ + Return whether this refers to the custom module LSTM flow. + """ + mod = _get_module(node, named_modules) + if qconfig is not None and qhandler is not None: + assert isinstance(qhandler, torch.ao.quantization.fx.quantize_handler.QuantizeHandler) # type: ignore[attr-defined] + return isinstance(mod, torch.nn.LSTM) and \ + activation_is_statically_quantized(qconfig) and \ + qhandler.is_custom_module() + else: + return isinstance(mod, torch.ao.nn.quantizable.LSTM) + +def _is_custom_module_mha( + node: Node, + named_modules: Dict[str, torch.nn.Module], + qconfig: QConfigAny = None, + # QuantizeHandler, but we cannot include the type here due to circular imports + qhandler: Optional[Any] = None, +) -> bool: + """ + Return whether this refers to the custom module MultiheadAttention flow. + """ + mod = _get_module(node, named_modules) + if qconfig is not None and qhandler is not None: + assert isinstance(qhandler, torch.ao.quantization.fx.quantize_handler.QuantizeHandler) # type: ignore[attr-defined] + return isinstance(mod, torch.nn.MultiheadAttention) and \ + activation_is_statically_quantized(qconfig) and \ + qhandler.is_custom_module() + else: + return isinstance(mod, torch.ao.nn.quantizable.MultiheadAttention) + +def _get_module(node: Node, named_modules: Dict[str, torch.nn.Module]) -> Optional[torch.nn.Module]: + """ + If `node` refers to a call_module node, return the module, else None. + """ + if node.op == "call_module" and str(node.target) in named_modules: + return named_modules[str(node.target)] + else: + return None + +def _insert_dequant_stub( + node: Node, + model: torch.nn.Module, + named_modules: Dict[str, torch.nn.Module], + graph: Graph, +) -> Node: + """ + Attach a `DeQuantStub` to the model and create a node that calls this + `DeQuantStub` on the output of `node`, similar to how observers are inserted. + """ + prefix = "dequant_stub_" + get_new_dequant_stub_name = get_new_attr_name_with_prefix(prefix) + dequant_stub_name = get_new_dequant_stub_name(model) + dequant_stub = DeQuantStub() + setattr(model, dequant_stub_name, dequant_stub) + named_modules[dequant_stub_name] = dequant_stub + with graph.inserting_after(node): + return graph.call_module(dequant_stub_name, (node,)) + +def _insert_dequant_stubs_for_custom_module_lstm_output( + node: Node, + model: torch.nn.Module, + named_modules: Dict[str, torch.nn.Module], + graph: Graph, +) -> Node: + """ + Insert DeQuantStubs after each internal output node of custom module LSTM. + + Custom module LSTM outputs are nested tuples of the structure (output, (hidden0, hidden1)), + Since we cannot dequantize a tuple as a whole, we must first break down the tuple into its + components through `getitem`. This function transforms the graph as follows: + + (1) Split the LSTM node into (output, (hidden0, hidden1)) + (2) Insert a DeQuantStub after each internal node + (3) Recombine the DeQuantStubs into the same structure as before + (4) Reroute all consumers of the original LSTM node and its sub-nodes + (e.g. lstm[0]) + + Before: + lstm_output + | + v + original_user(s) + After: + lstm_output + / \\ + / (getitem) \\ + / \\ + v v + output hidden + | / \\ + (DeQuantStub) (getitem) + | / \\ + v v v + output_dq hidden0 hidden1 + | | | + | (DeQuantStub) (DeQuantStub) + | | | + | v v + | hidden0_dq hidden1_dq + | \\ / + | (tuple) + | \\ / + | v v + | hidden_dq + \\ / + \\ (tuple) / + v v + lstm_output_dq + | + v + original_user(s) + + For step (4), reroute all users of the original LSTM node(s) as follows: + lstm_output -> lstm_output_dq + lstm_output[0] -> output_dq + lstm_output[1] -> hidden_dq + lstm_output[1][0] -> hidden0_dq + lstm_output[1][1] -> hidden1_dq + + Return the node `lstm_output_dq`. + """ + # (1) Split the LSTM node into (output, (hidden0, hidden1)) + # (2) Insert a DeQuantStub after each internal node + with graph.inserting_after(node): + output = graph.call_function(operator.getitem, (node, 0)) + output_dq = _insert_dequant_stub(output, model, named_modules, graph) + with graph.inserting_after(output_dq): + hidden = graph.call_function(operator.getitem, (node, 1)) + with graph.inserting_after(hidden): + hidden0 = graph.call_function(operator.getitem, (hidden, 0)) + hidden0_dq = _insert_dequant_stub(hidden0, model, named_modules, graph) + with graph.inserting_after(hidden0_dq): + hidden1 = graph.call_function(operator.getitem, (hidden, 1)) + hidden1_dq = _insert_dequant_stub(hidden1, model, named_modules, graph) + + # (3) Recombine the DeQuantStubs into the same structure as before + with graph.inserting_after(hidden1_dq): + hidden_dq = graph.call_function(tuple, ([hidden0_dq, hidden1_dq],)) + with graph.inserting_after(hidden_dq): + lstm_output_dq = graph.call_function(tuple, ([output_dq, hidden_dq],)) + + # (4) Reroute all consumers of the original LSTM node and its sub-nodes + for user in list(node.users.keys()): + if user != output and user != hidden: + user.replace_input_with(node, lstm_output_dq) + # The getitem and tuple nodes we added here may interfere with reference quantized + # pattern matching, so we need to redirect the consumers of internal nodes to the + # corresponding nodes with DeQuantStubs (e.g. lstm_output_dq[0] -> output_dq) attached, + # in order to preserve reference patterns like "dequantize - consumer - quantize". + _reroute_tuple_getitem_pattern(graph) + return lstm_output_dq + +def _maybe_get_custom_module_lstm_from_node_arg( + arg: Node, + named_modules: Dict[str, torch.nn.Module], +) -> Optional[Node]: + """ + Given an argument of a node, if the argument refers to the path through which the node + is a consumer of custom module LSTM, return the custom module LSTM node, or None otherwise. + + This is used to determine whether a node is a consumer of custom module LSTM, and, if so, + skip inserting input observers for this node. This is because custom module LSTM produces + quantized outputs, so inserting an input observer for the consumer of custom module LSTM + would unnecessarily quantize the outputs again. + + lstm -> consumer + + In practice, however, custom module LSTM outputs a tuple (output, (hidden0, hidden1)) with + DeQuantStubs attached to each internal node (see `_insert_dequant_stubs_for_custom_module_lstm_output`). + This tuple can be consumed in one of four ways: + + lstm -> getitem -> DeQuantStub -> consumer # consume lstm[0] + lstm -> getitem -> getitem -> DeQuantStub -> tuple -> consumer # consume lstm[1] + lstm -> getitem -> getitem -> DeQuantStub -> consumer # consume lstm[1][0] or lstm[1][1] + lstm -> getitem -> DeQuantStub -> tuple -> consumer # consume lstm + + Thus, we must match against the above patterns instead of simply checking the parent node + to determine whether this node is a consumer of a custom module LSTM. + """ + def match_dq(a): + return isinstance(_get_module(a, named_modules), DeQuantStub) + + def match_lstm(a): + return _is_custom_module_lstm(a, named_modules) + + def match_getitem(a): + return a.op == "call_function" and a.target == operator.getitem + + def match_tuple(a): + return a.op == "call_function" and a.target == tuple + + def _match_pattern(match_pattern: List[Callable]) -> Optional[Node]: + """ + Traverse up the graph and match the args one by one. + If there is a match, return the last matched node, or None otherwise. + """ + a = arg + for i, match in enumerate(match_pattern): + if not match(a): + return None + # Match next arg, for tuple the arg is a tuple of a list, e.g. ([dq_1, other_node],) + if i < len(match_pattern) - 1: + if match == match_tuple: + a = a.args[0][0] # type: ignore[assignment,index] + else: + a = a.args[0] # type: ignore[assignment] + return a + + all_match_patterns = [ + [match_dq, match_getitem, match_lstm], + [match_tuple, match_dq, match_getitem, match_getitem, match_lstm], + [match_dq, match_getitem, match_getitem, match_lstm], + [match_tuple, match_dq, match_getitem, match_lstm], + ] + + for p in all_match_patterns: + matched_node = _match_pattern(p) + if matched_node is not None: + return matched_node + return None + +def _reroute_tuple_getitem_pattern(graph: Graph): + """ + Search for patterns where N consecutive `tuple` call_function nodes are followed by + N consecutive `getitem` call_function nodes that are "reverses" of the `tuple` nodes. + If we find this pattern, reroute the consumers of the last `getitem` to skip these + N `tuple` and `getitem` nodes. + + Before: + + a b c + | \\ / + \\ tuple + \\ / + tuple + | + getitem(1) + | + getitem(0) + | + d + + After: + + b + | + d + """ + def find_patterns( + node: Node, + index_stack: List[int], + current_pattern: List[Node], + matched_patterns: List[List[Node]], + seen: Set[Tuple[Node, Tuple[int, ...]]]): + """ + Traverse the graph recursively to match for the N-tuple - N-getitem patterns, + starting at the given node. + + We use a stack to keep track of the expected `getitem` indices, since these are + reversed from the `tuple` indices. In the above example, the stack after + (b -> tuple -> tuple) will be [0, 1], which will be popped by getitem(1) first + and then by getitem(0). + + TODO: traverse upwards from the output and handle the case when tuple is not a + separate node, e.g. graph.call_function(operator.getitem, args=(a, (b, c))) + """ + if len(index_stack) == 0 and len(current_pattern) > 0: + matched_patterns.append(copy.copy(current_pattern)) + current_pattern.clear() + + # Avoid duplicating work + state = (node, tuple(index_stack)) + if state in seen: + return + seen.add(state) + + # Iterate through users of this node to find tuple/getitem nodes to match + for user in node.users: + if user.op == "call_function" and user.target == tuple: + for i, user_arg in enumerate(user.args[0]): # type: ignore[arg-type] + if user_arg == node: + index_stack.append(i) + current_pattern.append(user) + find_patterns(user, index_stack, current_pattern, matched_patterns, seen) + elif user.op == "call_function" and user.target == operator.getitem: + if len(index_stack) > 0: + if user.args[1] == index_stack[-1]: + index_stack.pop() + current_pattern.append(user) + find_patterns(user, index_stack, current_pattern, matched_patterns, seen) + return matched_patterns + + # Collect all matched patterns + matched_patterns: List[List[Node]] = [] + seen: Set[Tuple[Node, Tuple[int, ...]]] = set() # (node, index_stack) + for node in graph.nodes: + find_patterns(node, [], [], matched_patterns, seen) + + # For each pattern, redirect all consumers of the last getitem node to the correct input + # of the first tuple node + for pattern in matched_patterns: + first_tuple = pattern[0] + last_getitem = pattern[-1] + assert first_tuple.op == "call_function" and first_tuple.target == tuple + assert last_getitem.op == "call_function" and last_getitem.target == operator.getitem + last_getitem_index = last_getitem.args[1] + new_input = first_tuple.args[0][last_getitem_index] # type: ignore[index] + for user in list(last_getitem.users.keys()): + user.replace_input_with(last_getitem, new_input) + +def _get_observer_from_activation_post_process( + activation_post_process: Union[ObserverBase, FakeQuantizeBase], +) -> ObserverBase: + """ + If `activation_post_process` is an observer, return the observer. + If `activation_post_process` is a fake quantize, return the internal observer. + """ + if isinstance(activation_post_process, ObserverBase): + return activation_post_process + else: + assert isinstance(activation_post_process, FakeQuantizeBase) + return activation_post_process.activation_post_process # type: ignore[return-value] + +def _qconfig_satisfies_dtype_config_constraints( + qconfig: QConfigAny, + dtype_with_constraints: DTypeWithConstraints, + is_activation: bool = True) -> bool: + """ + Return whether `qconfig` satisfies the following constraints from the backend, + specified through the activation and weight DTypeWithConstraints. + + 1. QConfig specified a quantization range that falls within the backend's, if any + 2. QConfig specified a min scale value that is >= the backend's, if any + 3. QConfig specified a FixedQParamsObserver or FixedQParamsFakeQuantize that has + scale and zero point that match the backend's, if any + + If `is_activation` is True, we check `qconfig.activation`, else we check `qconfig.weight`. + If `qconfig` or `dtype_with_constraints.dtype` is None, or the dtypes do not match, return True. + """ + # TODO: log warnings only when the user enabled a debug flag + def _activation_post_process_satisfies_dtype_config_constraints( + activation_post_process: Union[ObserverBase, FakeQuantizeBase], + dtype_with_constraints: DTypeWithConstraints, + debug_string: str) -> bool: + observer = _get_observer_from_activation_post_process(activation_post_process) + app_quant_min = getattr(observer, "quant_min", None) + app_quant_max = getattr(observer, "quant_max", None) + # TODO: for now, just use the existing eps value as scale_min. In the future, we should + # resolve the differences between the two, either by renaming eps or some other way + app_scale_min = getattr(observer, "eps", None) + backend_quant_min = dtype_with_constraints.quant_min_lower_bound + backend_quant_max = dtype_with_constraints.quant_max_upper_bound + backend_scale_min = dtype_with_constraints.scale_min_lower_bound + backend_scale_exact_match = dtype_with_constraints.scale_exact_match + backend_zero_point_exact_match = dtype_with_constraints.zero_point_exact_match + # check quantization ranges + if backend_quant_min is not None and backend_quant_max is not None: + if app_quant_min is None or app_quant_max is None: + warnings.warn(f"QConfig {debug_string} must specify 'quant_min' and 'quant_max', ignoring {qconfig}") + return False + elif app_quant_min < backend_quant_min or app_quant_max > backend_quant_max: + warnings.warn( + f"QConfig {debug_string} quantization range must fall within the backend's:\n" + f"QConfig range = ({app_quant_min}, {app_quant_max}), " + f"BackendConfig range = ({backend_quant_min}, {backend_quant_max}), " + f"ignoring {qconfig}" + ) + return False + # check scale min + if backend_scale_min is not None: + if app_scale_min is None: + warnings.warn(f"QConfig {debug_string} must specify 'eps', ignoring {qconfig}") + return False + if app_scale_min < backend_scale_min: + warnings.warn( + f"QConfig {debug_string} eps ({app_scale_min}) must be greater than or equal to " + f"the backend's min scale value ({backend_scale_min}), ignoring {qconfig}" + ) + return False + # check fixed scale and zero point + if backend_scale_exact_match is not None and backend_zero_point_exact_match is not None: + # For tests only, accept the following qconfigs for now + # TODO: handle fp16 qconfigs properly + for accepted_qconfig in [float16_static_qconfig, float16_dynamic_qconfig]: + if qconfig_equals(qconfig, accepted_qconfig): + return True + suggestion_str = ( + "Please use torch.ao.quantization.get_default_qconfig_mapping or " + "torch.ao.quantization.get_default_qat_qconfig_mapping. Example:\n" + " qconfig_mapping = get_default_qconfig_mapping(\"fbgemm\")\n" + " model = prepare_fx(model, qconfig_mapping, example_inputs)" + ) + if not isinstance(activation_post_process, FixedQParamsObserver) and \ + not isinstance(activation_post_process, FixedQParamsFakeQuantize): + warnings.warn( + f"QConfig must specify a FixedQParamsObserver or a FixedQParamsFakeQuantize " + f"for fixed qparams ops, ignoring {qconfig}.\n{suggestion_str}" + ) + return False + if observer.scale != backend_scale_exact_match or observer.zero_point != backend_zero_point_exact_match: + warnings.warn( + f"QConfig fixed scale ({observer.scale}) and zero point ({observer.zero_point}) " + f"do not match the backend's ({backend_scale_exact_match} and {backend_zero_point_exact_match}), " + f"ignoring {qconfig}.\n{suggestion_str}" + ) + return False + return True + + if qconfig is None or dtype_with_constraints.dtype is None: + return True + + activation_post_process_ctr = qconfig.activation if is_activation else qconfig.weight + debug_string = "activation" if is_activation else "weight" + satisfies_constraints = True + if activation_post_process_ctr is not None: + activation_post_process = activation_post_process_ctr() + assert _is_activation_post_process(activation_post_process) + # If dtypes don't match, don't check the activation_post_process and return True early + if activation_post_process.dtype != dtype_with_constraints.dtype: + return True + satisfies_constraints = _activation_post_process_satisfies_dtype_config_constraints( + activation_post_process, dtype_with_constraints, debug_string) + return satisfies_constraints diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__init__.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..e65652573b1b30bc755ab1861d4f7de0359bcedc --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__init__.py @@ -0,0 +1,21 @@ +from .quantizer import ( + DerivedQuantizationSpec, + EdgeOrNode, + FixedQParamsQuantizationSpec, + QuantizationAnnotation, + QuantizationSpec, + QuantizationSpecBase, + Quantizer, + SharedQuantizationSpec, +) + +__all__ = [ + "EdgeOrNode", + "Quantizer", + "QuantizationSpecBase", + "QuantizationSpec", + "FixedQParamsQuantizationSpec", + "SharedQuantizationSpec", + "DerivedQuantizationSpec", + "QuantizationAnnotation", +] diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/__init__.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/__init__.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..f6f6b9860accd77f30cce35f34b38885e03ef023 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/__init__.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/composable_quantizer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/composable_quantizer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..b9c4cb9bab0e6505435c4fed6a4cbdb1451eedb4 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/composable_quantizer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/embedding_quantizer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/embedding_quantizer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..83469c4189ff8af13960e89c50b5f8b3a844beaa Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/embedding_quantizer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/quantizer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/quantizer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..6cf3475bfc9bb30a114d646fabacfeedb808501e Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/quantizer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/utils.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/utils.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..f55a4881b3e8d14a243dd1795373bd549aaae9b1 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/utils.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/x86_inductor_quantizer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/x86_inductor_quantizer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..cbe32ee9e7f87bdd2884ef4e6e9fda3e0acbf1c2 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/x86_inductor_quantizer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/xnnpack_quantizer.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/xnnpack_quantizer.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..a737b3244c4c58d9cb7c48369b3aad3c0d1ac650 Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/xnnpack_quantizer.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/xnnpack_quantizer_utils.cpython-310.pyc b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/xnnpack_quantizer_utils.cpython-310.pyc new file mode 100644 index 0000000000000000000000000000000000000000..57df5265e6137d5cae97f79198192752855a003d Binary files /dev/null and b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/__pycache__/xnnpack_quantizer_utils.cpython-310.pyc differ diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/composable_quantizer.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/composable_quantizer.py new file mode 100644 index 0000000000000000000000000000000000000000..ad55138191fbfcf83d6e25d584d6f8be1428dd70 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/composable_quantizer.py @@ -0,0 +1,71 @@ +from __future__ import annotations + +from typing import Dict, List + +import torch + +from torch.fx import Node + +from .quantizer import QuantizationAnnotation, Quantizer + +__all__ = [ + "ComposableQuantizer", +] + + +class ComposableQuantizer(Quantizer): + """ + ComposableQuantizer allows users to combine more than one quantizer into a single quantizer. + This allows users to quantize a model with multiple quantizers. E.g., embedding quantization + maybe supported by one quantizer while linear layers and other ops might be supported by another + quantizer. + + ComposableQuantizer is initialized with a list of `Quantizer` instances. + The order of the composition matters since that is the order in which the quantizers will be + applies. + Example: + ``` + embedding_quantizer = EmbeddingQuantizer() + linear_quantizer = MyLinearQuantizer() + xnnpack_quantizer = XNNPackQuantizer() # to handle ops not quantized by previous two quantizers + composed_quantizer = ComposableQuantizer([embedding_quantizer, linear_quantizer, xnnpack_quantizer]) + prepared_m = prepare_pt2e(model, composed_quantizer) + ``` + """ + + def __init__(self, quantizers: List[Quantizer]): + super().__init__() + self.quantizers = quantizers + self._graph_annotations: Dict[Node, QuantizationAnnotation] = {} + + def _record_and_validate_annotations( + self, gm: torch.fx.GraphModule, quantizer: Quantizer + ) -> None: + for n in gm.graph.nodes: + if "quantization_annotation" in n.meta: + # check if the annotation has been changed by + # comparing QuantizationAnnotation object id + if n in self._graph_annotations and ( + id(self._graph_annotations[n]) + != id(n.meta["quantization_annotation"]) + ): + raise RuntimeError( + f"Quantizer {quantizer.__class__.__name__} has changed annotations on node {n}" + ) + else: + self._graph_annotations[n] = n.meta["quantization_annotation"] + else: + if n in self._graph_annotations: + raise RuntimeError( + f"Quantizer {quantizer.__class__.__name__} has removed annotations on node {n}" + ) + + def annotate(self, model: torch.fx.GraphModule) -> torch.fx.GraphModule: + """just handling global spec for now""" + for quantizer in self.quantizers: + quantizer.annotate(model) + self._record_and_validate_annotations(model, quantizer) + return model + + def validate(self, model: torch.fx.GraphModule) -> None: + pass diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/embedding_quantizer.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/embedding_quantizer.py new file mode 100644 index 0000000000000000000000000000000000000000..8ffd2002e580db7cc6cae69161de1e88c788073c --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/embedding_quantizer.py @@ -0,0 +1,96 @@ +from __future__ import annotations + +import copy +from typing import List, Set + +import torch +import torch.nn.functional as F +from torch.ao.quantization.observer import PerChannelMinMaxObserver +from torch.ao.quantization.quantizer.quantizer import ( + QuantizationAnnotation, + QuantizationSpec, + Quantizer, +) +from torch.ao.quantization.quantizer.xnnpack_quantizer_utils import ( + OperatorConfig, + OperatorPatternType, + QuantizationConfig, +) + +__all__ = [ + "get_embedding_operators_config", + "EmbeddingQuantizer", +] + + +def get_embedding_operators_config() -> OperatorConfig: + weight_quantization_spec = QuantizationSpec( + dtype=torch.uint8, + qscheme=torch.per_channel_affine_float_qparams, + ch_axis=0, + observer_or_fake_quant_ctr=PerChannelMinMaxObserver.with_args(eps=2**-12), + ) + quantization_config = QuantizationConfig(None, None, weight_quantization_spec, None) + ops: List[OperatorPatternType] = [[torch.nn.Embedding]] + ops.append([F.embedding]) + supported_config_and_operators = OperatorConfig( + config=quantization_config, operators=ops + ) + return copy.deepcopy(supported_config_and_operators) + + +class EmbeddingQuantizer(Quantizer): + def __init__(self): + super().__init__() + + @classmethod + def get_supported_quantization_configs(cls) -> List[QuantizationConfig]: + op_configs: Set[QuantizationConfig] = set({}) + for spec, _ in cls.get_supported_operators(): + op_configs.add(spec) + return list(op_configs) + + @classmethod + def get_supported_operator_for_quantization_config( + cls, quantization_config: QuantizationConfig + ) -> List[OperatorPatternType]: + for config, ops in cls.get_supported_operators(): + # note: this assumes each entry in cls.supported_spec_and_operators + # corresponds to one spec, e.g. we don't have + # [(spec1, op_list1), (spec1, op_list2), (spec2, op_list3)] + # where the first and second entry have the same spec but did not + # merge the op list + if config == quantization_config: + return ops + return [] + + def annotate(self, model: torch.fx.GraphModule) -> torch.fx.GraphModule: + """just handling global spec for now""" + self._annotate_embedding_ops(model.graph) + return model + + def _annotate_embedding_ops(self, graph: torch.fx.Graph) -> None: + embedding_config: OperatorConfig = get_embedding_operators_config() + for node in graph.nodes: + # Keep node parsing based annotations instead of module partitioners + # just as an example of alternate ways of annotating + if ( + node.op == "call_function" + and node.target == torch.ops.aten.embedding.default + ): + if embedding_config.config.weight is None: + raise ValueError( + "Embedding config must have a valid weight quantization spec." + ) + node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map={ + node.args[0]: embedding_config.config.weight, + } + ) + + def validate(self, model: torch.fx.GraphModule) -> None: + pass + + @classmethod + def get_supported_operators(cls) -> List[OperatorConfig]: + return [get_embedding_operators_config()] diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/quantizer.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/quantizer.py new file mode 100644 index 0000000000000000000000000000000000000000..da9a66337cbec78ef6d3973391a638fe0bc8b42e --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/quantizer.py @@ -0,0 +1,183 @@ +from abc import ABC, abstractmethod +from dataclasses import dataclass, field +from typing import Callable, Dict, List, Optional, Tuple, Union + +import torch +from torch import Tensor +from torch.ao.quantization import ObserverOrFakeQuantize +from torch.ao.quantization.qconfig import _ObserverOrFakeQuantizeConstructor +from torch.fx import Node + +__all__ = [ + "Quantizer", + "QuantizationSpecBase", + "QuantizationSpec", + "FixedQParamsQuantizationSpec", + "EdgeOrNode", + "SharedQuantizationSpec", + "DerivedQuantizationSpec", + "QuantizationAnnotation", +] + +# TODO: maybe remove torch.float32 +SUPPORTED_DTYPES = [ + torch.uint8, + torch.int8, + torch.int16, + torch.int32, + torch.float16, + torch.float32, +] +SUPPORTED_QSCHEMES = [ + torch.per_tensor_affine, + torch.per_tensor_symmetric, + torch.per_channel_affine, + torch.per_channel_symmetric, + torch.per_channel_affine_float_qparams, +] + + +class QuantizationSpecBase(ABC): # noqa: B024 + """Base class for different types of quantization specs that allows users to + specify how to quantize a Tensor (input/output of a Node) in the model + """ + + pass + + +@dataclass(eq=True, frozen=True) +class QuantizationSpec(QuantizationSpecBase): + """Quantization spec for common operators that allows user to specify how to + quantize a Tensor, this includes dtype, quant_min, quant_max etc. + """ + + dtype: torch.dtype + # observer or fake_quantize constructor such as + # MinMaxObserver, PerChannelHistogramObserver etc. + # or we can attach some custom args to them + # e.g. MinMaxObserver.with_args(eps=eps) + observer_or_fake_quant_ctr: _ObserverOrFakeQuantizeConstructor + quant_min: Optional[int] = None + quant_max: Optional[int] = None + qscheme: Optional[torch.qscheme] = None + ch_axis: Optional[int] = None + is_dynamic: bool = False + + def __post_init__(self): + # check dtype is one of the supported types + if self.dtype not in SUPPORTED_DTYPES: + raise TypeError(f"Unsupported dtype {self.dtype}.") + + # quant_min must be less than quant_max + if ( + self.quant_min is not None + and self.quant_max is not None + and self.quant_min > self.quant_max + ): + raise ValueError( + f"quant_min {self.quant_min} must be <= quant_max {self.quant_max}." + ) + + # check qscheme is on of the supported ones + if self.qscheme is not None and self.qscheme not in SUPPORTED_QSCHEMES: + raise ValueError(f"Unsupported qscheme {self.qscheme}.") + + # ch_axis must be less than the number of channels + # but no way to check here. Just check that it is not < 0. + if self.ch_axis is not None and self.ch_axis < 0: + raise ValueError("Ch_axis is < 0.") + + +@dataclass(eq=True, frozen=True) +class FixedQParamsQuantizationSpec(QuantizationSpecBase): + dtype: torch.dtype + scale: float + zero_point: int + quant_min: Optional[int] = None + quant_max: Optional[int] = None + qscheme: Optional[torch.qscheme] = None + + +""" +The way we refer to other points of quantization in the graph will be either +an input edge or an output value +input edge is the connection between input node and the node consuming the input, so it's a Tuple[Node, Node] +output value is an fx Node +""" +EdgeOrNode = Union[Tuple[Node, Node], Node] +EdgeOrNode.__module__ = "torch.ao.quantization.quantizer.quantizer" + + +@dataclass(eq=True, frozen=True) +class SharedQuantizationSpec(QuantizationSpecBase): + """ + Quantization spec for the Tensors whose quantization parameters are shared with other Tensors + """ + + # the edge or node to share observer or fake quant instances with + edge_or_node: EdgeOrNode + + +@dataclass(eq=True, frozen=True) +class DerivedQuantizationSpec(QuantizationSpecBase): + """Quantization spec for the Tensors whose quantization parameters are derived from other Tensors""" + + derived_from: List[EdgeOrNode] + derive_qparams_fn: Callable[[List[ObserverOrFakeQuantize]], Tuple[Tensor, Tensor]] + dtype: torch.dtype + quant_min: Optional[int] = None + quant_max: Optional[int] = None + qscheme: Optional[torch.qscheme] = None + ch_axis: Optional[int] = None + + +@dataclass +class QuantizationAnnotation: + """How are input arguemnt or output should be quantized, + expressed as QuantizationSpec, this corresponds to how a Tensor in the + operator Graph is observed (PTQ) or fake quantized (QAT) + """ + + # a map from torch.fx.Node to a type of QuantizationSpecBase + input_qspec_map: Dict[Node, Optional[QuantizationSpecBase]] = field( + default_factory=dict + ) + + # How the output of this node is quantized, expressed as QuantizationSpec + # TODO: change the value to QuantizationSpec in a separate PR + output_qspec: Optional[QuantizationSpecBase] = None + + # For a Node: node1 and edge: (node1, node2), since they are observing the same + # Tensor, we may want to implicitly share observers, this flag allows people to + # turn off this behavior for the output of the node + allow_implicit_sharing: bool = True + + # whether the node is annotated or not + _annotated: bool = False + + +class Quantizer(ABC): + def transform_for_annotation( + self, model: torch.fx.GraphModule + ) -> torch.fx.GraphModule: + """Allows for user defined transforms to run before annotating the graph. + This allows quantizer to allow quantizing part of the model that are otherwise not quantizable. + For example quantizer can + a) decompose a compound operator like scaled dot product attention, + into bmm and softmax if quantizer knows how to quantize bmm/softmax but not sdpa + or b) transform scalars to tensor to allow quantizing scalares. + + Note: this is an optional method + """ + return model + + # annotate nodes in the graph with observer or fake quant constructors + # to convey the desired way of quantization + @abstractmethod + def annotate(self, model: torch.fx.GraphModule) -> torch.fx.GraphModule: + pass + + # validate the annotated graph is supported by the backend + @abstractmethod + def validate(self, model: torch.fx.GraphModule) -> None: + pass diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/utils.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/utils.py new file mode 100644 index 0000000000000000000000000000000000000000..f25d0916018bae597ee0ef4ab4f3ddfe24eb1ff0 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/utils.py @@ -0,0 +1,49 @@ +from typing import List + +from torch.ao.quantization.pt2e.utils import _is_sym_size_node + +from torch.ao.quantization.quantizer.quantizer import QuantizationAnnotation +from torch.fx import Node + + +def _annotate_input_qspec_map(node: Node, input_node: Node, qspec): + quantization_annotation = node.meta.get( + "quantization_annotation", QuantizationAnnotation() + ) + if quantization_annotation.input_qspec_map is None: + quantization_annotation.input_qspec_map = {} + quantization_annotation.input_qspec_map[input_node] = qspec + node.meta["quantization_annotation"] = quantization_annotation + + +def _annotate_output_qspec(node: Node, qspec): + quantization_annotation = node.meta.get( + "quantization_annotation", QuantizationAnnotation() + ) + quantization_annotation.output_qspec = qspec + node.meta["quantization_annotation"] = quantization_annotation + + +def _node_only_used_for_sym_size(node: Node, partition_nodes: List[Node]): + """ + This utility is used to handle cases when dynami_shape=True tracing leads + to symint nodes in the pattern of linear module. In those cases, we need to + distinguish between the nodes that are in input for just extracting value of + some dimentions (and symint nodes) vs. the one that is activation. + For example: + graph(x, y, weight): + size_0 = torch.ops.aten.sym_size([x], [0]) + size_1 = torch.ops.aten.sym_size([y], [1]) + view_size = size_0 * size_1 + size_3 = torch.ops.aten.sym_size([x], [2]) + vie_out = torch.ops.aten.view(x, [view_size, size_3]) + return mm(view_out, weight) + In the example above y node is not actual input. It exist only to extract size_1 + """ + if _is_sym_size_node(node): + return True + + return all( + ((user not in partition_nodes) or _is_sym_size_node(user)) + for user in node.users + ) diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/x86_inductor_quantizer.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/x86_inductor_quantizer.py new file mode 100644 index 0000000000000000000000000000000000000000..da8fd9782da35c9d8065f144122ac2e62fa7a41b --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/x86_inductor_quantizer.py @@ -0,0 +1,989 @@ +import copy +import functools +import itertools +import operator +from dataclasses import dataclass +from typing import Any, Dict, List, Optional, Sequence, Set, Tuple + +import torch +import torch.nn.functional as F +from torch.ao.quantization.fake_quantize import FusedMovingAvgObsFakeQuantize +from torch.ao.quantization.observer import ( + HistogramObserver, + MovingAveragePerChannelMinMaxObserver, + PerChannelMinMaxObserver, + PlaceholderObserver, +) +from torch.ao.quantization.pt2e.graph_utils import find_sequential_partitions +from torch.ao.quantization.qconfig import _ObserverOrFakeQuantizeConstructor +from torch.ao.quantization.quantizer.quantizer import ( + QuantizationAnnotation, + QuantizationSpec, + Quantizer, + SharedQuantizationSpec, +) +from torch.ao.quantization.quantizer.xnnpack_quantizer_utils import ( + _is_annotated, + get_bias_qspec, + get_input_act_qspec, + get_output_act_qspec, + get_weight_qspec, + OperatorConfig, + OperatorPatternType, + QuantizationConfig, +) +from torch.fx import Node +from torch.fx.passes.utils.source_matcher_utils import ( + get_source_partitions, + SourcePartition, +) + +__all__ = [ + "X86InductorQuantizer", + "get_default_x86_inductor_quantization_config", +] + + +@dataclass +class _X86InductorQuantizationAnnotation(QuantizationAnnotation): + # _is_output_of_quantized_pattern: + # * Node as output node of a fusion pattern. + # * The fusion pattern supports int8 data type. + # * The fusion pattern has inputs annotated to insert observer. + _is_output_of_quantized_pattern: bool = False + + +# Operations that: +# 1. Operations are optimized to run with int8 when int8 input provided. +# 2. Operations do not support int8 input and produce fp32 output. +int8_in_int8_out_ops_pt2e: Set = { + torch.ops.aten.max_pool2d.default, + torch.ops.aten.cat.default, + torch.ops.aten.avg_pool2d.default, + torch.ops.aten.adaptive_avg_pool2d.default, + torch.ops.aten.flatten.using_ints, +} + + +# Operations support the int8 data type and exclude operations such as conv and linear. +# A superset of int8_in_int8_out_ops_pt2e incorporating additional operators. +quantizable_ops_pt2e = copy.deepcopy(int8_in_int8_out_ops_pt2e) + +QUANT_ANNOTATION_KEY = "quantization_annotation" + + +def _mark_nodes_as_annotated(nodes: List[Node]): + for node in nodes: + if node is not None: + if QUANT_ANNOTATION_KEY not in node.meta: + node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation() + node.meta[QUANT_ANNOTATION_KEY]._annotated = True + + +def _is_node_annotated(_node): + """ + return True if the node is annotated, otherwise return False + """ + return ( + QUANT_ANNOTATION_KEY in _node.meta + and _node.meta[QUANT_ANNOTATION_KEY]._annotated + ) + + +def _is_any_annotated(nodes: List[Node]): + """ + Given a list of nodes (that represents an operator pattern), + check if any of the node is annotated, return True if any of the node + is annotated, otherwise return False. + """ + return any(_is_node_annotated(node) for node in nodes) + + +def _is_all_annotated(nodes: List[Node]): + """ + Given a list of nodes (that represents an operator pattern), + return True if all of the node is annotated, otherwise return False. + """ + return all(_is_node_annotated(node) for node in nodes) + + +def _is_quantized_op_pt2e(node: torch.fx.Node): + """ + Used for pt2e flow to check if the node is a quantized node: + Case1: the node has been annotated as output node of a fusion pattern. + Case2: the node has been annotated as single quantized node. + """ + if not _is_any_annotated([node]): + # The node has not been annotated, directly return False + return False + quantization_annotation = node.meta.get(QUANT_ANNOTATION_KEY, None) + assert isinstance(quantization_annotation, _X86InductorQuantizationAnnotation) + return quantization_annotation._is_output_of_quantized_pattern + + +def _supported_quantized_operators() -> Dict[str, List[OperatorPatternType]]: + # TODO: Add more supported operators here. + supported_operators: Dict[str, List[OperatorPatternType]] = { + "conv2d": [ + [torch.nn.Conv2d], + [F.conv2d], + ], + } + + # Append Conv Optional(Add) Optioinal(ReLU) + conv_add_relu_options = itertools.product( + [torch.nn.Conv2d, F.conv2d], + [torch.add, operator.add, None], # add + [torch.nn.ReLU, F.relu, None], # relu + ) + for conv_op, add_op, relu_op in conv_add_relu_options: + if add_op is None: + # Append Conv ReLU + supported_operators["conv2d"].append([conv_op, relu_op]) # type: ignore[list-item] + elif relu_op is None: + # Append Conv Add + supported_operators["conv2d"].append([conv_op, add_op]) # type: ignore[list-item] + else: + # Append Conv Add ReLU + supported_operators["conv2d"].append([conv_op, add_op, relu_op]) # type: ignore[list-item] + + return copy.deepcopy(supported_operators) + + +def _get_supported_x86_inductor_config_and_operators() -> List[OperatorConfig]: + supported_config_and_operators: List[OperatorConfig] = [] + for quantization_config in [ + get_default_x86_inductor_quantization_config(), + ]: + ops = _supported_quantized_operators() + for pattern_list in ops.values(): + supported_config_and_operators.append( + OperatorConfig(quantization_config, pattern_list) + ) + return copy.deepcopy(supported_config_and_operators) + + +@functools.lru_cache +def get_default_x86_inductor_quantization_config(is_qat: bool = False): + act_observer_or_fake_quant_ctr: _ObserverOrFakeQuantizeConstructor = ( + FusedMovingAvgObsFakeQuantize if is_qat else HistogramObserver + ) + + # Copy from x86 default qconfig from torch/ao/quantization/qconfig.py + act_quantization_spec = QuantizationSpec( + dtype=torch.uint8, + quant_min=0, + quant_max=255, # reduce_range=False + qscheme=torch.per_tensor_affine, + is_dynamic=False, + observer_or_fake_quant_ctr=act_observer_or_fake_quant_ctr.with_args( + eps=2**-12 + ), + ) + + weight_observer_or_fake_quant_ctr: _ObserverOrFakeQuantizeConstructor = ( + FusedMovingAvgObsFakeQuantize if is_qat else PerChannelMinMaxObserver + ) + + extra_args: Dict[str, Any] = {"eps": 2**-12} + if is_qat: + # Only support per channel quant for now + extra_args["observer"] = MovingAveragePerChannelMinMaxObserver # type: ignore[dict-item] + weight_quantization_spec = QuantizationSpec( + dtype=torch.int8, + quant_min=-128, + quant_max=127, + qscheme=torch.per_channel_symmetric, + ch_axis=0, # 0 corresponding to weight shape = (oc, ic, kh, kw) of conv + is_dynamic=False, + observer_or_fake_quant_ctr=weight_observer_or_fake_quant_ctr.with_args( + **extra_args + ), + ) + bias_observer_or_fake_quant_ctr: _ObserverOrFakeQuantizeConstructor = ( + PlaceholderObserver + ) + bias_quantization_spec = QuantizationSpec( + dtype=torch.float, observer_or_fake_quant_ctr=bias_observer_or_fake_quant_ctr + ) + quantization_config = QuantizationConfig( + act_quantization_spec, + act_quantization_spec, + weight_quantization_spec, + bias_quantization_spec, + is_qat, + ) + return quantization_config + + +def _get_supported_config_and_operators() -> List[OperatorConfig]: + return _get_supported_x86_inductor_config_and_operators() + + +class X86InductorQuantizer(Quantizer): + supported_config_and_operators = _get_supported_config_and_operators() + + def __init__(self): + super().__init__() + self.global_config: QuantizationConfig = None # type: ignore[assignment] + self.operator_type_config: Dict[str, Optional[QuantizationConfig]] = {} + + @classmethod + def get_supported_quantization_configs(cls) -> List[QuantizationConfig]: + op_configs: Set[QuantizationConfig] = set({}) + for spec, _ in cls.supported_config_and_operators: + op_configs.add(spec) + return list(op_configs) + + @classmethod + def get_supported_operator_for_quantization_config( + cls, quantization_config: Optional[QuantizationConfig] + ) -> List[OperatorPatternType]: + if quantization_config is None: + all_ops = [] + for _, ops in cls.supported_config_and_operators: + all_ops.extend(ops) + return all_ops + + for config, ops in cls.supported_config_and_operators: + if config == quantization_config: + return ops + return [] + + def set_global(self, quantization_config: QuantizationConfig): + self.global_config = quantization_config + return self + + def set_config_for_operator_type( + self, operator_type: str, quantization_config: QuantizationConfig + ): + self.operator_type_config[operator_type] = quantization_config + return self + + def _annotate_conv_node_helper( + self, + conv_node: torch.fx.Node, + annotate_output: bool, + quantization_config: QuantizationConfig, + ) -> None: + """Helper function to annotate the conv node""" + input_qspec_map = {} + input_node = conv_node.args[0] + assert isinstance(input_node, Node) + input_qspec_map[input_node] = get_input_act_qspec(quantization_config) + weight_node = conv_node.args[1] + assert isinstance(weight_node, Node) + input_qspec_map[weight_node] = get_weight_qspec(quantization_config) + bias_node = None if len(conv_node.args) == 2 else conv_node.args[2] + if isinstance(bias_node, Node): + input_qspec_map[bias_node] = get_bias_qspec(quantization_config) + if annotate_output: + conv_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + else: + conv_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + ) + + def _annotate_linear_node_helper( + self, + linear_node: torch.fx.Node, + annotate_output: bool, + quantization_config: QuantizationConfig, + ) -> None: + """Helper function to annotate the linear node""" + input_qspec_map = {} + assert linear_node.target in (torch.ops.aten.linear.default,) + has_bias = len(linear_node.args) == 3 + input_index = 0 + weight_index = 1 + bias_index = 2 + + input_node = linear_node.args[input_index] + assert isinstance(input_node, Node) + input_qspec_map[input_node] = get_input_act_qspec(quantization_config) + + weight_node = linear_node.args[weight_index] + assert isinstance(weight_node, Node) + input_qspec_map[weight_node] = get_weight_qspec(quantization_config) + + bias_node = linear_node.args[bias_index] if has_bias else None + if isinstance(bias_node, Node): + input_qspec_map[bias_node] = get_bias_qspec(quantization_config) + + if annotate_output: + linear_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + else: + linear_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=input_qspec_map, _annotated=True + ) + + def _get_output_nodes_of_partitions( + self, + partition_list: List[SourcePartition], + ) -> List[torch.fx.Node]: + """Helper function to get the output node list from partition list""" + output_node_list = [] + for partition in partition_list: + if len(partition.output_nodes) > 1: + raise ValueError("Input partition has more than one output node") + output_node = partition.output_nodes[0] + assert isinstance(output_node, Node) + output_node_list.append(output_node) + if len(output_node_list) != len(partition_list): + raise ValueError( + "length of output_node_list should equal to length of partition_list" + ) + return output_node_list + + def _get_input_idx_for_binary_node( + self, + conv_gemm_node: torch.fx.Node, + binary_node: torch.fx.Node, + ): + """Helper function to check conv_gemm and extra input node index + for binary node fused with conv_gemm. + """ + conv_gemm_node_idx = None + extra_input_node_idx = None + if (binary_node.args[0].op == "call_function") and ( # type: ignore[union-attr] + binary_node.args[0] == conv_gemm_node + ): + conv_gemm_node_idx = 0 + extra_input_node_idx = 1 + elif (binary_node.args[1].op == "call_function") and ( # type: ignore[union-attr] + binary_node.args[1] == conv_gemm_node + ): + conv_gemm_node_idx = 1 + extra_input_node_idx = 0 + extra_input_node = binary_node.args[extra_input_node_idx] # type: ignore[index] + assert isinstance(extra_input_node, Node) + return conv_gemm_node_idx, extra_input_node_idx + + def annotate(self, model: torch.fx.GraphModule) -> torch.fx.GraphModule: + """just handling global spec for now""" + model = self._annotate_for_static_quantization_config(model) + return model + + def _annotate_for_static_quantization_config( + self, model: torch.fx.GraphModule + ) -> torch.fx.GraphModule: + r""" + High-level description of quantization recipe for X86 Inductor Backend: + Step 1: Apply quantization recipe for fusion patterns of conv/linear to enable int8 data type actively. + Step 2: Propagate quantization annotation for patterns besides conv/linear. Go through the pattern in model + from start to the end. If a pattern supports computation with int8 data type and inputs connected to + quantized patterns, annotate its inputs as quantized pattern. + Step 3: Since in step 2, we only annotate the inputs of quantized pattern. For some quantized patterns, + such as maxpool2d, which only supports output with int8 data type when the input is with int8 data type, + we need to annotate the output of this pattern. + """ + + config = self.global_config + + # Step1: Recipe of fusion patterns like conv/linear. + if config.is_qat: + # Annotate QAT specific pattern: mainly due to BN not folded in prepare_qat + self._annotate_qat_conv2d_fusion_pattern(model, config) + + self._annotate_conv2d_fusion_pattern(model, config) + + # Step2: Recipe to propagate annotation for patterns beside conv/linear. + # Go through all the nodes from start to end. + # Recipe refer to https://github.com/intel/intel-extension-for-pytorch/blob/ + # 90d19323d96afc53fcc22ba5a7bb3fb07fdd6c1c/intel_extension_for_pytorch/quantization/_recipe.py#L538 + for node in model.graph.nodes: + self._annotation_propagation_quantizable_pattern(node, config) + + # Step3: For quantizable ops, such as maxpool2d, we need to quantize its output if it is quantized + # in inputs. So, we can fuse dq-operator-q into a quantized op. + # Refer to https://github.com/intel/intel-extension-for-pytorch/blob/ + # 90d19323d96afc53fcc22ba5a7bb3fb07fdd6c1c/intel_extension_for_pytorch/quantization/_recipe.py#L487 + for node in model.graph.nodes: + self._annotate_output_for_int8_in_int8_out_pattern(node, config) + + return model + + def _annotate_qat_conv2d_fusion_pattern( + self, model: torch.fx.GraphModule, config: QuantizationConfig + ): + # Annotate QAT Specific patterns + self._annotate_qat_conv2d_bn_binary_unary(model, config) + self._annotate_qat_conv2d_bn_binary(model, config) + self._annotate_qat_conv2d_bn_unary(model, config) + self._annotate_qat_conv2d_bn(model, config) + + def _annotate_qat_conv2d_bn_binary_unary( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + fused_partitions = find_sequential_partitions( + gm, [torch.nn.Conv2d, torch.nn.BatchNorm2d, operator.add, torch.nn.ReLU] + ) + for fused_partition in fused_partitions: + ( + conv_partition, + bn_partition, + binary_partition, + unary_partition, + ) = fused_partition + + ( + conv_node, + bn_output_node, + binary_node, + unary_node, + ) = self._get_output_nodes_of_partitions( + [conv_partition, bn_partition, binary_partition, unary_partition] + ) + if len(bn_output_node.users) != 1: + # Conv BN pattern should only has 1 user. + continue + ( + bn_output_node_idx, + extra_input_node_idx, + ) = self._get_input_idx_for_binary_node(bn_output_node, binary_node) + if (bn_output_node_idx is None) or (extra_input_node_idx is None): + continue + if bn_output_node != binary_node.args[bn_output_node_idx]: + raise ValueError(f"{bn_output_node} doesn't match input of binary node") + extra_input_node = binary_node.args[extra_input_node_idx] + + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + continue + + if _is_annotated([unary_node, binary_node, bn_output_node, conv_node]): + continue + + self._annotate_conv_node_helper(conv_node, False, quantization_config) + + binary_node_input_qspec_map = {} + binary_node_input_qspec_map[extra_input_node] = get_input_act_qspec( + quantization_config + ) + binary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=binary_node_input_qspec_map, + _annotated=True, + ) + unary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + # TODO Remove the annotate of output in QAT when qat util support pattern matcher. + output_qspec=get_output_act_qspec(quantization_config), # type: ignore[arg-type] + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + nodes_to_mark_annotated = list(conv_partition.nodes) + nodes_to_mark_annotated.extend(list(bn_partition.nodes)) + nodes_to_mark_annotated.extend(list(binary_partition.nodes)) + nodes_to_mark_annotated.extend(list(unary_partition.nodes)) + _mark_nodes_as_annotated(nodes_to_mark_annotated) + + def _annotate_qat_conv2d_bn_binary( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + fused_partitions = find_sequential_partitions( + gm, [torch.nn.Conv2d, torch.nn.BatchNorm2d, operator.add] + ) + for fused_partition in fused_partitions: + conv_partition, bn_partition, binary_partition = fused_partition + ( + conv_node, + bn_output_node, + binary_node, + ) = self._get_output_nodes_of_partitions( + [conv_partition, bn_partition, binary_partition] + ) + if len(bn_output_node.users) != 1: + # Conv BN pattern should only has 1 user. + continue + ( + bn_output_node_idx, + extra_input_node_idx, + ) = self._get_input_idx_for_binary_node(bn_output_node, binary_node) + if (bn_output_node_idx is None) or (extra_input_node_idx is None): + continue + if bn_output_node != binary_node.args[bn_output_node_idx]: + raise ValueError(f"{bn_output_node} doesn't match input of binary node") + + extra_input_node = binary_node.args[extra_input_node_idx] + + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + continue + + if _is_annotated([binary_node, bn_output_node, conv_node]): + continue + + self._annotate_conv_node_helper(conv_node, False, quantization_config) + + binary_node_input_qspec_map = {} + binary_node_input_qspec_map[extra_input_node] = get_input_act_qspec( + quantization_config + ) + binary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=binary_node_input_qspec_map, + # TODO Remove the annotate of output in QAT when qat util support pattern matcher. + output_qspec=get_output_act_qspec(quantization_config), # type: ignore[arg-type] + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + nodes_to_mark_annotated = list(conv_partition.nodes) + nodes_to_mark_annotated.extend(list(bn_partition.nodes)) + nodes_to_mark_annotated.extend(list(binary_partition.nodes)) + _mark_nodes_as_annotated(nodes_to_mark_annotated) + + def _annotate_qat_conv2d_bn_unary( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + fused_partitions = [] + unary_patterns = [ + [torch.nn.Conv2d, torch.nn.BatchNorm2d, torch.nn.ReLU], + [torch.nn.Conv2d, torch.nn.BatchNorm2d, torch.nn.Hardtanh], + [torch.nn.Conv2d, torch.nn.BatchNorm2d, torch.nn.ReLU6], + ] + for unary_pattern in unary_patterns: + partitions = find_sequential_partitions(gm, unary_pattern) + if partitions: + # Extend the fused_partitions if partitions is not empty + fused_partitions.extend(partitions) + + for fused_partition in fused_partitions: + conv_partition, bn_partition, unary_partition = fused_partition + ( + conv_node, + bn_output_node, + unary_node, + ) = self._get_output_nodes_of_partitions( + [conv_partition, bn_partition, unary_partition] + ) + + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + continue + + if _is_annotated([unary_node, bn_output_node, conv_node]): + continue + + self._annotate_conv_node_helper(conv_node, False, quantization_config) + unary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + # TODO Remove the annotate of output in QAT when qat util support pattern matcher. + output_qspec=get_output_act_qspec(quantization_config), # type: ignore[arg-type] + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + nodes_to_mark_annotated = list(conv_partition.nodes) + nodes_to_mark_annotated.extend(list(bn_partition.nodes)) + nodes_to_mark_annotated.extend(list(unary_partition.nodes)) + _mark_nodes_as_annotated(nodes_to_mark_annotated) + + def _annotate_qat_conv2d_bn( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + fused_partitions = find_sequential_partitions( + gm, [torch.nn.Conv2d, torch.nn.BatchNorm2d] + ) + for fused_partition in fused_partitions: + conv_partition, bn_partition = fused_partition + conv_node, bn_output_node = self._get_output_nodes_of_partitions( + [conv_partition, bn_partition] + ) + + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + continue + + if _is_annotated([bn_output_node, conv_node]): + continue + + self._annotate_conv_node_helper(conv_node, False, quantization_config) + bn_output_node.meta[ + QUANT_ANNOTATION_KEY + ] = _X86InductorQuantizationAnnotation( + # TODO Remove the annotate of output in QAT when qat util support pattern matcher. + output_qspec=get_output_act_qspec(quantization_config), # type: ignore[arg-type] + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + nodes_to_mark_annotated = list(conv_partition.nodes) + nodes_to_mark_annotated.extend(list(bn_partition.nodes)) + _mark_nodes_as_annotated(nodes_to_mark_annotated) + + def _annotate_conv2d_fusion_pattern( + self, model: torch.fx.GraphModule, config: QuantizationConfig + ): + self._annotate_conv2d_binary_unary(model, config) + self._annotate_conv2d_binary(model, config) + self._annotate_conv2d_unary(model, config) + self._annotate_conv2d(model, config) + self._annotate_linear_unary(model, config) + self._annotate_linear(model, config) + + def _annotate_conv2d_binary_unary( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + # Conv2d + add + unary op + fused_partitions = find_sequential_partitions( + gm, [torch.nn.Conv2d, operator.add, torch.nn.ReLU] + ) + for fused_partition in fused_partitions: + conv_partition, binary_partition, unary_partition = fused_partition + conv_node, binary_node, unary_node = self._get_output_nodes_of_partitions( + [conv_partition, binary_partition, unary_partition] + ) + if len(conv_node.users) != 1: + # Conv Node should only has 1 user node + continue + conv_node_idx, extra_input_node_idx = self._get_input_idx_for_binary_node( + conv_node, binary_node + ) + if (conv_node_idx is None) or (extra_input_node_idx is None): + continue + if conv_node != binary_node.args[conv_node_idx]: + raise ValueError(f"{conv_node} doesn't match input of binary node") + extra_input_node = binary_node.args[extra_input_node_idx] + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + # No conv node found to be fused with add + continue + if _is_annotated([unary_node, binary_node, conv_node]): + continue + self._annotate_conv_node_helper(conv_node, False, quantization_config) + binary_node_input_qspec_map = {} + binary_node_input_qspec_map[extra_input_node] = get_input_act_qspec( + quantization_config + ) + binary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=binary_node_input_qspec_map, + _annotated=True, + ) + unary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + + def _annotate_conv2d_binary( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + # Conv2d + add + fused_partitions = find_sequential_partitions( + gm, [torch.nn.Conv2d, operator.add] + ) + for fused_partition in fused_partitions: + conv_partition, binary_partition = fused_partition + conv_node, binary_node = self._get_output_nodes_of_partitions( + [conv_partition, binary_partition] + ) + if len(conv_node.users) != 1: + # Conv Node should only has 1 user node + continue + conv_node_idx, extra_input_node_idx = self._get_input_idx_for_binary_node( + conv_node, binary_node + ) + if (conv_node_idx is None) or (extra_input_node_idx is None): + continue + if conv_node != binary_node.args[conv_node_idx]: + raise ValueError(f"{conv_node} doesn't match input of binary node") + extra_input_node = binary_node.args[extra_input_node_idx] + assert isinstance(conv_node, Node) + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + # No conv node found to be fused with add + continue + if _is_annotated([binary_node, conv_node]): + continue + self._annotate_conv_node_helper(conv_node, False, quantization_config) + binary_node_input_qspec_map = {} + binary_node_input_qspec_map[extra_input_node] = get_input_act_qspec( + quantization_config + ) + binary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=binary_node_input_qspec_map, + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + + def _annotate_conv2d_unary( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + fused_partitions = [] + unary_patterns = [ + [torch.nn.Conv2d, torch.nn.ReLU], + [torch.nn.Conv2d, torch.nn.Hardtanh], + [torch.nn.Conv2d, torch.nn.ReLU6], + ] + for unary_pattern in unary_patterns: + partitions = find_sequential_partitions(gm, unary_pattern) + if partitions: + # Extend the fused_partitions if partitions is not empty + fused_partitions.extend(partitions) + + for fused_partition in fused_partitions: + conv_partition, unary_partition = fused_partition + conv_node, unary_node = self._get_output_nodes_of_partitions( + [conv_partition, unary_partition] + ) + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + continue + if _is_annotated([unary_node, conv_node]): + continue + self._annotate_conv_node_helper(conv_node, False, quantization_config) + unary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + + def _annotate_conv2d( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + conv_partitions = get_source_partitions( + gm.graph, [torch.nn.Conv2d, torch.nn.functional.conv2d] + ) + conv_partitions = list(itertools.chain(*conv_partitions.values())) + for conv_partition in conv_partitions: + if len(conv_partition.output_nodes) > 1: + raise ValueError("conv partition has more than one output node") + conv_node = conv_partition.output_nodes[0] + if ( + conv_node.op != "call_function" + or conv_node.target != torch.ops.aten.conv2d.default + ): + raise ValueError(f"{conv_node} is not an aten conv2d operator") + # skip annotation if it is already annotated + if _is_annotated([conv_node]): + continue + self._annotate_conv_node_helper(conv_node, True, quantization_config) + + def _annotate_maxpool2d( + self, node: Node, quantization_config: QuantizationConfig + ) -> None: + if node.target is not torch.ops.aten.max_pool2d.default: + return + maxpool_node = node + if _is_any_annotated( + [ + maxpool_node, + ] + ): + return + input_node = maxpool_node.args[0] + assert isinstance(input_node, Node) + input_qspec_map = {} + input_qspec_map[input_node] = get_input_act_qspec(quantization_config) + maxpool_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + + def _annotate_cat( + self, node: Node, quantization_config: QuantizationConfig + ) -> None: + cat_node = node + input_nodes = cat_node.args[0] + assert isinstance(input_nodes, Sequence) + first_input_node = input_nodes[0] + input_qspec_map = {} + assert isinstance(first_input_node, Node) + assert isinstance(cat_node, Node) + input_qspec_map[first_input_node] = get_input_act_qspec(quantization_config) + share_qparams_with_input_act0_qspec = SharedQuantizationSpec( + (first_input_node, cat_node) + ) + + for input_node in input_nodes[1:]: + if input_node not in input_qspec_map: + # There has the case of cat same nodes: torch.cat([input0, input0], 1) + assert isinstance(input_node, Node) + input_qspec_map[input_node] = share_qparams_with_input_act0_qspec + + cat_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + + def _annotation_propagation_quantizable_pattern( + self, node: Node, quantization_config: QuantizationConfig + ) -> None: + # Propagate annotation to quantizable patterns. + if ( + (node.target in quantizable_ops_pt2e) + and (not _is_any_annotated([node])) + and (node.op == "call_function") + ): + + def is_all_inputs_connected_to_quantized_op(input_nodes): + # Ensure all the inputs connect to fusion pattern or quantized node + for input_node in input_nodes: + if not _is_quantized_op_pt2e(input_node): + return False + return True + + if node.target is torch.ops.aten.max_pool2d.default: + # Recipe of maxpool2d: check input arg[0] of maxpool2d is quantized or not + input_nodes_to_check = [node.all_input_nodes[0]] + if not is_all_inputs_connected_to_quantized_op(input_nodes_to_check): + return + self._annotate_maxpool2d(node, quantization_config) + return + elif node.target is torch.ops.aten.cat.default: + input_nodes_to_check = node.all_input_nodes + if not is_all_inputs_connected_to_quantized_op(input_nodes_to_check): + return + self._annotate_cat(node, quantization_config) + else: + input_node = node.all_input_nodes[0] + if not is_all_inputs_connected_to_quantized_op( + [ + input_node, + ] + ): + return + input_qspec_map = {} + input_qspec_map[input_node] = get_input_act_qspec(quantization_config) + node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + return + + def _annotate_output_share_observer_as_input( + self, input_node: Node, source_node: Node + ): + source_node_quantization_annotation = ( + source_node.meta[QUANT_ANNOTATION_KEY] + if QUANT_ANNOTATION_KEY in source_node.meta + else None + ) + if ( + source_node_quantization_annotation + and source_node_quantization_annotation._is_output_of_quantized_pattern + ): + edge_or_node = (input_node, source_node) + source_node_quantization_annotation.output_qspec = SharedQuantizationSpec( + edge_or_node + ) + return + + def _annotate_output_for_int8_in_int8_out_pattern( + self, node: Node, quantization_config: QuantizationConfig + ) -> None: + r""" + Check and insert observer at output of node in int8_in_int8_out_ops_pt2e if needed. + Recipe refers to https://github.com/intel/intel-extension-for-pytorch/blob/ + 90d19323d96afc53fcc22ba5a7bb3fb07fdd6c1c/intel_extension_for_pytorch/quantization/_utils.py#L495 + """ + edge_or_node: Tuple[Node, Node] + if (node.target in int8_in_int8_out_ops_pt2e) and (_is_any_annotated([node])): + if node.target == torch.ops.aten.max_pool2d.default: + maxpool_node = node + if not _is_all_annotated( + [ + maxpool_node, + ] + ): + return + # Get the quantization_annotation from getitem_node + maxpool_node_quantization_annotation = ( + maxpool_node.meta[QUANT_ANNOTATION_KEY] + if QUANT_ANNOTATION_KEY in maxpool_node.meta + else None + ) + if ( + maxpool_node_quantization_annotation + and maxpool_node_quantization_annotation._is_output_of_quantized_pattern + ): + # Annotate the output_qspec of getitem_node + input_act = maxpool_node.args[0] + assert isinstance(input_act, Node) + assert isinstance(maxpool_node, Node) + edge_or_node = (input_act, maxpool_node) + maxpool_node_quantization_annotation.output_qspec = ( + SharedQuantizationSpec(edge_or_node) + ) + else: + input_node = node.all_input_nodes[0] + self._annotate_output_share_observer_as_input(input_node, node) + return + + def _annotate_linear( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + linear_partitions = get_source_partitions( + gm.graph, [torch.nn.Linear, torch.nn.functional.linear] + ) + linear_partitions = list(itertools.chain(*linear_partitions.values())) + for partition in linear_partitions: + if len(partition.output_nodes) > 1: + raise ValueError( + "Linear partition cannot have more than one output node" + ) + linear_node = partition.output_nodes[0] + if linear_node.op != "call_function" or linear_node.target not in ( + torch.ops.aten.linear.default, + ): + raise ValueError(f"{linear_node} is not an aten linear operator") + # skip annotation if it is already annotated + if _is_annotated([linear_node]): + continue + self._annotate_linear_node_helper(linear_node, True, quantization_config) + + def _annotate_linear_unary( + self, gm: torch.fx.GraphModule, quantization_config: QuantizationConfig + ) -> None: + postop_list = [ + torch.nn.ReLU, + torch.nn.LeakyReLU, + torch.nn.Tanh, + ] + fused_partitions: List[tuple] = [] + for postop in postop_list: + fused_partitions = fused_partitions + find_sequential_partitions( + gm, [torch.nn.Linear, postop] + ) + for fused_partition in fused_partitions: + linear_partition, unary_partition = fused_partition + linear_node, unary_node = self._get_output_nodes_of_partitions( + [linear_partition, unary_partition] + ) + if linear_node.op != "call_function" or linear_node.target not in ( + torch.ops.aten.linear.default, + ): + continue + if _is_annotated([unary_node, linear_node]): + continue + self._annotate_linear_node_helper(linear_node, False, quantization_config) + unary_node.meta[QUANT_ANNOTATION_KEY] = _X86InductorQuantizationAnnotation( + _annotated=True, + _is_output_of_quantized_pattern=True, + ) + + def validate(self, model: torch.fx.GraphModule) -> None: + pass + + @classmethod + def get_supported_operators(cls) -> List[OperatorConfig]: + return cls.supported_config_and_operators diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/xnnpack_quantizer.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/xnnpack_quantizer.py new file mode 100644 index 0000000000000000000000000000000000000000..d01164489cf06fd579bc9b7f9866fc5f56feff65 --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/xnnpack_quantizer.py @@ -0,0 +1,449 @@ +from __future__ import annotations + +import copy +import functools + +from typing import Any, Callable, Dict, List, Optional, Set + +import torch +import torch._dynamo as torchdynamo +import torch.nn.functional as F +from torch.ao.quantization.fake_quantize import ( + FakeQuantize, + FusedMovingAvgObsFakeQuantize, +) +from torch.ao.quantization.observer import ( + HistogramObserver, + MinMaxObserver, + MovingAverageMinMaxObserver, + MovingAveragePerChannelMinMaxObserver, + PerChannelMinMaxObserver, + PlaceholderObserver, +) + +from torch.ao.quantization.qconfig import _ObserverOrFakeQuantizeConstructor + +from torch.ao.quantization.quantizer import QuantizationSpec, Quantizer + +from torch.ao.quantization.quantizer.xnnpack_quantizer_utils import ( + _convert_scalars_to_attrs, + OP_TO_ANNOTATOR, + OperatorConfig, + OperatorPatternType, + propagate_annotation, + QuantizationConfig, +) + +from torch.fx import Node + + +__all__ = [ + "XNNPACKQuantizer", + "get_symmetric_quantization_config", +] + + +def _get_dynamo_graph(function: Callable, inputs) -> torch.fx.Graph: + gm, _ = torchdynamo.export(function, aten_graph=True)(*inputs) + gm.graph.eliminate_dead_code() + return gm.graph + + +def _get_linear_patterns(input_size: List[int]): + in_channels = input_size[-1] + out_channels = 8 # hard coding but this should not matter + weight = torch.ones((out_channels, in_channels)) + bias = torch.ones((out_channels,)) + act = torch.ones(input_size) + + def linear_op(act, weight, bias=None): + return F.linear(act, weight, bias) + + pattern_w_bias = _get_dynamo_graph(linear_op, (act, weight, bias)) + pattern_wo_bias = _get_dynamo_graph(linear_op, (act, weight)) + return [pattern_w_bias, pattern_wo_bias] + + +def _supported_symmetric_quantized_operators() -> Dict[str, List[OperatorPatternType]]: + supported_operators: Dict[str, List[OperatorPatternType]] = { + # Both conv and linear should be able to handle relu + hardtanh fusion since + # those are clamp ops + "conv2d": [ + [torch.nn.Conv2d, torch.nn.ReLU], + [torch.nn.Conv2d, F.relu], + [F.conv2d, torch.nn.ReLU], + [F.conv2d, F.relu], + ], + "linear": [[torch.nn.Linear], [F.linear]], + "add": [[torch.add]], + "max_pool2d": [[torch.nn.MaxPool2d], [F.max_pool2d]], + "adaptive_avg_pool2d": [ + [torch.nn.AdaptiveAvgPool2d], + [F.adaptive_avg_pool2d], + ], + } + return copy.deepcopy(supported_operators) + + +def _get_supported_symmetric_config_and_operators() -> List[OperatorConfig]: + supported_config_and_operators: List[OperatorConfig] = [] + for quantization_config in [ + get_symmetric_quantization_config(), + get_symmetric_quantization_config(is_qat=True), + get_symmetric_quantization_config(is_per_channel=True), + get_symmetric_quantization_config(is_per_channel=True, is_qat=True), + ]: + ops = _supported_symmetric_quantized_operators() + for pattern_list in ops.values(): + supported_config_and_operators.append( + OperatorConfig(quantization_config, pattern_list) + ) + return copy.deepcopy(supported_config_and_operators) + + +@functools.lru_cache +def get_symmetric_quantization_config( + is_per_channel: bool = False, + is_qat: bool = False, + is_dynamic: bool = False, +): + extra_args: Dict[str, Any] = {"eps": 2**-12} + if is_qat: + if is_dynamic: + act_observer_or_fake_quant_ctr = FakeQuantize + dynamic_quant_observer = MovingAverageMinMaxObserver.with_args( + averaging_constant=1 + ) + extra_args["observer"] = dynamic_quant_observer + else: + act_observer_or_fake_quant_ctr = FusedMovingAvgObsFakeQuantize # type: ignore[assignment] + else: + if is_dynamic: + act_observer_or_fake_quant_ctr = PlaceholderObserver # type: ignore[assignment] + else: + act_observer_or_fake_quant_ctr = HistogramObserver # type: ignore[assignment] + + act_quantization_spec = QuantizationSpec( + dtype=torch.int8, + quant_min=-128, + quant_max=127, + qscheme=torch.per_tensor_affine, + is_dynamic=is_dynamic, + observer_or_fake_quant_ctr=act_observer_or_fake_quant_ctr.with_args( + **extra_args, + ), + ) + weight_qscheme = ( + torch.per_channel_symmetric if is_per_channel else torch.per_tensor_symmetric + ) + weight_observer_or_fake_quant_ctr: _ObserverOrFakeQuantizeConstructor = ( + MinMaxObserver + ) + if is_qat: + # TODO: qat + per channel? + weight_observer_or_fake_quant_ctr = FusedMovingAvgObsFakeQuantize + elif is_per_channel: + weight_observer_or_fake_quant_ctr = PerChannelMinMaxObserver + + extra_args: Dict[str, Any] = {"eps": 2**-12} + if is_qat: + if weight_qscheme == torch.per_tensor_symmetric: + extra_args["observer"] = MovingAverageMinMaxObserver + else: + extra_args["observer"] = MovingAveragePerChannelMinMaxObserver # type: ignore[dict-item] + weight_quantization_spec = QuantizationSpec( + dtype=torch.int8, + quant_min=-127, + quant_max=127, + qscheme=weight_qscheme, + ch_axis=0, + is_dynamic=False, + observer_or_fake_quant_ctr=weight_observer_or_fake_quant_ctr.with_args( + **extra_args + ), + ) + + bias_quantization_spec = None + if is_dynamic: + quantization_config = QuantizationConfig( + act_quantization_spec, + None, + weight_quantization_spec, + bias_quantization_spec, + is_qat, + ) + else: + quantization_config = QuantizationConfig( + act_quantization_spec, + act_quantization_spec, + weight_quantization_spec, + bias_quantization_spec, + is_qat, + ) + return quantization_config + + +def _get_supported_config_and_operators() -> List[OperatorConfig]: + return _get_supported_symmetric_config_and_operators() + + +def _get_module_name_filter(module_name: str): + """Get the module_name_filter function for a given module name, the filter accepts + a node and checks if the node comes from a module that has certain module name + + For example: + node: linear_op = call_function[...](...) # comes from a module with name blocks.sub.linear1 + + + >> module_name_filter = _get_module_name_filter("blocks.sub") + >> print(module_name_filter(node)) + True # the node is from "blocks.sub" based on the fully qualified name "blocks.sub.linear1" + """ + + def module_name_filter(n: Node) -> bool: + # example: { + # 'L__self___sub': ("L['self'].sub", ), + # 'L__self___sub_linear': ("L['self'].sub.linear", ) + # } + # get_attr nodes doesn't have nn_module_stack? + nn_module_stack = n.meta.get("nn_module_stack", {}) + names = [ + n[len("L__self___") :].replace("_", ".") for n in nn_module_stack.keys() + ] + return module_name in names + + return module_name_filter + + +def _get_module_type_filter(tp: Callable): + """Get the module_type_filter function for a given module type, the filter accepts + a node and checks if the node comes from a module that has certain module type + + For example: + node: linear_op = call_function[...](...) # comes from a module with type Block -> Sub -> Linear + + + >> module_type_filter = _get_module_type_filter(Sub) # submodule with type `Sub`, under the `Block` submodule + >> print(module_type_filter(node)) + True # the node is from the submodule `Sub` (same for `Block` and `Linear` as well) + """ + + def module_type_filter(n: Node) -> bool: + # example: { + # 'L__self___sub': ("L['self'].sub", ), + # 'L__self___sub_linear': ("L['self'].sub.linear", ) + # } + nn_module_stack = n.meta.get("nn_module_stack", {}) + types = [t for _, t in nn_module_stack.values()] + return tp in types + + return module_type_filter + + +def _get_not_module_type_or_name_filter( + tp_list: List[Callable], module_name_list: List[str] +) -> Callable[[Node], bool]: + module_type_filters = [_get_module_type_filter(tp) for tp in tp_list] + module_name_list_filters = [_get_module_name_filter(m) for m in module_name_list] + + def not_module_type_or_name_filter(n: Node) -> bool: + return not any(f(n) for f in module_type_filters + module_name_list_filters) + + return not_module_type_or_name_filter + + +class XNNPACKQuantizer(Quantizer): + supported_config_and_operators = _get_supported_config_and_operators() + STATIC_QAT_ONLY_OPS = [ + "conv_bn_relu", + "conv_bn", + ] + + # static quantization ops (both PTQ and QAT) + STATIC_OPS = [ + "linear", + "conv_relu", + "conv", + "adaptive_avg_pool2d", + # TODO: move this to BoltNNQuantizer? + "gru_io_only", + "max_pool2d", + "add_relu", + "add", + "mul_relu", + "mul", + "cat", + ] + + DYNAMIC_OPS = [ + "linear", + ] + + def __init__(self): + super().__init__() + self.global_config: Optional[QuantizationConfig] = None + self.operator_type_config: Dict[ + torch._ops.OpOverloadPacket, Optional[QuantizationConfig] + ] = {} + self.module_type_config: Dict[Callable, Optional[QuantizationConfig]] = {} + self.module_name_config: Dict[str, Optional[QuantizationConfig]] = {} + + @classmethod + def get_supported_quantization_configs(cls) -> List[QuantizationConfig]: + op_configs: Set[QuantizationConfig] = set({}) + for spec, _ in cls.supported_config_and_operators: + op_configs.add(spec) + return list(op_configs) + + @classmethod + def get_supported_operator_for_quantization_config( + cls, quantization_config: Optional[QuantizationConfig] + ) -> List[OperatorPatternType]: + if quantization_config is None: + all_ops = [] + for _, ops in cls.supported_config_and_operators: + all_ops.extend(ops) + return all_ops + + for config, ops in cls.supported_config_and_operators: + # note: this assumes each entry in cls.supported_spec_and_operators + # corresponds to one spec, e.g. we don't have + # [(spec1, op_list1), (spec1, op_list2), (spec2, op_list3)] + # where the first and second entry have the same spec but did not + # merge the op list + if config == quantization_config: + return ops + return [] + + def set_global(self, quantization_config: QuantizationConfig) -> XNNPACKQuantizer: + self.global_config = quantization_config + return self + + def set_operator_type( + self, + operator_type: torch._ops.OpOverloadPacket, + quantization_config: QuantizationConfig, + ) -> XNNPACKQuantizer: + self.operator_type_config[operator_type] = quantization_config + return self + + def set_module_type( + self, module_type: Callable, quantization_config: QuantizationConfig + ): + """Set quantization_config for a submodule with type: `module_type`, for example: + quantizer.set_module_name(Sub) or quantizer.set_module_name(nn.Linear), it will quantize all supported operator/operator + patterns in the submodule with this module type with the given `quantization_config` + """ + self.module_type_config[module_type] = quantization_config + return self + + def set_module_name( + self, module_name: str, quantization_config: Optional[QuantizationConfig] + ): + """Set quantization_config for a submodule with name: `module_name`, for example: + quantizer.set_module_name("blocks.sub"), it will quantize all supported operator/operator + patterns in the submodule with this module name with the given `quantization_config` + """ + assert ( + quantization_config is not None + ), " quantization_config == None is not supported yet" + self.module_name_config[module_name] = quantization_config + return self + + def transform_for_annotation( + self, model: torch.fx.GraphModule + ) -> torch.fx.GraphModule: + """Transforms scalar values to tensor attributes""" + return _convert_scalars_to_attrs(model) + + def annotate(self, model: torch.fx.GraphModule) -> torch.fx.GraphModule: + """just handling global spec for now""" + # hacked for handling dynamic linear quant. will fix later. + if self.global_config and self.global_config.input_activation.is_dynamic: # type: ignore[union-attr] + model = self._annotate_for_dynamic_quantization_config(model) + else: + model = self._annotate_for_static_quantization_config(model) + propagate_annotation(model) + return model + + def _annotate_all_static_patterns( + self, + model: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, + ) -> torch.fx.GraphModule: + # TODO: implement the support for None to be canceling out previous annotations + if quantization_config is None: + return model + + if quantization_config.is_qat: + for op in self.STATIC_QAT_ONLY_OPS: + OP_TO_ANNOTATOR[op](model, quantization_config, filter_fn) + for op in self.STATIC_OPS: + OP_TO_ANNOTATOR[op](model, quantization_config, filter_fn) + return model + + def _annotate_all_dynamic_patterns( + self, + model: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, + ) -> torch.fx.GraphModule: + # TODO: implement the support for None to be canceling out previous annotations + if quantization_config is None: + return model + + for op in self.DYNAMIC_OPS: + OP_TO_ANNOTATOR[op](model, quantization_config, filter_fn) + return model + + def _annotate_for_static_quantization_config( + self, model: torch.fx.GraphModule + ) -> torch.fx.GraphModule: + module_name_list = list(self.module_name_config.keys()) + for module_name, config in self.module_name_config.items(): + self._annotate_all_static_patterns( + model, config, _get_module_name_filter(module_name) + ) + + tp_list = list(self.module_type_config.keys()) + for module_type, config in self.module_type_config.items(): + self._annotate_all_static_patterns( + model, config, _get_module_type_filter(module_type) + ) + + self._annotate_all_static_patterns( + model, + self.global_config, + _get_not_module_type_or_name_filter(tp_list, module_name_list), + ) + return model + + def _annotate_for_dynamic_quantization_config( + self, model: torch.fx.GraphModule + ) -> torch.fx.GraphModule: + module_name_list = list(self.module_name_config.keys()) + for module_name, config in self.module_name_config.items(): + self._annotate_all_dynamic_patterns( + model, config, _get_module_name_filter(module_name) + ) + + tp_list = list(self.module_type_config.keys()) + for module_type, config in self.module_type_config.items(): + self._annotate_all_dynamic_patterns( + model, config, _get_module_type_filter(module_type) + ) + + self._annotate_all_dynamic_patterns( + model, + self.global_config, + _get_not_module_type_or_name_filter(tp_list, module_name_list), + ) + return model + + def validate(self, model: torch.fx.GraphModule) -> None: + pass + + @classmethod + def get_supported_operators(cls) -> List[OperatorConfig]: + return cls.supported_config_and_operators diff --git a/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/xnnpack_quantizer_utils.py b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/xnnpack_quantizer_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..df5f0d09f63f67d1d2728ff6eb6f85316105eefb --- /dev/null +++ b/videollama2/lib/python3.10/site-packages/torch/ao/quantization/quantizer/xnnpack_quantizer_utils.py @@ -0,0 +1,938 @@ +import itertools +import operator +from dataclasses import dataclass +from typing import Callable, Dict, List, NamedTuple, Optional + +import torch +import torch.nn.functional as F +from torch.ao.quantization.fx.utils import get_new_attr_name_with_prefix +from torch.ao.quantization.pt2e.graph_utils import find_sequential_partitions +from torch.ao.quantization.pt2e.utils import ( + _conv1d_bn_example_inputs, + _conv2d_bn_example_inputs, + get_aten_graph_module, +) +from torch.ao.quantization.quantizer import ( + QuantizationAnnotation, + QuantizationSpec, + QuantizationSpecBase, + SharedQuantizationSpec, +) + +from torch.ao.quantization.quantizer.utils import ( + _annotate_input_qspec_map, + _annotate_output_qspec, +) +from torch.fx import Node +from torch.fx.passes.utils.matcher_with_name_node_map_utils import ( + SubgraphMatcherWithNameNodeMap, +) +from torch.fx.passes.utils.source_matcher_utils import get_source_partitions + + +__all__ = [ + "OperatorConfig", + "OperatorPatternType", + "QuantizationConfig", + "get_input_act_qspec", + "get_output_act_qspec", + "get_weight_qspec", + "get_bias_qspec", + "OP_TO_ANNOTATOR", + "propagate_annotation", +] + + +# In the absence of better name, just winging it with QuantizationConfig +@dataclass(eq=True, frozen=True) +class QuantizationConfig: + input_activation: Optional[QuantizationSpec] + output_activation: Optional[QuantizationSpec] + weight: Optional[QuantizationSpec] + bias: Optional[QuantizationSpec] + # TODO: remove, since we can use observer_or_fake_quant_ctr to express this + is_qat: bool = False + + +OperatorPatternType = List[Callable] +OperatorPatternType.__module__ = ( + "torch.ao.quantization.quantizer.xnnpack_quantizer_utils" +) + +AnnotatorType = Callable[ + [ + torch.fx.GraphModule, + Optional[QuantizationConfig], + Optional[Callable[[Node], bool]], + ], + Optional[List[List[Node]]], +] +OP_TO_ANNOTATOR: Dict[str, AnnotatorType] = {} + + +def register_annotator(op: str): + def decorator(annotator: AnnotatorType): + OP_TO_ANNOTATOR[op] = annotator + + return decorator + + +class OperatorConfig(NamedTuple): + # fix List[str] with List[List[Union[nn.Module, FunctionType, BuiltinFunctionType]]] + # Basically we are mapping a quantization config to some list of patterns. + # a pattern is defined as a list of nn module, function or builtin function names + # e.g. [nn.Conv2d, torch.relu, torch.add] + # We have not resolved whether fusion can be considered internal details of the + # quantizer hence it does not need communication to user. + # Note this pattern is not really informative since it does not really + # tell us the graph structure resulting from the list of ops. + config: QuantizationConfig + operators: List[OperatorPatternType] + + +def _is_annotated(nodes: List[Node]): + """ + Given a list of nodes (that represents an operator pattern), + check if any of the node is annotated, return True if any of the node + is annotated, otherwise return False + """ + annotated = False + for node in nodes: + annotated = annotated or ( + "quantization_annotation" in node.meta + and node.meta["quantization_annotation"]._annotated + ) + return annotated + + +def _mark_nodes_as_annotated(nodes: List[Node]): + for node in nodes: + if node is not None: + if "quantization_annotation" not in node.meta: + node.meta["quantization_annotation"] = QuantizationAnnotation() + node.meta["quantization_annotation"]._annotated = True + + +def get_input_act_qspec(quantization_config: Optional[QuantizationConfig]): + if quantization_config is None: + return None + if quantization_config.input_activation is None: + return None + quantization_spec: QuantizationSpec = quantization_config.input_activation + assert quantization_spec.qscheme in [ + torch.per_tensor_affine, + torch.per_tensor_symmetric, + ] + return quantization_spec + + +def get_output_act_qspec(quantization_config: Optional[QuantizationConfig]): + if quantization_config is None: + return None + if quantization_config.output_activation is None: + return None + quantization_spec: QuantizationSpec = quantization_config.output_activation + assert quantization_spec.qscheme in [ + torch.per_tensor_affine, + torch.per_tensor_symmetric, + ] + return quantization_spec + + +def get_weight_qspec(quantization_config: Optional[QuantizationConfig]): + if quantization_config is None: + return None + assert quantization_config is not None + if quantization_config.weight is None: + return None + quantization_spec: QuantizationSpec = quantization_config.weight + if quantization_spec.qscheme not in [ + torch.per_tensor_symmetric, + torch.per_channel_symmetric, + ]: + raise ValueError( + f"Unsupported quantization_spec {quantization_spec} for weight" + ) + return quantization_spec + + +def get_bias_qspec(quantization_config: Optional[QuantizationConfig]): + if quantization_config is None: + return None + assert quantization_config is not None + if quantization_config.bias is None: + return None + quantization_spec: QuantizationSpec = quantization_config.bias + assert ( + quantization_spec.dtype == torch.float + ), "Only float dtype for bias is supported for bias right now" + return quantization_spec + + +@register_annotator("linear") +def _annotate_linear( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + annotated_partitions = [] + input_act_qspec = get_input_act_qspec(quantization_config) + output_act_qspec = get_output_act_qspec(quantization_config) + weight_qspec = get_weight_qspec(quantization_config) + bias_qspec = get_bias_qspec(quantization_config) + for node in gm.graph.nodes: + if node.op != "call_function" or node.target != torch.ops.aten.linear.default: + continue + if filter_fn and not filter_fn(node): + continue + act_node = node.args[0] + weight_node = node.args[1] + bias_node = None + if len(node.args) > 2: + bias_node = node.args[2] + + if _is_annotated([node]) is False: # type: ignore[list-item] + _annotate_input_qspec_map( + node, + act_node, + input_act_qspec, + ) + _annotate_input_qspec_map( + node, + weight_node, + weight_qspec, + ) + nodes_to_mark_annotated = [node, weight_node] + if bias_node: + _annotate_input_qspec_map( + node, + bias_node, + bias_qspec, + ) + nodes_to_mark_annotated.append(bias_node) + _annotate_output_qspec(node, output_act_qspec) + _mark_nodes_as_annotated(nodes_to_mark_annotated) + annotated_partitions.append(nodes_to_mark_annotated) + + return annotated_partitions + + +@register_annotator("conv") +def _annotate_conv( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + annotated_partitions = [] + for n in gm.graph.nodes: + if n.op != "call_function" or n.target not in [ + torch.ops.aten.conv1d.default, + torch.ops.aten.conv2d.default, + ]: + continue + conv_node = n + + input_qspec_map = {} + input_act = conv_node.args[0] + assert isinstance(input_act, Node) + input_qspec_map[input_act] = get_input_act_qspec(quantization_config) + + weight = conv_node.args[1] + assert isinstance(weight, Node) + input_qspec_map[weight] = get_weight_qspec(quantization_config) + + # adding weight node to the partition as well + partition = [conv_node, conv_node.args[1]] + + bias = conv_node.args[2] if len(conv_node.args) > 2 else None + if isinstance(bias, Node): + input_qspec_map[bias] = get_bias_qspec(quantization_config) + partition.append(bias) + + if _is_annotated(partition): + continue + + if filter_fn and any(not filter_fn(n) for n in partition): + continue + + conv_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, + output_qspec=get_output_act_qspec(quantization_config), + _annotated=True, + ) + _mark_nodes_as_annotated(partition) + annotated_partitions.append(partition) + return annotated_partitions + + +@register_annotator("conv_relu") +def _annotate_conv_relu( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + annotated_partitions = [] + for n in gm.graph.nodes: + if n.op != "call_function" or n.target not in [ + torch.ops.aten.relu.default, + torch.ops.aten.relu_.default, + ]: + continue + relu_node = n + maybe_conv_node = n.args[0] + if ( + not isinstance(maybe_conv_node, Node) + or maybe_conv_node.op != "call_function" + or maybe_conv_node.target + not in [ + torch.ops.aten.conv1d.default, + torch.ops.aten.conv2d.default, + ] + ): + continue + conv_node = maybe_conv_node + + input_qspec_map = {} + input_act = conv_node.args[0] + assert isinstance(input_act, Node) + input_qspec_map[input_act] = get_input_act_qspec(quantization_config) + + weight = conv_node.args[1] + assert isinstance(weight, Node) + input_qspec_map[weight] = get_weight_qspec(quantization_config) + + # adding weight node to the partition as well + partition = [relu_node, conv_node, conv_node.args[1]] + bias = conv_node.args[2] if len(conv_node.args) > 2 else None + if isinstance(bias, Node): + input_qspec_map[bias] = get_bias_qspec(quantization_config) + partition.append(bias) + + if _is_annotated(partition): + continue + + if filter_fn and any(not filter_fn(n) for n in partition): + continue + + conv_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, _annotated=True + ) + relu_node.meta["quantization_annotation"] = QuantizationAnnotation( + output_qspec=get_output_act_qspec(quantization_config), # type: ignore[arg-type] + _annotated=True, + ) + _mark_nodes_as_annotated(partition) + annotated_partitions.append(partition) + return annotated_partitions + + +@register_annotator("conv_bn") +def _annotate_conv_bn( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + """ + Find conv + batchnorm parititions + Note: This is only used for QAT. In PTQ, batchnorm should already be fused into the conv. + """ + return _do_annotate_conv_bn(gm, quantization_config, filter_fn, has_relu=False) + + +@register_annotator("conv_bn_relu") +def _annotate_conv_bn_relu( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + """ + Find conv + batchnorm + relu parititions + Note: This is only used for QAT. In PTQ, batchnorm should already be fused into the conv. + """ + return _do_annotate_conv_bn(gm, quantization_config, filter_fn, has_relu=True) + + +def _do_annotate_conv_bn( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]], + has_relu: bool, +) -> List[List[Node]]: + """ + Given a function that takes in a `conv_fn` and returns a conv-bn[-relu] pattern, + return a list of annotated partitions. + + The output of the pattern must include a dictionary from string name to node + for the following names: "input", "conv", "weight", "bias", and "output". + """ + + def get_pattern(conv_fn: Callable, relu_is_inplace: bool): + def _conv_bn(x, conv_weight, conv_bias, bn_weight, bn_bias, bn_rm, bn_rv): + conv = conv_fn(x, conv_weight, conv_bias) + bn = F.batch_norm(conv, bn_rm, bn_rv, bn_weight, bn_bias, training=True) + if has_relu: + output = F.relu_(bn) if relu_is_inplace else F.relu(bn) + else: + output = bn + return output, { + "input": x, + "conv": conv, + "weight": conv_weight, + "bias": conv_bias, + "output": output, + } + + return _conv_bn + + # Needed for matching, otherwise the matches gets filtered out due to unused + # nodes returned by batch norm + gm.graph.eliminate_dead_code() + gm.recompile() + + matches = [] + combinations = [ + (F.conv1d, _conv1d_bn_example_inputs), + (F.conv2d, _conv2d_bn_example_inputs), + ] + + # Add `is_cuda` and `relu_is_inplace` dimensions + combinations = itertools.product( + combinations, + [True, False] if torch.cuda.is_available() else [False], # is_cuda + [True, False] if has_relu else [False], # relu_is_inplace + ) + + # Match against all conv dimensions and cuda variants + for (conv_fn, example_inputs), is_cuda, relu_is_inplace in combinations: + pattern = get_pattern(conv_fn, relu_is_inplace) + pattern = get_aten_graph_module(pattern, example_inputs, is_cuda) + pattern.graph.eliminate_dead_code() + pattern.recompile() + matcher = SubgraphMatcherWithNameNodeMap(pattern, ignore_literals=True) + matches.extend(matcher.match(gm.graph)) + + # Annotate nodes returned in the matches + annotated_partitions = [] + for match in matches: + name_node_map = match.name_node_map + input_node = name_node_map["input"] + conv_node = name_node_map["conv"] + weight_node = name_node_map["weight"] + bias_node = name_node_map["bias"] + output_node = name_node_map["output"] + + # TODO: annotate the uses of input, weight, and bias separately instead + # of assuming they come from a single conv node. This is not possible today + # because input may have multiple users, and we can't rely on the conv node + # always being the first user. This was the case in models with skip + # connections like resnet18 + + # Validate conv args + if conv_node.args[0] is not input_node: + raise ValueError("Conv arg did not contain input node ", input_node) + if conv_node.args[1] is not weight_node: + raise ValueError("Conv arg did not contain weight node ", weight_node) + if len(conv_node.args) > 2 and conv_node.args[2] is not bias_node: + raise ValueError("Conv arg did not contain bias node ", bias_node) + + # Skip if the partition is already annotated or is filtered out by the user + partition = [conv_node, weight_node] + if bias_node is not None: + partition.append(bias_node) + if _is_annotated(partition): + continue + if filter_fn and any(not filter_fn(n) for n in partition): + continue + + # Annotate conv inputs and pattern output + input_qspec_map = {} + input_qspec_map[input_node] = get_input_act_qspec(quantization_config) + input_qspec_map[weight_node] = get_weight_qspec(quantization_config) + if bias_node is not None: + input_qspec_map[bias_node] = get_bias_qspec(quantization_config) + conv_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + ) + output_node.meta["quantization_annotation"] = QuantizationAnnotation( + output_qspec=get_output_act_qspec(quantization_config), # type: ignore[arg-type] + _annotated=True, + ) + _mark_nodes_as_annotated(partition) + annotated_partitions.append(partition) + return annotated_partitions + + +@register_annotator("gru_io_only") +def _annotate_gru_io_only( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + gru_partitions = get_source_partitions(gm.graph, [torch.nn.GRU], filter_fn) + gru_partitions = list(itertools.chain(*gru_partitions.values())) + annotated_partitions = [] + for gru_partition in gru_partitions: + annotated_partitions.append(gru_partition.nodes) + output_nodes = gru_partition.output_nodes + input_nodes = gru_partition.input_nodes + # skip annotation if it is already annotated + if _is_annotated(input_nodes + output_nodes): + continue + # inside each GRU partition, we should be able to annotate each linear + # subgraph + input_qspec_map: Dict[Node, QuantizationSpecBase] = {} + input_act = input_nodes[0] + input_act_user = next(iter(input_act.users.keys())) + assert isinstance(input_act, Node) + assert isinstance(input_act_user, Node) + input_act_user.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map={ + input_act: get_input_act_qspec(quantization_config), + }, + _annotated=True, + ) + + hidden_state = input_nodes[1] + hidden_state_user = next(iter(hidden_state.users.keys())) + assert isinstance(hidden_state, Node) + assert isinstance(hidden_state_user, Node) + hidden_state_user.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map={ + hidden_state: get_input_act_qspec(quantization_config), + }, + _annotated=True, + ) + + assert len(output_nodes) == 2, "expecting GRU to have two outputs" + for output in output_nodes: + output.meta["quantization_annotation"] = QuantizationAnnotation( + output_qspec=get_output_act_qspec(quantization_config), + _annotated=True, + ) + nodes_to_mark_annotated = list(gru_partition.nodes) + _mark_nodes_as_annotated(nodes_to_mark_annotated) + return annotated_partitions + + +@register_annotator("max_pool2d") +def _annotate_max_pool2d( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + module_partitions = get_source_partitions( + gm.graph, [torch.nn.MaxPool2d, torch.nn.functional.max_pool2d], filter_fn + ) + maxpool_partitions = list(itertools.chain(*module_partitions.values())) + annotated_partitions = [] + for maxpool_partition in maxpool_partitions: + annotated_partitions.append(maxpool_partition.nodes) + output_node = maxpool_partition.output_nodes[0] + maxpool_node = None + for n in maxpool_partition.nodes: + if n.target == torch.ops.aten.max_pool2d.default: + maxpool_node = n + assert ( + maxpool_node is not None + ), "XNNPACKQuantizer only works with torch.ops.aten.max_pool2d.default, " + "please make sure you are exporting the model correctly" + if _is_annotated([output_node, maxpool_node]): # type: ignore[list-item] + continue + + input_act = maxpool_node.args[0] # type: ignore[union-attr] + assert isinstance(input_act, Node) + + # only annotate maxpool when the output of the input node is annotated + if ( + "quantization_annotation" not in input_act.meta + or not input_act.meta["quantization_annotation"]._annotated + or input_act.meta["quantization_annotation"].output_qspec is None + ): + continue + # input and output of maxpool will share quantization parameter with input of maxpool + act_qspec = SharedQuantizationSpec(input_act) + # act_qspec = get_act_qspec(quantization_config) + maxpool_node.meta["quantization_annotation"] = QuantizationAnnotation( # type: ignore[union-attr] + input_qspec_map={ + input_act: act_qspec, + }, + _annotated=True, + ) + output_node.meta["quantization_annotation"] = QuantizationAnnotation( + output_qspec=act_qspec, + _annotated=True, + ) + return annotated_partitions + + +@register_annotator("adaptive_avg_pool2d") +def _annotate_adaptive_avg_pool2d( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + """Always annotate adaptive_avg_pool2d op""" + module_partitions = get_source_partitions( + gm.graph, [torch.nn.AdaptiveAvgPool2d, F.adaptive_avg_pool2d], filter_fn + ) + partitions = list(itertools.chain(*module_partitions.values())) + annotated_partitions = [] + for partition in partitions: + pool_node = partition.output_nodes[0] + if ( + pool_node.op != "call_function" + or pool_node.target != torch.ops.aten.adaptive_avg_pool2d.default + ): + raise ValueError(f"{pool_node} is not an aten adaptive_avg_pool2d operator") + + if _is_annotated([pool_node]): + continue + + annotated_partitions.append(partition.nodes) + input_act = pool_node.args[0] + assert isinstance(input_act, Node) + + # only annotate input output sharing operator + # when the output of the input node is annotated + if ( + "quantization_annotation" not in input_act.meta + or not input_act.meta["quantization_annotation"]._annotated + or input_act.meta["quantization_annotation"].output_qspec is None + ): + input_act_qspec = get_input_act_qspec(quantization_config) + else: + input_act_qspec = SharedQuantizationSpec(input_act) + + # output sharing with input + output_act_qspec = SharedQuantizationSpec((input_act, pool_node)) + pool_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map={ + input_act: input_act_qspec, + }, + output_qspec=output_act_qspec, + _annotated=True, + ) + return annotated_partitions + + +def _is_input_large_scalar(node: Node, gm: torch.fx.GraphModule): + """Check if input is a large scalar value. So that we can skip quantization for the node + since histc op (in HistogramObserver) only works for values up to certain upper bound + """ + if node.op == "get_attr": + tensor = getattr(gm, node.target) # type: ignore[arg-type] + # torch.histc works until this upper bound + HISTC_UPPER_BOUND = 3.4028235e15 + return tensor.numel() == 1 and abs(tensor.item()) > HISTC_UPPER_BOUND + return False + + +@register_annotator("add_relu") +def _annotate_add_relu( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + fused_partitions = find_sequential_partitions( + gm, [torch.add, torch.nn.ReLU], filter_fn + ) + annotated_partitions = [] + for fused_partition in fused_partitions: + add_partition, relu_partition = fused_partition + annotated_partitions.append(add_partition.nodes + relu_partition.nodes) + if len(relu_partition.output_nodes) > 1: + raise ValueError("Relu partition has more than one output node") + relu_node = relu_partition.output_nodes[0] + if len(add_partition.output_nodes) > 1: + raise ValueError("add partition has more than one output node") + add_node = add_partition.output_nodes[0] + + if _is_annotated([relu_node, add_node]): + continue + + input_act_qspec = get_input_act_qspec(quantization_config) + output_act_qspec = get_output_act_qspec(quantization_config) + + input_qspec_map = {} + input_act0 = add_node.args[0] + if isinstance(input_act0, Node): + if _is_input_large_scalar(input_act0, gm): + continue + input_qspec_map[input_act0] = input_act_qspec + + input_act1 = add_node.args[1] + if isinstance(input_act1, Node): + if _is_input_large_scalar(input_act1, gm): + continue + input_qspec_map[input_act1] = input_act_qspec + + add_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + ) + relu_node.meta["quantization_annotation"] = QuantizationAnnotation( + output_qspec=output_act_qspec, + _annotated=True, + ) + return annotated_partitions + + +@register_annotator("add") +def _annotate_add( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + add_partitions = get_source_partitions( + gm.graph, [operator.add, torch.add, operator.iadd], filter_fn + ) + add_partitions = list(itertools.chain(*add_partitions.values())) + annotated_partitions = [] + for add_partition in add_partitions: + annotated_partitions.append(add_partition.nodes) + add_node = add_partition.output_nodes[0] + if _is_annotated([add_node]): + continue + + input_act_qspec = get_input_act_qspec(quantization_config) + output_act_qspec = get_output_act_qspec(quantization_config) + + input_qspec_map = {} + input_act0 = add_node.args[0] + if isinstance(input_act0, Node): + if _is_input_large_scalar(input_act0, gm): + continue + input_qspec_map[input_act0] = input_act_qspec + + input_act1 = add_node.args[1] + if isinstance(input_act1, Node): + if _is_input_large_scalar(input_act1, gm): + continue + input_qspec_map[input_act1] = input_act_qspec + + add_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, + output_qspec=output_act_qspec, + _annotated=True, + ) + return annotated_partitions + + +@register_annotator("mul_relu") +def _annotate_mul_relu( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + fused_partitions = find_sequential_partitions( + gm, [torch.mul, torch.nn.ReLU], filter_fn + ) + annotated_partitions = [] + for fused_partition in fused_partitions: + mul_partition, relu_partition = fused_partition + annotated_partitions.append(mul_partition.nodes + relu_partition.nodes) + if len(relu_partition.output_nodes) > 1: + raise ValueError("Relu partition has more than one output node") + relu_node = relu_partition.output_nodes[0] + if len(mul_partition.output_nodes) > 1: + raise ValueError("mul partition has more than one output node") + mul_node = mul_partition.output_nodes[0] + + if _is_annotated([relu_node, mul_node]): + continue + + input_act_qspec = get_input_act_qspec(quantization_config) + output_act_qspec = get_output_act_qspec(quantization_config) + + input_qspec_map = {} + input_act0 = mul_node.args[0] + if isinstance(input_act0, Node): + if _is_input_large_scalar(input_act0, gm): + continue + input_qspec_map[input_act0] = input_act_qspec + + input_act1 = mul_node.args[1] + if isinstance(input_act1, Node): + if _is_input_large_scalar(input_act1, gm): + continue + input_qspec_map[input_act1] = input_act_qspec + + mul_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, + _annotated=True, + ) + relu_node.meta["quantization_annotation"] = QuantizationAnnotation( + output_qspec=output_act_qspec, + _annotated=True, + ) + return annotated_partitions + + +@register_annotator("mul") +def _annotate_mul( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + mul_partitions = get_source_partitions( + gm.graph, ["mul", "mul_", operator.mul, torch.mul, operator.imul], filter_fn + ) + mul_partitions = list(itertools.chain(*mul_partitions.values())) + annotated_partitions = [] + for mul_partition in mul_partitions: + annotated_partitions.append(mul_partition.nodes) + mul_node = mul_partition.output_nodes[0] + if _is_annotated([mul_node]): + continue + + input_act_qspec = get_input_act_qspec(quantization_config) + output_act_qspec = get_output_act_qspec(quantization_config) + + input_qspec_map = {} + input_act0 = mul_node.args[0] + if isinstance(input_act0, Node): + if _is_input_large_scalar(input_act0, gm): + continue + input_qspec_map[input_act0] = input_act_qspec + + input_act1 = mul_node.args[1] + if isinstance(input_act1, Node): + if _is_input_large_scalar(input_act1, gm): + continue + input_qspec_map[input_act1] = input_act_qspec + + mul_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, + output_qspec=output_act_qspec, + _annotated=True, + ) + return annotated_partitions + + +# TODO: remove Optional in return type, fix annotated_partitions logic +@register_annotator("cat") +def _annotate_cat( + gm: torch.fx.GraphModule, + quantization_config: Optional[QuantizationConfig], + filter_fn: Optional[Callable[[Node], bool]] = None, +) -> Optional[List[List[Node]]]: + cat_partitions = get_source_partitions(gm.graph, [torch.cat], filter_fn) + cat_partitions = list(itertools.chain(*cat_partitions.values())) + annotated_partitions = [] + for cat_partition in cat_partitions: + cat_node = cat_partition.output_nodes[0] + if _is_annotated([cat_node]): + continue + + if cat_node.target != torch.ops.aten.cat.default: + # TODO: change this to AnnotationException + raise Exception( + f"Expected cat node: torch.ops.aten.cat.default, but found {cat_node.target}" + " please check if you are calling the correct capture API" + ) + + annotated_partitions.append(cat_partition.nodes) + + input_act_qspec = get_input_act_qspec(quantization_config) + inputs = cat_node.args[0] + + input_qspec_map = {} + input_act0 = inputs[0] + if isinstance(input_act0, Node): + input_qspec_map[input_act0] = input_act_qspec + + shared_with_input0_qspec = SharedQuantizationSpec((input_act0, cat_node)) + for input_act in inputs[1:]: + input_qspec_map[input_act] = shared_with_input0_qspec + + output_act_qspec = shared_with_input0_qspec + + cat_node.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map=input_qspec_map, + output_qspec=output_act_qspec, + _annotated=True, + ) + return annotated_partitions + + +def _is_share_obs_or_fq_op(op: Callable) -> bool: + return op in [ + torch.ops.aten.hardtanh.default, + torch.ops.aten.hardtanh_.default, + torch.ops.aten.mean.default, + torch.ops.aten.mean.dim, + torch.ops.aten.permute.default, + torch.ops.aten.permute_copy.default, + torch.ops.aten.squeeze.dim, + torch.ops.aten.squeeze_copy.dim, + # TODO: remove? + torch.ops.aten.adaptive_avg_pool2d.default, + torch.ops.aten.view_copy.default, + torch.ops.aten.view.default, + torch.ops.aten.slice_copy.Tensor, + torch.ops.aten.flatten.using_ints, + ] + + +def propagate_annotation(model: torch.fx.GraphModule) -> None: + for n in model.graph.nodes: + if n.op != "call_function" or not _is_share_obs_or_fq_op(n.target): + continue + + prev_node = n.args[0] + if not isinstance(prev_node, Node): + continue + + quantization_annotation = prev_node.meta.get("quantization_annotation", None) + if not quantization_annotation: + continue + + output_qspec = quantization_annotation.output_qspec + if not output_qspec: + continue + + # make sure current node is not annotated + if ( + "quantization_annotation" in n.meta + and n.meta["quantization_annotation"]._annotated + ): + continue + + shared_qspec = SharedQuantizationSpec(prev_node) + # propagate the previous output_qspec to the current node + n.meta["quantization_annotation"] = QuantizationAnnotation( + input_qspec_map={ + prev_node: shared_qspec, + }, + output_qspec=shared_qspec, + _annotated=True, + ) + + +# TODO: make the list of ops customizable +def _convert_scalars_to_attrs(model: torch.fx.GraphModule) -> torch.fx.GraphModule: + for n in model.graph.nodes: + if n.op != "call_function" or n.target not in [ + torch.ops.aten.add.Tensor, + torch.ops.aten.mul.Tensor, + ]: + continue + args = list(n.args) + new_args = [] + for i in range(len(args)): + if isinstance(args[i], torch.fx.Node): + new_args.append(args[i]) + continue + prefix = "_tensor_constant_" + get_new_attr_name = get_new_attr_name_with_prefix(prefix) + tensor_constant_name = get_new_attr_name(model) + model.register_buffer(tensor_constant_name, torch.tensor(float(args[i]))) + with model.graph.inserting_before(n): + get_attr_node = model.graph.create_node( + "get_attr", tensor_constant_name, (), {} + ) + new_args.append(get_attr_node) + n.args = tuple(new_args) + model.recompile() + return model diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_compute.pyx b/vllm/lib/python3.10/site-packages/pyarrow/_compute.pyx new file mode 100644 index 0000000000000000000000000000000000000000..d39120934d5fd582c14b9ea478585118659bff6b --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_compute.pyx @@ -0,0 +1,3274 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: language_level = 3 + +import sys + +from cpython.object cimport Py_LT, Py_EQ, Py_GT, Py_LE, Py_NE, Py_GE +from cython.operator cimport dereference as deref + +from collections import namedtuple + +from pyarrow.lib import frombytes, tobytes, ArrowInvalid +from pyarrow.lib cimport * +from pyarrow.includes.common cimport * +from pyarrow.includes.libarrow cimport * +import pyarrow.lib as lib +from pyarrow.util import _DEPR_MSG +from libcpp cimport bool as c_bool + +import inspect +try: + import numpy as np +except ImportError: + np = None +import warnings + + +__pas = None +_substrait_msg = ( + "The pyarrow installation is not built with support for Substrait." +) + + +SUPPORTED_INPUT_ARR_TYPES = (list, tuple) +if np is not None: + SUPPORTED_INPUT_ARR_TYPES += (np.ndarray, ) + + +def _pas(): + global __pas + if __pas is None: + try: + import pyarrow.substrait as pas + __pas = pas + except ImportError: + raise ImportError(_substrait_msg) + return __pas + + +def _forbid_instantiation(klass, subclasses_instead=True): + msg = '{} is an abstract class thus cannot be initialized.'.format( + klass.__name__ + ) + if subclasses_instead: + subclasses = [cls.__name__ for cls in klass.__subclasses__] + msg += ' Use one of the subclasses instead: {}'.format( + ', '.join(subclasses) + ) + raise TypeError(msg) + + +cdef wrap_scalar_function(const shared_ptr[CFunction]& sp_func): + """ + Wrap a C++ scalar Function in a ScalarFunction object. + """ + cdef ScalarFunction func = ScalarFunction.__new__(ScalarFunction) + func.init(sp_func) + return func + + +cdef wrap_vector_function(const shared_ptr[CFunction]& sp_func): + """ + Wrap a C++ vector Function in a VectorFunction object. + """ + cdef VectorFunction func = VectorFunction.__new__(VectorFunction) + func.init(sp_func) + return func + + +cdef wrap_scalar_aggregate_function(const shared_ptr[CFunction]& sp_func): + """ + Wrap a C++ aggregate Function in a ScalarAggregateFunction object. + """ + cdef ScalarAggregateFunction func = \ + ScalarAggregateFunction.__new__(ScalarAggregateFunction) + func.init(sp_func) + return func + + +cdef wrap_hash_aggregate_function(const shared_ptr[CFunction]& sp_func): + """ + Wrap a C++ aggregate Function in a HashAggregateFunction object. + """ + cdef HashAggregateFunction func = \ + HashAggregateFunction.__new__(HashAggregateFunction) + func.init(sp_func) + return func + + +cdef wrap_meta_function(const shared_ptr[CFunction]& sp_func): + """ + Wrap a C++ meta Function in a MetaFunction object. + """ + cdef MetaFunction func = MetaFunction.__new__(MetaFunction) + func.init(sp_func) + return func + + +cdef wrap_function(const shared_ptr[CFunction]& sp_func): + """ + Wrap a C++ Function in a Function object. + + This dispatches to specialized wrappers depending on the function kind. + """ + if sp_func.get() == NULL: + raise ValueError("Function was NULL") + + cdef FunctionKind c_kind = sp_func.get().kind() + if c_kind == FunctionKind_SCALAR: + return wrap_scalar_function(sp_func) + elif c_kind == FunctionKind_VECTOR: + return wrap_vector_function(sp_func) + elif c_kind == FunctionKind_SCALAR_AGGREGATE: + return wrap_scalar_aggregate_function(sp_func) + elif c_kind == FunctionKind_HASH_AGGREGATE: + return wrap_hash_aggregate_function(sp_func) + elif c_kind == FunctionKind_META: + return wrap_meta_function(sp_func) + else: + raise NotImplementedError("Unknown Function::Kind") + + +cdef wrap_scalar_kernel(const CScalarKernel* c_kernel): + if c_kernel == NULL: + raise ValueError("Kernel was NULL") + cdef ScalarKernel kernel = ScalarKernel.__new__(ScalarKernel) + kernel.init(c_kernel) + return kernel + + +cdef wrap_vector_kernel(const CVectorKernel* c_kernel): + if c_kernel == NULL: + raise ValueError("Kernel was NULL") + cdef VectorKernel kernel = VectorKernel.__new__(VectorKernel) + kernel.init(c_kernel) + return kernel + + +cdef wrap_scalar_aggregate_kernel(const CScalarAggregateKernel* c_kernel): + if c_kernel == NULL: + raise ValueError("Kernel was NULL") + cdef ScalarAggregateKernel kernel = \ + ScalarAggregateKernel.__new__(ScalarAggregateKernel) + kernel.init(c_kernel) + return kernel + + +cdef wrap_hash_aggregate_kernel(const CHashAggregateKernel* c_kernel): + if c_kernel == NULL: + raise ValueError("Kernel was NULL") + cdef HashAggregateKernel kernel = \ + HashAggregateKernel.__new__(HashAggregateKernel) + kernel.init(c_kernel) + return kernel + + +cdef class Kernel(_Weakrefable): + """ + A kernel object. + + Kernels handle the execution of a Function for a certain signature. + """ + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly" + .format(self.__class__.__name__)) + + +cdef class ScalarKernel(Kernel): + cdef const CScalarKernel* kernel + + cdef void init(self, const CScalarKernel* kernel) except *: + self.kernel = kernel + + def __repr__(self): + return ("ScalarKernel<{}>" + .format(frombytes(self.kernel.signature.get().ToString()))) + + +cdef class VectorKernel(Kernel): + cdef const CVectorKernel* kernel + + cdef void init(self, const CVectorKernel* kernel) except *: + self.kernel = kernel + + def __repr__(self): + return ("VectorKernel<{}>" + .format(frombytes(self.kernel.signature.get().ToString()))) + + +cdef class ScalarAggregateKernel(Kernel): + cdef const CScalarAggregateKernel* kernel + + cdef void init(self, const CScalarAggregateKernel* kernel) except *: + self.kernel = kernel + + def __repr__(self): + return ("ScalarAggregateKernel<{}>" + .format(frombytes(self.kernel.signature.get().ToString()))) + + +cdef class HashAggregateKernel(Kernel): + cdef const CHashAggregateKernel* kernel + + cdef void init(self, const CHashAggregateKernel* kernel) except *: + self.kernel = kernel + + def __repr__(self): + return ("HashAggregateKernel<{}>" + .format(frombytes(self.kernel.signature.get().ToString()))) + + +FunctionDoc = namedtuple( + "FunctionDoc", + ("summary", "description", "arg_names", "options_class", + "options_required")) + + +cdef class Function(_Weakrefable): + """ + A compute function. + + A function implements a certain logical computation over a range of + possible input signatures. Each signature accepts a range of input + types and is implemented by a given Kernel. + + Functions can be of different kinds: + + * "scalar" functions apply an item-wise computation over all items + of their inputs. Each item in the output only depends on the values + of the inputs at the same position. Examples: addition, comparisons, + string predicates... + + * "vector" functions apply a collection-wise computation, such that + each item in the output may depend on the values of several items + in each input. Examples: dictionary encoding, sorting, extracting + unique values... + + * "scalar_aggregate" functions reduce the dimensionality of the inputs by + applying a reduction function. Examples: sum, min_max, mode... + + * "hash_aggregate" functions apply a reduction function to an input + subdivided by grouping criteria. They may not be directly called. + Examples: hash_sum, hash_min_max... + + * "meta" functions dispatch to other functions. + """ + + cdef: + shared_ptr[CFunction] sp_func + CFunction* base_func + + _kind_map = { + FunctionKind_SCALAR: "scalar", + FunctionKind_VECTOR: "vector", + FunctionKind_SCALAR_AGGREGATE: "scalar_aggregate", + FunctionKind_HASH_AGGREGATE: "hash_aggregate", + FunctionKind_META: "meta", + } + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly" + .format(self.__class__.__name__)) + + cdef void init(self, const shared_ptr[CFunction]& sp_func) except *: + self.sp_func = sp_func + self.base_func = sp_func.get() + + def __repr__(self): + return ("arrow.compute.Function" + .format(self.name, self.kind, self.arity, self.num_kernels)) + + def __reduce__(self): + # Reduction uses the global registry + return get_function, (self.name,) + + @property + def name(self): + """ + The function name. + """ + return frombytes(self.base_func.name()) + + @property + def arity(self): + """ + The function arity. + + If Ellipsis (i.e. `...`) is returned, the function takes a variable + number of arguments. + """ + cdef CArity arity = self.base_func.arity() + if arity.is_varargs: + return ... + else: + return arity.num_args + + @property + def kind(self): + """ + The function kind. + """ + cdef FunctionKind c_kind = self.base_func.kind() + try: + return self._kind_map[c_kind] + except KeyError: + raise NotImplementedError("Unknown Function::Kind") + + @property + def _doc(self): + """ + The C++-like function documentation (for internal use). + """ + cdef CFunctionDoc c_doc = self.base_func.doc() + return FunctionDoc(frombytes(c_doc.summary), + frombytes(c_doc.description), + [frombytes(s) for s in c_doc.arg_names], + frombytes(c_doc.options_class), + c_doc.options_required) + + @property + def num_kernels(self): + """ + The number of kernels implementing this function. + """ + return self.base_func.num_kernels() + + def call(self, args, FunctionOptions options=None, + MemoryPool memory_pool=None, length=None): + """ + Call the function on the given arguments. + + Parameters + ---------- + args : iterable + The arguments to pass to the function. Accepted types depend + on the specific function. + options : FunctionOptions, optional + Options instance for executing this function. This should have + the right concrete options type. + memory_pool : pyarrow.MemoryPool, optional + If not passed, will allocate memory from the default memory pool. + length : int, optional + Batch size for execution, for nullary (no argument) functions. If + not passed, will be inferred from passed data. + """ + cdef: + const CFunctionOptions* c_options = NULL + CMemoryPool* pool = maybe_unbox_memory_pool(memory_pool) + CExecContext c_exec_ctx = CExecContext(pool) + CExecBatch c_batch + CDatum result + + _pack_compute_args(args, &c_batch.values) + + if options is not None: + c_options = options.get_options() + + if length is not None: + c_batch.length = length + with nogil: + result = GetResultValue( + self.base_func.Execute(c_batch, c_options, &c_exec_ctx) + ) + else: + with nogil: + result = GetResultValue( + self.base_func.Execute(c_batch.values, c_options, + &c_exec_ctx) + ) + + return wrap_datum(result) + + +cdef class ScalarFunction(Function): + cdef const CScalarFunction* func + + cdef void init(self, const shared_ptr[CFunction]& sp_func) except *: + Function.init(self, sp_func) + self.func = sp_func.get() + + @property + def kernels(self): + """ + The kernels implementing this function. + """ + cdef vector[const CScalarKernel*] kernels = self.func.kernels() + return [wrap_scalar_kernel(k) for k in kernels] + + +cdef class VectorFunction(Function): + cdef const CVectorFunction* func + + cdef void init(self, const shared_ptr[CFunction]& sp_func) except *: + Function.init(self, sp_func) + self.func = sp_func.get() + + @property + def kernels(self): + """ + The kernels implementing this function. + """ + cdef vector[const CVectorKernel*] kernels = self.func.kernels() + return [wrap_vector_kernel(k) for k in kernels] + + +cdef class ScalarAggregateFunction(Function): + cdef const CScalarAggregateFunction* func + + cdef void init(self, const shared_ptr[CFunction]& sp_func) except *: + Function.init(self, sp_func) + self.func = sp_func.get() + + @property + def kernels(self): + """ + The kernels implementing this function. + """ + cdef vector[const CScalarAggregateKernel*] kernels = \ + self.func.kernels() + return [wrap_scalar_aggregate_kernel(k) for k in kernels] + + +cdef class HashAggregateFunction(Function): + cdef const CHashAggregateFunction* func + + cdef void init(self, const shared_ptr[CFunction]& sp_func) except *: + Function.init(self, sp_func) + self.func = sp_func.get() + + @property + def kernels(self): + """ + The kernels implementing this function. + """ + cdef vector[const CHashAggregateKernel*] kernels = self.func.kernels() + return [wrap_hash_aggregate_kernel(k) for k in kernels] + + +cdef class MetaFunction(Function): + cdef const CMetaFunction* func + + cdef void init(self, const shared_ptr[CFunction]& sp_func) except *: + Function.init(self, sp_func) + self.func = sp_func.get() + + # Since num_kernels is exposed, also expose a kernels property + @property + def kernels(self): + """ + The kernels implementing this function. + """ + return [] + + +cdef _pack_compute_args(object values, vector[CDatum]* out): + for val in values: + if isinstance(val, SUPPORTED_INPUT_ARR_TYPES): + val = lib.asarray(val) + + if isinstance(val, Array): + out.push_back(CDatum(( val).sp_array)) + continue + elif isinstance(val, ChunkedArray): + out.push_back(CDatum(( val).sp_chunked_array)) + continue + elif isinstance(val, Scalar): + out.push_back(CDatum(( val).unwrap())) + continue + elif isinstance(val, RecordBatch): + out.push_back(CDatum(( val).sp_batch)) + continue + elif isinstance(val, Table): + out.push_back(CDatum(( val).sp_table)) + continue + else: + # Is it a Python scalar? + try: + scal = lib.scalar(val) + except Exception: + # Raise dedicated error below + pass + else: + out.push_back(CDatum(( scal).unwrap())) + continue + + raise TypeError(f"Got unexpected argument type {type(val)} " + "for compute function") + + +cdef class FunctionRegistry(_Weakrefable): + cdef CFunctionRegistry* registry + + def __init__(self): + self.registry = GetFunctionRegistry() + + def list_functions(self): + """ + Return all function names in the registry. + """ + cdef vector[c_string] names = self.registry.GetFunctionNames() + return [frombytes(name) for name in names] + + def get_function(self, name): + """ + Look up a function by name in the registry. + + Parameters + ---------- + name : str + The name of the function to lookup + """ + cdef: + c_string c_name = tobytes(name) + shared_ptr[CFunction] func + with nogil: + func = GetResultValue(self.registry.GetFunction(c_name)) + return wrap_function(func) + + +cdef FunctionRegistry _global_func_registry = FunctionRegistry() + + +def function_registry(): + return _global_func_registry + + +def get_function(name): + """ + Get a function by name. + + The function is looked up in the global registry + (as returned by `function_registry()`). + + Parameters + ---------- + name : str + The name of the function to lookup + """ + return _global_func_registry.get_function(name) + + +def list_functions(): + """ + Return all function names in the global registry. + """ + return _global_func_registry.list_functions() + + +def call_function(name, args, options=None, memory_pool=None, length=None): + """ + Call a named function. + + The function is looked up in the global registry + (as returned by `function_registry()`). + + Parameters + ---------- + name : str + The name of the function to call. + args : list + The arguments to the function. + options : optional + options provided to the function. + memory_pool : MemoryPool, optional + memory pool to use for allocations during function execution. + length : int, optional + Batch size for execution, for nullary (no argument) functions. If not + passed, inferred from data. + """ + func = _global_func_registry.get_function(name) + return func.call(args, options=options, memory_pool=memory_pool, + length=length) + + +cdef class FunctionOptions(_Weakrefable): + __slots__ = () # avoid mistakingly creating attributes + + cdef const CFunctionOptions* get_options(self) except NULL: + return self.wrapped.get() + + cdef void init(self, const shared_ptr[CFunctionOptions]& sp): + self.wrapped = sp + + cdef inline shared_ptr[CFunctionOptions] unwrap(self): + return self.wrapped + + def serialize(self): + cdef: + CResult[shared_ptr[CBuffer]] res = self.get_options().Serialize() + shared_ptr[CBuffer] c_buf = GetResultValue(res) + return pyarrow_wrap_buffer(c_buf) + + @staticmethod + def deserialize(buf): + """ + Deserialize options for a function. + + Parameters + ---------- + buf : Buffer + The buffer containing the data to deserialize. + """ + cdef: + shared_ptr[CBuffer] c_buf = pyarrow_unwrap_buffer(buf) + CResult[unique_ptr[CFunctionOptions]] maybe_options = \ + DeserializeFunctionOptions(deref(c_buf)) + shared_ptr[CFunctionOptions] c_options + c_options = to_shared(GetResultValue(move(maybe_options))) + type_name = frombytes(c_options.get().options_type().type_name()) + module = globals() + if type_name not in module: + raise ValueError(f'Cannot deserialize "{type_name}"') + klass = module[type_name] + options = klass.__new__(klass) + ( options).init(c_options) + return options + + def __repr__(self): + type_name = self.__class__.__name__ + # Remove {} so we can use our own braces + string_repr = frombytes(self.get_options().ToString())[1:-1] + return f"{type_name}({string_repr})" + + def __eq__(self, FunctionOptions other): + return self.get_options().Equals(deref(other.get_options())) + + +def _raise_invalid_function_option(value, description, *, + exception_class=ValueError): + raise exception_class(f"\"{value}\" is not a valid {description}") + + +# NOTE: +# To properly expose the constructor signature of FunctionOptions +# subclasses, we use a two-level inheritance: +# 1. a C extension class that implements option validation and setting +# (won't expose function signatures because of +# https://github.com/cython/cython/issues/3873) +# 2. a Python derived class that implements the constructor + +cdef class _CastOptions(FunctionOptions): + cdef CCastOptions* options + + cdef void init(self, const shared_ptr[CFunctionOptions]& sp): + FunctionOptions.init(self, sp) + self.options = self.wrapped.get() + + def _set_options(self, DataType target_type, allow_int_overflow, + allow_time_truncate, allow_time_overflow, + allow_decimal_truncate, allow_float_truncate, + allow_invalid_utf8): + cdef: + shared_ptr[CCastOptions] wrapped = make_shared[CCastOptions]() + self.init( wrapped) + self._set_type(target_type) + if allow_int_overflow is not None: + self.allow_int_overflow = allow_int_overflow + if allow_time_truncate is not None: + self.allow_time_truncate = allow_time_truncate + if allow_time_overflow is not None: + self.allow_time_overflow = allow_time_overflow + if allow_decimal_truncate is not None: + self.allow_decimal_truncate = allow_decimal_truncate + if allow_float_truncate is not None: + self.allow_float_truncate = allow_float_truncate + if allow_invalid_utf8 is not None: + self.allow_invalid_utf8 = allow_invalid_utf8 + + def _set_type(self, target_type=None): + if target_type is not None: + deref(self.options).to_type = \ + ( ensure_type(target_type)).sp_type + + def _set_safe(self): + self.init(shared_ptr[CFunctionOptions]( + new CCastOptions(CCastOptions.Safe()))) + + def _set_unsafe(self): + self.init(shared_ptr[CFunctionOptions]( + new CCastOptions(CCastOptions.Unsafe()))) + + def is_safe(self): + return not (deref(self.options).allow_int_overflow or + deref(self.options).allow_time_truncate or + deref(self.options).allow_time_overflow or + deref(self.options).allow_decimal_truncate or + deref(self.options).allow_float_truncate or + deref(self.options).allow_invalid_utf8) + + @property + def allow_int_overflow(self): + return deref(self.options).allow_int_overflow + + @allow_int_overflow.setter + def allow_int_overflow(self, c_bool flag): + deref(self.options).allow_int_overflow = flag + + @property + def allow_time_truncate(self): + return deref(self.options).allow_time_truncate + + @allow_time_truncate.setter + def allow_time_truncate(self, c_bool flag): + deref(self.options).allow_time_truncate = flag + + @property + def allow_time_overflow(self): + return deref(self.options).allow_time_overflow + + @allow_time_overflow.setter + def allow_time_overflow(self, c_bool flag): + deref(self.options).allow_time_overflow = flag + + @property + def allow_decimal_truncate(self): + return deref(self.options).allow_decimal_truncate + + @allow_decimal_truncate.setter + def allow_decimal_truncate(self, c_bool flag): + deref(self.options).allow_decimal_truncate = flag + + @property + def allow_float_truncate(self): + return deref(self.options).allow_float_truncate + + @allow_float_truncate.setter + def allow_float_truncate(self, c_bool flag): + deref(self.options).allow_float_truncate = flag + + @property + def allow_invalid_utf8(self): + return deref(self.options).allow_invalid_utf8 + + @allow_invalid_utf8.setter + def allow_invalid_utf8(self, c_bool flag): + deref(self.options).allow_invalid_utf8 = flag + + +class CastOptions(_CastOptions): + """ + Options for the `cast` function. + + Parameters + ---------- + target_type : DataType, optional + The PyArrow type to cast to. + allow_int_overflow : bool, default False + Whether integer overflow is allowed when casting. + allow_time_truncate : bool, default False + Whether time precision truncation is allowed when casting. + allow_time_overflow : bool, default False + Whether date/time range overflow is allowed when casting. + allow_decimal_truncate : bool, default False + Whether decimal precision truncation is allowed when casting. + allow_float_truncate : bool, default False + Whether floating-point precision truncation is allowed when casting. + allow_invalid_utf8 : bool, default False + Whether producing invalid utf8 data is allowed when casting. + """ + + def __init__(self, target_type=None, *, allow_int_overflow=None, + allow_time_truncate=None, allow_time_overflow=None, + allow_decimal_truncate=None, allow_float_truncate=None, + allow_invalid_utf8=None): + self._set_options(target_type, allow_int_overflow, allow_time_truncate, + allow_time_overflow, allow_decimal_truncate, + allow_float_truncate, allow_invalid_utf8) + + @staticmethod + def safe(target_type=None): + """" + Create a CastOptions for a safe cast. + + Parameters + ---------- + target_type : optional + Target cast type for the safe cast. + """ + self = CastOptions() + self._set_safe() + self._set_type(target_type) + return self + + @staticmethod + def unsafe(target_type=None): + """" + Create a CastOptions for an unsafe cast. + + Parameters + ---------- + target_type : optional + Target cast type for the unsafe cast. + """ + self = CastOptions() + self._set_unsafe() + self._set_type(target_type) + return self + + +def _skip_nulls_doc(): + # (note the weird indent because of how the string is inserted + # by callers) + return """skip_nulls : bool, default True + Whether to skip (ignore) nulls in the input. + If False, any null in the input forces the output to null. +""" + + +def _min_count_doc(*, default): + return f"""min_count : int, default {default} + Minimum number of non-null values in the input. If the number + of non-null values is below `min_count`, the output is null. +""" + + +cdef class _ElementWiseAggregateOptions(FunctionOptions): + def _set_options(self, skip_nulls): + self.wrapped.reset(new CElementWiseAggregateOptions(skip_nulls)) + + +class ElementWiseAggregateOptions(_ElementWiseAggregateOptions): + __doc__ = f""" + Options for element-wise aggregate functions. + + Parameters + ---------- + {_skip_nulls_doc()} + """ + + def __init__(self, *, skip_nulls=True): + self._set_options(skip_nulls) + + +cdef CRoundMode unwrap_round_mode(round_mode) except *: + if round_mode == "down": + return CRoundMode_DOWN + elif round_mode == "up": + return CRoundMode_UP + elif round_mode == "towards_zero": + return CRoundMode_TOWARDS_ZERO + elif round_mode == "towards_infinity": + return CRoundMode_TOWARDS_INFINITY + elif round_mode == "half_down": + return CRoundMode_HALF_DOWN + elif round_mode == "half_up": + return CRoundMode_HALF_UP + elif round_mode == "half_towards_zero": + return CRoundMode_HALF_TOWARDS_ZERO + elif round_mode == "half_towards_infinity": + return CRoundMode_HALF_TOWARDS_INFINITY + elif round_mode == "half_to_even": + return CRoundMode_HALF_TO_EVEN + elif round_mode == "half_to_odd": + return CRoundMode_HALF_TO_ODD + _raise_invalid_function_option(round_mode, "round mode") + + +cdef class _RoundOptions(FunctionOptions): + def _set_options(self, ndigits, round_mode): + self.wrapped.reset( + new CRoundOptions(ndigits, unwrap_round_mode(round_mode)) + ) + + +class RoundOptions(_RoundOptions): + """ + Options for rounding numbers. + + Parameters + ---------- + ndigits : int, default 0 + Number of fractional digits to round to. + round_mode : str, default "half_to_even" + Rounding and tie-breaking mode. + Accepted values are "down", "up", "towards_zero", "towards_infinity", + "half_down", "half_up", "half_towards_zero", "half_towards_infinity", + "half_to_even", "half_to_odd". + """ + + def __init__(self, ndigits=0, round_mode="half_to_even"): + self._set_options(ndigits, round_mode) + + +cdef class _RoundBinaryOptions(FunctionOptions): + def _set_options(self, round_mode): + self.wrapped.reset( + new CRoundBinaryOptions(unwrap_round_mode(round_mode)) + ) + + +class RoundBinaryOptions(_RoundBinaryOptions): + """ + Options for rounding numbers when ndigits is provided by a second array + + Parameters + ---------- + round_mode : str, default "half_to_even" + Rounding and tie-breaking mode. + Accepted values are "down", "up", "towards_zero", "towards_infinity", + "half_down", "half_up", "half_towards_zero", "half_towards_infinity", + "half_to_even", "half_to_odd". + """ + + def __init__(self, round_mode="half_to_even"): + self._set_options(round_mode) + + +cdef CCalendarUnit unwrap_round_temporal_unit(unit) except *: + if unit == "nanosecond": + return CCalendarUnit_NANOSECOND + elif unit == "microsecond": + return CCalendarUnit_MICROSECOND + elif unit == "millisecond": + return CCalendarUnit_MILLISECOND + elif unit == "second": + return CCalendarUnit_SECOND + elif unit == "minute": + return CCalendarUnit_MINUTE + elif unit == "hour": + return CCalendarUnit_HOUR + elif unit == "day": + return CCalendarUnit_DAY + elif unit == "week": + return CCalendarUnit_WEEK + elif unit == "month": + return CCalendarUnit_MONTH + elif unit == "quarter": + return CCalendarUnit_QUARTER + elif unit == "year": + return CCalendarUnit_YEAR + _raise_invalid_function_option(unit, "Calendar unit") + + +cdef class _RoundTemporalOptions(FunctionOptions): + def _set_options(self, multiple, unit, week_starts_monday, + ceil_is_strictly_greater, calendar_based_origin): + self.wrapped.reset( + new CRoundTemporalOptions( + multiple, unwrap_round_temporal_unit(unit), + week_starts_monday, ceil_is_strictly_greater, + calendar_based_origin) + ) + + +class RoundTemporalOptions(_RoundTemporalOptions): + """ + Options for rounding temporal values. + + Parameters + ---------- + multiple : int, default 1 + Number of units to round to. + unit : str, default "day" + The unit in which `multiple` is expressed. + Accepted values are "year", "quarter", "month", "week", "day", + "hour", "minute", "second", "millisecond", "microsecond", + "nanosecond". + week_starts_monday : bool, default True + If True, weeks start on Monday; if False, on Sunday. + ceil_is_strictly_greater : bool, default False + If True, ceil returns a rounded value that is strictly greater than the + input. For example: ceiling 1970-01-01T00:00:00 to 3 hours would + yield 1970-01-01T03:00:00 if set to True and 1970-01-01T00:00:00 + if set to False. + This applies to the ceil_temporal function only. + calendar_based_origin : bool, default False + By default, the origin is 1970-01-01T00:00:00. By setting this to True, + rounding origin will be beginning of one less precise calendar unit. + E.g.: rounding to hours will use beginning of day as origin. + + By default time is rounded to a multiple of units since + 1970-01-01T00:00:00. By setting calendar_based_origin to true, + time will be rounded to number of units since the last greater + calendar unit. + For example: rounding to multiple of days since the beginning of the + month or to hours since the beginning of the day. + Exceptions: week and quarter are not used as greater units, + therefore days will be rounded to the beginning of the month not + week. Greater unit of week is a year. + Note that ceiling and rounding might change sorting order of an array + near greater unit change. For example rounding YYYY-mm-dd 23:00:00 to + 5 hours will ceil and round to YYYY-mm-dd+1 01:00:00 and floor to + YYYY-mm-dd 20:00:00. On the other hand YYYY-mm-dd+1 00:00:00 will + ceil, round and floor to YYYY-mm-dd+1 00:00:00. This can break the + order of an already ordered array. + + """ + + def __init__(self, multiple=1, unit="day", *, week_starts_monday=True, + ceil_is_strictly_greater=False, + calendar_based_origin=False): + self._set_options(multiple, unit, week_starts_monday, + ceil_is_strictly_greater, + calendar_based_origin) + + +cdef class _RoundToMultipleOptions(FunctionOptions): + def _set_options(self, multiple, round_mode): + if not isinstance(multiple, Scalar): + try: + multiple = lib.scalar(multiple) + except Exception: + _raise_invalid_function_option( + multiple, "multiple type for RoundToMultipleOptions", + exception_class=TypeError) + + self.wrapped.reset( + new CRoundToMultipleOptions( + pyarrow_unwrap_scalar(multiple), unwrap_round_mode(round_mode)) + ) + + +class RoundToMultipleOptions(_RoundToMultipleOptions): + """ + Options for rounding numbers to a multiple. + + Parameters + ---------- + multiple : numeric scalar, default 1.0 + Multiple to round to. Should be a scalar of a type compatible + with the argument to be rounded. + round_mode : str, default "half_to_even" + Rounding and tie-breaking mode. + Accepted values are "down", "up", "towards_zero", "towards_infinity", + "half_down", "half_up", "half_towards_zero", "half_towards_infinity", + "half_to_even", "half_to_odd". + """ + + def __init__(self, multiple=1.0, round_mode="half_to_even"): + self._set_options(multiple, round_mode) + + +cdef class _JoinOptions(FunctionOptions): + _null_handling_map = { + "emit_null": CJoinNullHandlingBehavior_EMIT_NULL, + "skip": CJoinNullHandlingBehavior_SKIP, + "replace": CJoinNullHandlingBehavior_REPLACE, + } + + def _set_options(self, null_handling, null_replacement): + try: + self.wrapped.reset( + new CJoinOptions(self._null_handling_map[null_handling], + tobytes(null_replacement)) + ) + except KeyError: + _raise_invalid_function_option(null_handling, "null handling") + + +class JoinOptions(_JoinOptions): + """ + Options for the `binary_join_element_wise` function. + + Parameters + ---------- + null_handling : str, default "emit_null" + How to handle null values in the inputs. + Accepted values are "emit_null", "skip", "replace". + null_replacement : str, default "" + Replacement string to emit for null inputs if `null_handling` + is "replace". + """ + + def __init__(self, null_handling="emit_null", null_replacement=""): + self._set_options(null_handling, null_replacement) + + +cdef class _MatchSubstringOptions(FunctionOptions): + def _set_options(self, pattern, ignore_case): + self.wrapped.reset( + new CMatchSubstringOptions(tobytes(pattern), ignore_case) + ) + + +class MatchSubstringOptions(_MatchSubstringOptions): + """ + Options for looking for a substring. + + Parameters + ---------- + pattern : str + Substring pattern to look for inside input values. + ignore_case : bool, default False + Whether to perform a case-insensitive match. + """ + + def __init__(self, pattern, *, ignore_case=False): + self._set_options(pattern, ignore_case) + + +cdef class _PadOptions(FunctionOptions): + def _set_options(self, width, padding, lean_left_on_odd_padding): + self.wrapped.reset(new CPadOptions(width, tobytes(padding), lean_left_on_odd_padding)) + + +class PadOptions(_PadOptions): + """ + Options for padding strings. + + Parameters + ---------- + width : int + Desired string length. + padding : str, default " " + What to pad the string with. Should be one byte or codepoint. + lean_left_on_odd_padding : bool, default True + What to do if there is an odd number of padding characters (in case + of centered padding). Defaults to aligning on the left (i.e. adding + the extra padding character on the right). + """ + + def __init__(self, width, padding=' ', lean_left_on_odd_padding=True): + self._set_options(width, padding, lean_left_on_odd_padding) + + +cdef class _TrimOptions(FunctionOptions): + def _set_options(self, characters): + self.wrapped.reset(new CTrimOptions(tobytes(characters))) + + +class TrimOptions(_TrimOptions): + """ + Options for trimming characters from strings. + + Parameters + ---------- + characters : str + Individual characters to be trimmed from the string. + """ + + def __init__(self, characters): + self._set_options(tobytes(characters)) + + +cdef class _ReplaceSubstringOptions(FunctionOptions): + def _set_options(self, pattern, replacement, max_replacements): + self.wrapped.reset( + new CReplaceSubstringOptions(tobytes(pattern), + tobytes(replacement), + max_replacements) + ) + + +class ReplaceSubstringOptions(_ReplaceSubstringOptions): + """ + Options for replacing matched substrings. + + Parameters + ---------- + pattern : str + Substring pattern to look for inside input values. + replacement : str + What to replace the pattern with. + max_replacements : int or None, default None + The maximum number of strings to replace in each + input value (unlimited if None). + """ + + def __init__(self, pattern, replacement, *, max_replacements=None): + if max_replacements is None: + max_replacements = -1 + self._set_options(pattern, replacement, max_replacements) + + +cdef class _ExtractRegexOptions(FunctionOptions): + def _set_options(self, pattern): + self.wrapped.reset(new CExtractRegexOptions(tobytes(pattern))) + + +class ExtractRegexOptions(_ExtractRegexOptions): + """ + Options for the `extract_regex` function. + + Parameters + ---------- + pattern : str + Regular expression with named capture fields. + """ + + def __init__(self, pattern): + self._set_options(pattern) + + +cdef class _SliceOptions(FunctionOptions): + def _set_options(self, start, stop, step): + self.wrapped.reset(new CSliceOptions(start, stop, step)) + + +class SliceOptions(_SliceOptions): + """ + Options for slicing. + + Parameters + ---------- + start : int + Index to start slicing at (inclusive). + stop : int or None, default None + If given, index to stop slicing at (exclusive). + If not given, slicing will stop at the end. + step : int, default 1 + Slice step. + """ + + def __init__(self, start, stop=None, step=1): + if stop is None: + stop = sys.maxsize + if step < 0: + stop = -stop + self._set_options(start, stop, step) + + +cdef class _ListSliceOptions(FunctionOptions): + cpdef _set_options(self, start, stop=None, step=1, return_fixed_size_list=None): + cdef: + CListSliceOptions* opts + opts = new CListSliceOptions( + start, + nullopt if stop is None + else (stop), + step, + nullopt if return_fixed_size_list is None + else (return_fixed_size_list) + ) + self.wrapped.reset(opts) + + +class ListSliceOptions(_ListSliceOptions): + """ + Options for list array slicing. + + Parameters + ---------- + start : int + Index to start slicing inner list elements (inclusive). + stop : Optional[int], default None + If given, index to stop slicing at (exclusive). + If not given, slicing will stop at the end. (NotImplemented) + step : int, default 1 + Slice step. + return_fixed_size_list : Optional[bool], default None + Whether to return a FixedSizeListArray. If true _and_ stop is after + a list element's length, nulls will be appended to create the + requested slice size. The default of `None` will return the same + type which was passed in. + """ + + def __init__(self, start, stop=None, step=1, return_fixed_size_list=None): + self._set_options(start, stop, step, return_fixed_size_list) + + +cdef class _ReplaceSliceOptions(FunctionOptions): + def _set_options(self, start, stop, replacement): + self.wrapped.reset( + new CReplaceSliceOptions(start, stop, tobytes(replacement)) + ) + + +class ReplaceSliceOptions(_ReplaceSliceOptions): + """ + Options for replacing slices. + + Parameters + ---------- + start : int + Index to start slicing at (inclusive). + stop : int + Index to stop slicing at (exclusive). + replacement : str + What to replace the slice with. + """ + + def __init__(self, start, stop, replacement): + self._set_options(start, stop, replacement) + + +cdef class _FilterOptions(FunctionOptions): + _null_selection_map = { + "drop": CFilterNullSelectionBehavior_DROP, + "emit_null": CFilterNullSelectionBehavior_EMIT_NULL, + } + + def _set_options(self, null_selection_behavior): + try: + self.wrapped.reset( + new CFilterOptions( + self._null_selection_map[null_selection_behavior] + ) + ) + except KeyError: + _raise_invalid_function_option(null_selection_behavior, + "null selection behavior") + + +class FilterOptions(_FilterOptions): + """ + Options for selecting with a boolean filter. + + Parameters + ---------- + null_selection_behavior : str, default "drop" + How to handle nulls in the selection filter. + Accepted values are "drop", "emit_null". + """ + + def __init__(self, null_selection_behavior="drop"): + self._set_options(null_selection_behavior) + + +cdef class _DictionaryEncodeOptions(FunctionOptions): + _null_encoding_map = { + "encode": CDictionaryEncodeNullEncodingBehavior_ENCODE, + "mask": CDictionaryEncodeNullEncodingBehavior_MASK, + } + + def _set_options(self, null_encoding): + try: + self.wrapped.reset( + new CDictionaryEncodeOptions( + self._null_encoding_map[null_encoding] + ) + ) + except KeyError: + _raise_invalid_function_option(null_encoding, "null encoding") + + +class DictionaryEncodeOptions(_DictionaryEncodeOptions): + """ + Options for dictionary encoding. + + Parameters + ---------- + null_encoding : str, default "mask" + How to encode nulls in the input. + Accepted values are "mask" (null inputs emit a null in the indices + array), "encode" (null inputs emit a non-null index pointing to + a null value in the dictionary array). + """ + + def __init__(self, null_encoding="mask"): + self._set_options(null_encoding) + + +cdef class _RunEndEncodeOptions(FunctionOptions): + def _set_options(self, run_end_type): + run_end_ty = ensure_type(run_end_type) + self.wrapped.reset(new CRunEndEncodeOptions(pyarrow_unwrap_data_type(run_end_ty))) + + +class RunEndEncodeOptions(_RunEndEncodeOptions): + """ + Options for run-end encoding. + + Parameters + ---------- + run_end_type : DataType, default pyarrow.int32() + The data type of the run_ends array. + + Accepted values are pyarrow.{int16(), int32(), int64()}. + """ + + def __init__(self, run_end_type=lib.int32()): + self._set_options(run_end_type) + + +cdef class _TakeOptions(FunctionOptions): + def _set_options(self, boundscheck): + self.wrapped.reset(new CTakeOptions(boundscheck)) + + +class TakeOptions(_TakeOptions): + """ + Options for the `take` and `array_take` functions. + + Parameters + ---------- + boundscheck : boolean, default True + Whether to check indices are within bounds. If False and an + index is out of bounds, behavior is undefined (the process + may crash). + """ + + def __init__(self, *, boundscheck=True): + self._set_options(boundscheck) + + +cdef class _MakeStructOptions(FunctionOptions): + def _set_options(self, field_names, field_nullability, field_metadata): + cdef: + vector[c_string] c_field_names + vector[shared_ptr[const CKeyValueMetadata]] c_field_metadata + for name in field_names: + c_field_names.push_back(tobytes(name)) + for metadata in field_metadata: + c_field_metadata.push_back(pyarrow_unwrap_metadata(metadata)) + self.wrapped.reset( + new CMakeStructOptions(c_field_names, field_nullability, + c_field_metadata) + ) + + +class MakeStructOptions(_MakeStructOptions): + """ + Options for the `make_struct` function. + + Parameters + ---------- + field_names : sequence of str + Names of the struct fields to create. + field_nullability : sequence of bool, optional + Nullability information for each struct field. + If omitted, all fields are nullable. + field_metadata : sequence of KeyValueMetadata, optional + Metadata for each struct field. + """ + + def __init__(self, field_names=(), *, field_nullability=None, + field_metadata=None): + if field_nullability is None: + field_nullability = [True] * len(field_names) + if field_metadata is None: + field_metadata = [None] * len(field_names) + self._set_options(field_names, field_nullability, field_metadata) + + +cdef CFieldRef _ensure_field_ref(value) except *: + cdef: + CFieldRef field_ref + const CFieldRef* field_ref_ptr + + if isinstance(value, (list, tuple)): + value = Expression._nested_field(tuple(value)) + + if isinstance(value, Expression): + field_ref_ptr = (value).unwrap().field_ref() + if field_ref_ptr is NULL: + raise ValueError("Unable to get FieldRef from Expression") + field_ref = deref(field_ref_ptr) + elif isinstance(value, (bytes, str)): + if value.startswith(b'.' if isinstance(value, bytes) else '.'): + field_ref = GetResultValue( + CFieldRef.FromDotPath(tobytes(value))) + else: + field_ref = CFieldRef(tobytes(value)) + elif isinstance(value, int): + field_ref = CFieldRef( value) + else: + raise TypeError("Expected a field reference as a str or int, list of " + f"str or int, or Expression. Got {type(value)} instead.") + return field_ref + + +cdef class _StructFieldOptions(FunctionOptions): + def _set_options(self, indices): + + if isinstance(indices, (list, tuple)) and not len(indices): + # Allow empty indices; effectively return same array + self.wrapped.reset( + new CStructFieldOptions(indices)) + return + + cdef CFieldRef field_ref = _ensure_field_ref(indices) + self.wrapped.reset(new CStructFieldOptions(field_ref)) + + +class StructFieldOptions(_StructFieldOptions): + """ + Options for the `struct_field` function. + + Parameters + ---------- + indices : List[str], List[bytes], List[int], Expression, bytes, str, or int + List of indices for chained field lookup, for example `[4, 1]` + will look up the second nested field in the fifth outer field. + """ + + def __init__(self, indices): + self._set_options(indices) + + +cdef class _ScalarAggregateOptions(FunctionOptions): + def _set_options(self, skip_nulls, min_count): + self.wrapped.reset(new CScalarAggregateOptions(skip_nulls, min_count)) + + +class ScalarAggregateOptions(_ScalarAggregateOptions): + __doc__ = f""" + Options for scalar aggregations. + + Parameters + ---------- + {_skip_nulls_doc()} + {_min_count_doc(default=1)} + """ + + def __init__(self, *, skip_nulls=True, min_count=1): + self._set_options(skip_nulls, min_count) + + +cdef class _CountOptions(FunctionOptions): + _mode_map = { + "only_valid": CCountMode_ONLY_VALID, + "only_null": CCountMode_ONLY_NULL, + "all": CCountMode_ALL, + } + + def _set_options(self, mode): + try: + self.wrapped.reset(new CCountOptions(self._mode_map[mode])) + except KeyError: + _raise_invalid_function_option(mode, "count mode") + + +class CountOptions(_CountOptions): + """ + Options for the `count` function. + + Parameters + ---------- + mode : str, default "only_valid" + Which values to count in the input. + Accepted values are "only_valid", "only_null", "all". + """ + + def __init__(self, mode="only_valid"): + self._set_options(mode) + + +cdef class _IndexOptions(FunctionOptions): + def _set_options(self, scalar): + self.wrapped.reset(new CIndexOptions(pyarrow_unwrap_scalar(scalar))) + + +class IndexOptions(_IndexOptions): + """ + Options for the `index` function. + + Parameters + ---------- + value : Scalar + The value to search for. + """ + + def __init__(self, value): + self._set_options(value) + + +cdef class _MapLookupOptions(FunctionOptions): + _occurrence_map = { + "all": CMapLookupOccurrence_ALL, + "first": CMapLookupOccurrence_FIRST, + "last": CMapLookupOccurrence_LAST, + } + + def _set_options(self, query_key, occurrence): + try: + self.wrapped.reset( + new CMapLookupOptions( + pyarrow_unwrap_scalar(query_key), + self._occurrence_map[occurrence] + ) + ) + except KeyError: + _raise_invalid_function_option(occurrence, + "Should either be first, last, or all") + + +class MapLookupOptions(_MapLookupOptions): + """ + Options for the `map_lookup` function. + + Parameters + ---------- + query_key : Scalar or Object can be converted to Scalar + The key to search for. + occurrence : str + The occurrence(s) to return from the Map + Accepted values are "first", "last", or "all". + """ + + def __init__(self, query_key, occurrence): + if not isinstance(query_key, lib.Scalar): + query_key = lib.scalar(query_key) + + self._set_options(query_key, occurrence) + + +cdef class _ModeOptions(FunctionOptions): + def _set_options(self, n, skip_nulls, min_count): + self.wrapped.reset(new CModeOptions(n, skip_nulls, min_count)) + + +class ModeOptions(_ModeOptions): + __doc__ = f""" + Options for the `mode` function. + + Parameters + ---------- + n : int, default 1 + Number of distinct most-common values to return. + {_skip_nulls_doc()} + {_min_count_doc(default=0)} + """ + + def __init__(self, n=1, *, skip_nulls=True, min_count=0): + self._set_options(n, skip_nulls, min_count) + + +cdef class _SetLookupOptions(FunctionOptions): + def _set_options(self, value_set, c_bool skip_nulls): + cdef unique_ptr[CDatum] valset + if isinstance(value_set, Array): + valset.reset(new CDatum(( value_set).sp_array)) + elif isinstance(value_set, ChunkedArray): + valset.reset( + new CDatum(( value_set).sp_chunked_array) + ) + elif isinstance(value_set, Scalar): + valset.reset(new CDatum(( value_set).unwrap())) + else: + _raise_invalid_function_option(value_set, "value set", + exception_class=TypeError) + + self.wrapped.reset(new CSetLookupOptions(deref(valset), skip_nulls)) + + +class SetLookupOptions(_SetLookupOptions): + """ + Options for the `is_in` and `index_in` functions. + + Parameters + ---------- + value_set : Array + Set of values to look for in the input. + skip_nulls : bool, default False + If False, nulls in the input are matched in the value_set just + like regular values. + If True, nulls in the input always fail matching. + """ + + def __init__(self, value_set, *, skip_nulls=False): + self._set_options(value_set, skip_nulls) + + +cdef class _StrptimeOptions(FunctionOptions): + _unit_map = { + "s": TimeUnit_SECOND, + "ms": TimeUnit_MILLI, + "us": TimeUnit_MICRO, + "ns": TimeUnit_NANO, + } + + def _set_options(self, format, unit, error_is_null): + try: + self.wrapped.reset( + new CStrptimeOptions(tobytes(format), self._unit_map[unit], + error_is_null) + ) + except KeyError: + _raise_invalid_function_option(unit, "time unit") + + +class StrptimeOptions(_StrptimeOptions): + """ + Options for the `strptime` function. + + Parameters + ---------- + format : str + Pattern for parsing input strings as timestamps, such as "%Y/%m/%d". + Note that the semantics of the format follow the C/C++ strptime, not the Python one. + There are differences in behavior, for example how the "%y" placeholder + handles years with less than four digits. + unit : str + Timestamp unit of the output. + Accepted values are "s", "ms", "us", "ns". + error_is_null : boolean, default False + Return null on parsing errors if true or raise if false. + """ + + def __init__(self, format, unit, error_is_null=False): + self._set_options(format, unit, error_is_null) + + +cdef class _StrftimeOptions(FunctionOptions): + def _set_options(self, format, locale): + self.wrapped.reset( + new CStrftimeOptions(tobytes(format), tobytes(locale)) + ) + + +class StrftimeOptions(_StrftimeOptions): + """ + Options for the `strftime` function. + + Parameters + ---------- + format : str, default "%Y-%m-%dT%H:%M:%S" + Pattern for formatting input values. + locale : str, default "C" + Locale to use for locale-specific format specifiers. + """ + + def __init__(self, format="%Y-%m-%dT%H:%M:%S", locale="C"): + self._set_options(format, locale) + + +cdef class _DayOfWeekOptions(FunctionOptions): + def _set_options(self, count_from_zero, week_start): + self.wrapped.reset( + new CDayOfWeekOptions(count_from_zero, week_start) + ) + + +class DayOfWeekOptions(_DayOfWeekOptions): + """ + Options for the `day_of_week` function. + + Parameters + ---------- + count_from_zero : bool, default True + If True, number days from 0, otherwise from 1. + week_start : int, default 1 + Which day does the week start with (Monday=1, Sunday=7). + How this value is numbered is unaffected by `count_from_zero`. + """ + + def __init__(self, *, count_from_zero=True, week_start=1): + self._set_options(count_from_zero, week_start) + + +cdef class _WeekOptions(FunctionOptions): + def _set_options(self, week_starts_monday, count_from_zero, + first_week_is_fully_in_year): + self.wrapped.reset( + new CWeekOptions(week_starts_monday, count_from_zero, + first_week_is_fully_in_year) + ) + + +class WeekOptions(_WeekOptions): + """ + Options for the `week` function. + + Parameters + ---------- + week_starts_monday : bool, default True + If True, weeks start on Monday; if False, on Sunday. + count_from_zero : bool, default False + If True, dates at the start of a year that fall into the last week + of the previous year emit 0. + If False, they emit 52 or 53 (the week number of the last week + of the previous year). + first_week_is_fully_in_year : bool, default False + If True, week number 0 is fully in January. + If False, a week that begins on December 29, 30 or 31 is considered + to be week number 0 of the following year. + """ + + def __init__(self, *, week_starts_monday=True, count_from_zero=False, + first_week_is_fully_in_year=False): + self._set_options(week_starts_monday, + count_from_zero, first_week_is_fully_in_year) + + +cdef class _AssumeTimezoneOptions(FunctionOptions): + _ambiguous_map = { + "raise": CAssumeTimezoneAmbiguous_AMBIGUOUS_RAISE, + "earliest": CAssumeTimezoneAmbiguous_AMBIGUOUS_EARLIEST, + "latest": CAssumeTimezoneAmbiguous_AMBIGUOUS_LATEST, + } + _nonexistent_map = { + "raise": CAssumeTimezoneNonexistent_NONEXISTENT_RAISE, + "earliest": CAssumeTimezoneNonexistent_NONEXISTENT_EARLIEST, + "latest": CAssumeTimezoneNonexistent_NONEXISTENT_LATEST, + } + + def _set_options(self, timezone, ambiguous, nonexistent): + if ambiguous not in self._ambiguous_map: + _raise_invalid_function_option(ambiguous, + "'ambiguous' timestamp handling") + if nonexistent not in self._nonexistent_map: + _raise_invalid_function_option(nonexistent, + "'nonexistent' timestamp handling") + self.wrapped.reset( + new CAssumeTimezoneOptions(tobytes(timezone), + self._ambiguous_map[ambiguous], + self._nonexistent_map[nonexistent]) + ) + + +class AssumeTimezoneOptions(_AssumeTimezoneOptions): + """ + Options for the `assume_timezone` function. + + Parameters + ---------- + timezone : str + Timezone to assume for the input. + ambiguous : str, default "raise" + How to handle timestamps that are ambiguous in the assumed timezone. + Accepted values are "raise", "earliest", "latest". + nonexistent : str, default "raise" + How to handle timestamps that don't exist in the assumed timezone. + Accepted values are "raise", "earliest", "latest". + """ + + def __init__(self, timezone, *, ambiguous="raise", nonexistent="raise"): + self._set_options(timezone, ambiguous, nonexistent) + + +cdef class _NullOptions(FunctionOptions): + def _set_options(self, nan_is_null): + self.wrapped.reset(new CNullOptions(nan_is_null)) + + +class NullOptions(_NullOptions): + """ + Options for the `is_null` function. + + Parameters + ---------- + nan_is_null : bool, default False + Whether floating-point NaN values are considered null. + """ + + def __init__(self, *, nan_is_null=False): + self._set_options(nan_is_null) + + +cdef class _VarianceOptions(FunctionOptions): + def _set_options(self, ddof, skip_nulls, min_count): + self.wrapped.reset(new CVarianceOptions(ddof, skip_nulls, min_count)) + + +class VarianceOptions(_VarianceOptions): + __doc__ = f""" + Options for the `variance` and `stddev` functions. + + Parameters + ---------- + ddof : int, default 0 + Number of degrees of freedom. + {_skip_nulls_doc()} + {_min_count_doc(default=0)} + """ + + def __init__(self, *, ddof=0, skip_nulls=True, min_count=0): + self._set_options(ddof, skip_nulls, min_count) + + +cdef class _SplitOptions(FunctionOptions): + def _set_options(self, max_splits, reverse): + self.wrapped.reset(new CSplitOptions(max_splits, reverse)) + + +class SplitOptions(_SplitOptions): + """ + Options for splitting on whitespace. + + Parameters + ---------- + max_splits : int or None, default None + Maximum number of splits for each input value (unlimited if None). + reverse : bool, default False + Whether to start splitting from the end of each input value. + This only has an effect if `max_splits` is not None. + """ + + def __init__(self, *, max_splits=None, reverse=False): + if max_splits is None: + max_splits = -1 + self._set_options(max_splits, reverse) + + +cdef class _SplitPatternOptions(FunctionOptions): + def _set_options(self, pattern, max_splits, reverse): + self.wrapped.reset( + new CSplitPatternOptions(tobytes(pattern), max_splits, reverse) + ) + + +class SplitPatternOptions(_SplitPatternOptions): + """ + Options for splitting on a string pattern. + + Parameters + ---------- + pattern : str + String pattern to split on. + max_splits : int or None, default None + Maximum number of splits for each input value (unlimited if None). + reverse : bool, default False + Whether to start splitting from the end of each input value. + This only has an effect if `max_splits` is not None. + """ + + def __init__(self, pattern, *, max_splits=None, reverse=False): + if max_splits is None: + max_splits = -1 + self._set_options(pattern, max_splits, reverse) + + +cdef CSortOrder unwrap_sort_order(order) except *: + if order == "ascending": + return CSortOrder_Ascending + elif order == "descending": + return CSortOrder_Descending + _raise_invalid_function_option(order, "sort order") + + +cdef CNullPlacement unwrap_null_placement(null_placement) except *: + if null_placement == "at_start": + return CNullPlacement_AtStart + elif null_placement == "at_end": + return CNullPlacement_AtEnd + _raise_invalid_function_option(null_placement, "null placement") + + +cdef class _PartitionNthOptions(FunctionOptions): + def _set_options(self, pivot, null_placement): + self.wrapped.reset(new CPartitionNthOptions( + pivot, unwrap_null_placement(null_placement))) + + +class PartitionNthOptions(_PartitionNthOptions): + """ + Options for the `partition_nth_indices` function. + + Parameters + ---------- + pivot : int + Index into the equivalent sorted array of the pivot element. + null_placement : str, default "at_end" + Where nulls in the input should be partitioned. + Accepted values are "at_start", "at_end". + """ + + def __init__(self, pivot, *, null_placement="at_end"): + self._set_options(pivot, null_placement) + + +cdef class _CumulativeOptions(FunctionOptions): + def _set_options(self, start, skip_nulls): + if start is None: + self.wrapped.reset(new CCumulativeOptions(skip_nulls)) + elif isinstance(start, Scalar): + self.wrapped.reset(new CCumulativeOptions( + pyarrow_unwrap_scalar(start), skip_nulls)) + else: + try: + start = lib.scalar(start) + self.wrapped.reset(new CCumulativeOptions( + pyarrow_unwrap_scalar(start), skip_nulls)) + except Exception: + _raise_invalid_function_option( + start, "`start` type for CumulativeOptions", TypeError) + + +class CumulativeOptions(_CumulativeOptions): + """ + Options for `cumulative_*` functions. + + - cumulative_sum + - cumulative_sum_checked + - cumulative_prod + - cumulative_prod_checked + - cumulative_max + - cumulative_min + + Parameters + ---------- + start : Scalar, default None + Starting value for the cumulative operation. If none is given, + a default value depending on the operation and input type is used. + skip_nulls : bool, default False + When false, the first encountered null is propagated. + """ + + def __init__(self, start=None, *, skip_nulls=False): + self._set_options(start, skip_nulls) + + +class CumulativeSumOptions(_CumulativeOptions): + """ + Options for `cumulative_sum` function. + + Parameters + ---------- + start : Scalar, default None + Starting value for sum computation + skip_nulls : bool, default False + When false, the first encountered null is propagated. + """ + + def __init__(self, start=None, *, skip_nulls=False): + warnings.warn( + _DEPR_MSG.format("CumulativeSumOptions", "14.0", "CumulativeOptions"), + FutureWarning, + stacklevel=2 + ) + self._set_options(start, skip_nulls) + + +cdef class _PairwiseOptions(FunctionOptions): + def _set_options(self, period): + self.wrapped.reset(new CPairwiseOptions(period)) + + +class PairwiseOptions(_PairwiseOptions): + """ + Options for `pairwise` functions. + + Parameters + ---------- + period : int, default 1 + Period for applying the period function. + """ + + def __init__(self, period=1): + self._set_options(period) + + +cdef class _ListFlattenOptions(FunctionOptions): + def _set_options(self, recursive): + self.wrapped.reset(new CListFlattenOptions(recursive)) + + +class ListFlattenOptions(_ListFlattenOptions): + """ + Options for `list_flatten` function + + Parameters + ---------- + recursive : bool, default False + When True, the list array is flattened recursively until an array + of non-list values is formed. + """ + + def __init__(self, recursive=False): + self._set_options(recursive) + + +cdef class _ArraySortOptions(FunctionOptions): + def _set_options(self, order, null_placement): + self.wrapped.reset(new CArraySortOptions( + unwrap_sort_order(order), unwrap_null_placement(null_placement))) + + +class ArraySortOptions(_ArraySortOptions): + """ + Options for the `array_sort_indices` function. + + Parameters + ---------- + order : str, default "ascending" + Which order to sort values in. + Accepted values are "ascending", "descending". + null_placement : str, default "at_end" + Where nulls in the input should be sorted. + Accepted values are "at_start", "at_end". + """ + + def __init__(self, order="ascending", *, null_placement="at_end"): + self._set_options(order, null_placement) + + +cdef class _SortOptions(FunctionOptions): + def _set_options(self, sort_keys, null_placement): + cdef vector[CSortKey] c_sort_keys + for name, order in sort_keys: + c_sort_keys.push_back( + CSortKey(_ensure_field_ref(name), unwrap_sort_order(order)) + ) + self.wrapped.reset(new CSortOptions( + c_sort_keys, unwrap_null_placement(null_placement))) + + +class SortOptions(_SortOptions): + """ + Options for the `sort_indices` function. + + Parameters + ---------- + sort_keys : sequence of (name, order) tuples + Names of field/column keys to sort the input on, + along with the order each field/column is sorted in. + Accepted values for `order` are "ascending", "descending". + The field name can be a string column name or expression. + null_placement : str, default "at_end" + Where nulls in input should be sorted, only applying to + columns/fields mentioned in `sort_keys`. + Accepted values are "at_start", "at_end". + """ + + def __init__(self, sort_keys=(), *, null_placement="at_end"): + self._set_options(sort_keys, null_placement) + + +cdef class _SelectKOptions(FunctionOptions): + def _set_options(self, k, sort_keys): + cdef vector[CSortKey] c_sort_keys + for name, order in sort_keys: + c_sort_keys.push_back( + CSortKey(_ensure_field_ref(name), unwrap_sort_order(order)) + ) + self.wrapped.reset(new CSelectKOptions(k, c_sort_keys)) + + +class SelectKOptions(_SelectKOptions): + """ + Options for top/bottom k-selection. + + Parameters + ---------- + k : int + Number of leading values to select in sorted order + (i.e. the largest values if sort order is "descending", + the smallest otherwise). + sort_keys : sequence of (name, order) tuples + Names of field/column keys to sort the input on, + along with the order each field/column is sorted in. + Accepted values for `order` are "ascending", "descending". + The field name can be a string column name or expression. + """ + + def __init__(self, k, sort_keys): + self._set_options(k, sort_keys) + + +cdef class _QuantileOptions(FunctionOptions): + _interp_map = { + "linear": CQuantileInterp_LINEAR, + "lower": CQuantileInterp_LOWER, + "higher": CQuantileInterp_HIGHER, + "nearest": CQuantileInterp_NEAREST, + "midpoint": CQuantileInterp_MIDPOINT, + } + + def _set_options(self, quantiles, interp, skip_nulls, min_count): + try: + self.wrapped.reset( + new CQuantileOptions(quantiles, self._interp_map[interp], + skip_nulls, min_count) + ) + except KeyError: + _raise_invalid_function_option(interp, "quantile interpolation") + + +class QuantileOptions(_QuantileOptions): + __doc__ = f""" + Options for the `quantile` function. + + Parameters + ---------- + q : double or sequence of double, default 0.5 + Probability levels of the quantiles to compute. All values must be in + [0, 1]. + interpolation : str, default "linear" + How to break ties between competing data points for a given quantile. + Accepted values are: + + - "linear": compute an interpolation + - "lower": always use the smallest of the two data points + - "higher": always use the largest of the two data points + - "nearest": select the data point that is closest to the quantile + - "midpoint": compute the (unweighted) mean of the two data points + {_skip_nulls_doc()} + {_min_count_doc(default=0)} + """ + + def __init__(self, q=0.5, *, interpolation="linear", skip_nulls=True, + min_count=0): + if not isinstance(q, SUPPORTED_INPUT_ARR_TYPES): + q = [q] + self._set_options(q, interpolation, skip_nulls, min_count) + + +cdef class _TDigestOptions(FunctionOptions): + def _set_options(self, quantiles, delta, buffer_size, skip_nulls, + min_count): + self.wrapped.reset( + new CTDigestOptions(quantiles, delta, buffer_size, skip_nulls, + min_count) + ) + + +class TDigestOptions(_TDigestOptions): + __doc__ = f""" + Options for the `tdigest` function. + + Parameters + ---------- + q : double or sequence of double, default 0.5 + Probability levels of the quantiles to approximate. All values must be + in [0, 1]. + delta : int, default 100 + Compression parameter for the T-digest algorithm. + buffer_size : int, default 500 + Buffer size for the T-digest algorithm. + {_skip_nulls_doc()} + {_min_count_doc(default=0)} + """ + + def __init__(self, q=0.5, *, delta=100, buffer_size=500, skip_nulls=True, + min_count=0): + if not isinstance(q, SUPPORTED_INPUT_ARR_TYPES): + q = [q] + self._set_options(q, delta, buffer_size, skip_nulls, min_count) + + +cdef class _Utf8NormalizeOptions(FunctionOptions): + _form_map = { + "NFC": CUtf8NormalizeForm_NFC, + "NFKC": CUtf8NormalizeForm_NFKC, + "NFD": CUtf8NormalizeForm_NFD, + "NFKD": CUtf8NormalizeForm_NFKD, + } + + def _set_options(self, form): + try: + self.wrapped.reset( + new CUtf8NormalizeOptions(self._form_map[form]) + ) + except KeyError: + _raise_invalid_function_option(form, + "Unicode normalization form") + + +class Utf8NormalizeOptions(_Utf8NormalizeOptions): + """ + Options for the `utf8_normalize` function. + + Parameters + ---------- + form : str + Unicode normalization form. + Accepted values are "NFC", "NFKC", "NFD", NFKD". + """ + + def __init__(self, form): + self._set_options(form) + + +cdef class _RandomOptions(FunctionOptions): + def _set_options(self, initializer): + if initializer == 'system': + self.wrapped.reset(new CRandomOptions( + CRandomOptions.FromSystemRandom())) + return + + if not isinstance(initializer, int): + try: + initializer = hash(initializer) + except TypeError: + raise TypeError( + f"initializer should be 'system', an integer, " + f"or a hashable object; got {initializer!r}") + + if initializer < 0: + initializer += 2**64 + self.wrapped.reset(new CRandomOptions( + CRandomOptions.FromSeed(initializer))) + + +class RandomOptions(_RandomOptions): + """ + Options for random generation. + + Parameters + ---------- + initializer : int or str + How to initialize the underlying random generator. + If an integer is given, it is used as a seed. + If "system" is given, the random generator is initialized with + a system-specific source of (hopefully true) randomness. + Other values are invalid. + """ + + def __init__(self, *, initializer='system'): + self._set_options(initializer) + + +cdef class _RankOptions(FunctionOptions): + + _tiebreaker_map = { + "min": CRankOptionsTiebreaker_Min, + "max": CRankOptionsTiebreaker_Max, + "first": CRankOptionsTiebreaker_First, + "dense": CRankOptionsTiebreaker_Dense, + } + + def _set_options(self, sort_keys, null_placement, tiebreaker): + cdef vector[CSortKey] c_sort_keys + if isinstance(sort_keys, str): + c_sort_keys.push_back( + CSortKey(_ensure_field_ref(""), unwrap_sort_order(sort_keys)) + ) + else: + for name, order in sort_keys: + c_sort_keys.push_back( + CSortKey(_ensure_field_ref(name), unwrap_sort_order(order)) + ) + try: + self.wrapped.reset( + new CRankOptions(c_sort_keys, + unwrap_null_placement(null_placement), + self._tiebreaker_map[tiebreaker]) + ) + except KeyError: + _raise_invalid_function_option(tiebreaker, "tiebreaker") + + +class RankOptions(_RankOptions): + """ + Options for the `rank` function. + + Parameters + ---------- + sort_keys : sequence of (name, order) tuples or str, default "ascending" + Names of field/column keys to sort the input on, + along with the order each field/column is sorted in. + Accepted values for `order` are "ascending", "descending". + The field name can be a string column name or expression. + Alternatively, one can simply pass "ascending" or "descending" as a string + if the input is array-like. + null_placement : str, default "at_end" + Where nulls in input should be sorted. + Accepted values are "at_start", "at_end". + tiebreaker : str, default "first" + Configure how ties between equal values are handled. + Accepted values are: + + - "min": Ties get the smallest possible rank in sorted order. + - "max": Ties get the largest possible rank in sorted order. + - "first": Ranks are assigned in order of when ties appear in the + input. This ensures the ranks are a stable permutation + of the input. + - "dense": The ranks span a dense [1, M] interval where M is the + number of distinct values in the input. + """ + + def __init__(self, sort_keys="ascending", *, null_placement="at_end", tiebreaker="first"): + self._set_options(sort_keys, null_placement, tiebreaker) + + +cdef class Expression(_Weakrefable): + """ + A logical expression to be evaluated against some input. + + To create an expression: + + - Use the factory function ``pyarrow.compute.scalar()`` to create a + scalar (not necessary when combined, see example below). + - Use the factory function ``pyarrow.compute.field()`` to reference + a field (column in table). + - Compare fields and scalars with ``<``, ``<=``, ``==``, ``>=``, ``>``. + - Combine expressions using python operators ``&`` (logical and), + ``|`` (logical or) and ``~`` (logical not). + Note: python keywords ``and``, ``or`` and ``not`` cannot be used + to combine expressions. + - Create expression predicates using Expression methods such as + ``pyarrow.compute.Expression.isin()``. + + Examples + -------- + + >>> import pyarrow.compute as pc + >>> (pc.field("a") < pc.scalar(3)) | (pc.field("b") > 7) + 7))> + >>> pc.field('a') != 3 + + >>> pc.field('a').isin([1, 2, 3]) + + """ + + def __init__(self): + msg = 'Expression is an abstract class thus cannot be initialized.' + raise TypeError(msg) + + cdef void init(self, const CExpression& sp): + self.expr = sp + + @staticmethod + cdef wrap(const CExpression& sp): + cdef Expression self = Expression.__new__(Expression) + self.init(sp) + return self + + cdef inline CExpression unwrap(self): + return self.expr + + def equals(self, Expression other): + """ + Parameters + ---------- + other : pyarrow.dataset.Expression + + Returns + ------- + bool + """ + return self.expr.Equals(other.unwrap()) + + def __str__(self): + return frombytes(self.expr.ToString()) + + def __repr__(self): + return "".format( + self.__class__.__name__, str(self) + ) + + @staticmethod + def from_substrait(object buffer not None): + """ + Deserialize an expression from Substrait + + The serialized message must be an ExtendedExpression message that has + only a single expression. The name of the expression and the schema + the expression was bound to will be ignored. Use + pyarrow.substrait.deserialize_expressions if this information is needed + or if the message might contain multiple expressions. + + Parameters + ---------- + buffer : bytes or Buffer + The Substrait message to deserialize + + Returns + ------- + Expression + The deserialized expression + """ + expressions = _pas().deserialize_expressions(buffer).expressions + if len(expressions) == 0: + raise ValueError("Substrait message did not contain any expressions") + if len(expressions) > 1: + raise ValueError( + "Substrait message contained multiple expressions. Use pyarrow.substrait.deserialize_expressions instead") + return next(iter(expressions.values())) + + def to_substrait(self, Schema schema not None, c_bool allow_arrow_extensions=False): + """ + Serialize the expression using Substrait + + The expression will be serialized as an ExtendedExpression message that has a + single expression named "expression" + + Parameters + ---------- + schema : Schema + The input schema the expression will be bound to + allow_arrow_extensions : bool, default False + If False then only functions that are part of the core Substrait function + definitions will be allowed. Set this to True to allow pyarrow-specific functions + but the result may not be accepted by other compute libraries. + + Returns + ------- + Buffer + A buffer containing the serialized Protobuf plan. + """ + return _pas().serialize_expressions([self], ["expression"], schema, allow_arrow_extensions=allow_arrow_extensions) + + @staticmethod + def _deserialize(Buffer buffer not None): + return Expression.wrap(GetResultValue(CDeserializeExpression( + pyarrow_unwrap_buffer(buffer)))) + + def __reduce__(self): + buffer = pyarrow_wrap_buffer(GetResultValue( + CSerializeExpression(self.expr))) + return Expression._deserialize, (buffer,) + + @staticmethod + cdef Expression _expr_or_scalar(object expr): + if isinstance(expr, Expression): + return ( expr) + return ( Expression._scalar(expr)) + + @staticmethod + def _call(str function_name, list arguments, FunctionOptions options=None): + cdef: + vector[CExpression] c_arguments + shared_ptr[CFunctionOptions] c_options + + for argument in arguments: + if not isinstance(argument, Expression): + # Attempt to help convert this to an expression + try: + argument = Expression._scalar(argument) + except ArrowInvalid: + raise TypeError( + "only other expressions allowed as arguments") + c_arguments.push_back(( argument).expr) + + if options is not None: + c_options = options.unwrap() + + return Expression.wrap(CMakeCallExpression( + tobytes(function_name), move(c_arguments), c_options)) + + def __richcmp__(self, other, int op): + other = Expression._expr_or_scalar(other) + return Expression._call({ + Py_EQ: "equal", + Py_NE: "not_equal", + Py_GT: "greater", + Py_GE: "greater_equal", + Py_LT: "less", + Py_LE: "less_equal", + }[op], [self, other]) + + def __bool__(self): + raise ValueError( + "An Expression cannot be evaluated to python True or False. " + "If you are using the 'and', 'or' or 'not' operators, use '&', " + "'|' or '~' instead." + ) + + def __invert__(self): + return Expression._call("invert", [self]) + + def __and__(Expression self, other): + other = Expression._expr_or_scalar(other) + return Expression._call("and_kleene", [self, other]) + + def __or__(Expression self, other): + other = Expression._expr_or_scalar(other) + return Expression._call("or_kleene", [self, other]) + + def __add__(Expression self, other): + other = Expression._expr_or_scalar(other) + return Expression._call("add_checked", [self, other]) + + def __mul__(Expression self, other): + other = Expression._expr_or_scalar(other) + return Expression._call("multiply_checked", [self, other]) + + def __sub__(Expression self, other): + other = Expression._expr_or_scalar(other) + return Expression._call("subtract_checked", [self, other]) + + def __truediv__(Expression self, other): + other = Expression._expr_or_scalar(other) + return Expression._call("divide_checked", [self, other]) + + def is_valid(self): + """ + Check whether the expression is not-null (valid). + + This creates a new expression equivalent to calling the + `is_valid` compute function on this expression. + + Returns + ------- + is_valid : Expression + """ + return Expression._call("is_valid", [self]) + + def is_null(self, bint nan_is_null=False): + """ + Check whether the expression is null. + + This creates a new expression equivalent to calling the + `is_null` compute function on this expression. + + Parameters + ---------- + nan_is_null : boolean, default False + Whether floating-point NaNs are considered null. + + Returns + ------- + is_null : Expression + """ + options = NullOptions(nan_is_null=nan_is_null) + return Expression._call("is_null", [self], options) + + def is_nan(self): + """ + Check whether the expression is NaN. + + This creates a new expression equivalent to calling the + `is_nan` compute function on this expression. + + Returns + ------- + is_nan : Expression + """ + return Expression._call("is_nan", [self]) + + def cast(self, type=None, safe=None, options=None): + """ + Explicitly set or change the expression's data type. + + This creates a new expression equivalent to calling the + `cast` compute function on this expression. + + Parameters + ---------- + type : DataType, default None + Type to cast array to. + safe : boolean, default True + Whether to check for conversion errors such as overflow. + options : CastOptions, default None + Additional checks pass by CastOptions + + Returns + ------- + cast : Expression + """ + safe_vars_passed = (safe is not None) or (type is not None) + + if safe_vars_passed and (options is not None): + raise ValueError("Must either pass values for 'type' and 'safe' or pass a " + "value for 'options'") + + if options is None: + type = ensure_type(type, allow_none=False) + if safe is False: + options = CastOptions.unsafe(type) + else: + options = CastOptions.safe(type) + return Expression._call("cast", [self], options) + + def isin(self, values): + """ + Check whether the expression is contained in values. + + This creates a new expression equivalent to calling the + `is_in` compute function on this expression. + + Parameters + ---------- + values : Array or iterable + The values to check for. + + Returns + ------- + isin : Expression + A new expression that, when evaluated, checks whether + this expression's value is contained in `values`. + """ + if not isinstance(values, Array): + values = lib.array(values) + + options = SetLookupOptions(values) + return Expression._call("is_in", [self], options) + + @staticmethod + def _field(name_or_idx not None): + cdef: + CFieldRef c_field + + if isinstance(name_or_idx, int): + return Expression.wrap(CMakeFieldExpressionByIndex(name_or_idx)) + else: + c_field = CFieldRef( tobytes(name_or_idx)) + return Expression.wrap(CMakeFieldExpression(c_field)) + + @staticmethod + def _nested_field(tuple names not None): + cdef: + vector[CFieldRef] nested + + if len(names) == 0: + raise ValueError("nested field reference should be non-empty") + nested.reserve(len(names)) + for name in names: + if isinstance(name, int): + nested.push_back(CFieldRef(name)) + else: + nested.push_back(CFieldRef( tobytes(name))) + return Expression.wrap(CMakeFieldExpression(CFieldRef(move(nested)))) + + @staticmethod + def _scalar(value): + cdef: + Scalar scalar + + if isinstance(value, Scalar): + scalar = value + else: + scalar = lib.scalar(value) + + return Expression.wrap(CMakeScalarExpression(scalar.unwrap())) + + +_deserialize = Expression._deserialize +cdef CExpression _true = CMakeScalarExpression( + make_shared[CBooleanScalar](True) +) + + +cdef CExpression _bind(Expression filter, Schema schema) except *: + assert schema is not None + + if filter is None: + return _true + + return GetResultValue(filter.unwrap().Bind( + deref(pyarrow_unwrap_schema(schema).get()))) + + +cdef class UdfContext: + """ + Per-invocation function context/state. + + This object will always be the first argument to a user-defined + function. It should not be used outside of a call to the function. + """ + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly" + .format(self.__class__.__name__)) + + cdef void init(self, const CUdfContext &c_context): + self.c_context = c_context + + @property + def batch_length(self): + """ + The common length of all input arguments (int). + + In the case that all arguments are scalars, this value + is used to pass the "actual length" of the arguments, + e.g. because the scalar values are encoding a column + with a constant value. + """ + return self.c_context.batch_length + + @property + def memory_pool(self): + """ + A memory pool for allocations (:class:`MemoryPool`). + + This is the memory pool supplied by the user when they invoked + the function and it should be used in any calls to arrow that the + UDF makes if that call accepts a memory_pool. + """ + return box_memory_pool(self.c_context.pool) + + +cdef inline CFunctionDoc _make_function_doc(dict func_doc) except *: + """ + Helper function to generate the FunctionDoc + This function accepts a dictionary and expects the + summary(str), description(str) and arg_names(List[str]) keys. + """ + cdef: + CFunctionDoc f_doc + vector[c_string] c_arg_names + + f_doc.summary = tobytes(func_doc["summary"]) + f_doc.description = tobytes(func_doc["description"]) + for arg_name in func_doc["arg_names"]: + c_arg_names.push_back(tobytes(arg_name)) + f_doc.arg_names = c_arg_names + # UDFOptions integration: + # TODO: https://issues.apache.org/jira/browse/ARROW-16041 + f_doc.options_class = b"" + f_doc.options_required = False + return f_doc + + +cdef object box_udf_context(const CUdfContext& c_context): + cdef UdfContext context = UdfContext.__new__(UdfContext) + context.init(c_context) + return context + + +cdef _udf_callback(user_function, const CUdfContext& c_context, inputs): + """ + Helper callback function used to wrap the UdfContext from Python to C++ + execution. + """ + context = box_udf_context(c_context) + return user_function(context, *inputs) + + +def _get_udf_context(memory_pool, batch_length): + cdef CUdfContext c_context + c_context.pool = maybe_unbox_memory_pool(memory_pool) + c_context.batch_length = batch_length + context = box_udf_context(c_context) + return context + + +ctypedef CStatus (*CRegisterUdf)(PyObject* function, function[CallbackUdf] wrapper, + const CUdfOptions& options, CFunctionRegistry* registry) + +cdef class RegisterUdf(_Weakrefable): + cdef CRegisterUdf register_func + + cdef void init(self, const CRegisterUdf register_func): + self.register_func = register_func + + +cdef get_register_scalar_function(): + cdef RegisterUdf reg = RegisterUdf.__new__(RegisterUdf) + reg.register_func = RegisterScalarFunction + return reg + + +cdef get_register_tabular_function(): + cdef RegisterUdf reg = RegisterUdf.__new__(RegisterUdf) + reg.register_func = RegisterTabularFunction + return reg + + +cdef get_register_aggregate_function(): + cdef RegisterUdf reg = RegisterUdf.__new__(RegisterUdf) + reg.register_func = RegisterAggregateFunction + return reg + +cdef get_register_vector_function(): + cdef RegisterUdf reg = RegisterUdf.__new__(RegisterUdf) + reg.register_func = RegisterVectorFunction + return reg + + +def register_scalar_function(func, function_name, function_doc, in_types, out_type, + func_registry=None): + """ + Register a user-defined scalar function. + + This API is EXPERIMENTAL. + + A scalar function is a function that executes elementwise + operations on arrays or scalars, i.e. a scalar function must + be computed row-by-row with no state where each output row + is computed only from its corresponding input row. + In other words, all argument arrays have the same length, + and the output array is of the same length as the arguments. + Scalar functions are the only functions allowed in query engine + expressions. + + Parameters + ---------- + func : callable + A callable implementing the user-defined function. + The first argument is the context argument of type + UdfContext. + Then, it must take arguments equal to the number of + in_types defined. It must return an Array or Scalar + matching the out_type. It must return a Scalar if + all arguments are scalar, else it must return an Array. + + To define a varargs function, pass a callable that takes + *args. The last in_type will be the type of all varargs + arguments. + function_name : str + Name of the function. There should only be one function + registered with this name in the function registry. + function_doc : dict + A dictionary object with keys "summary" (str), + and "description" (str). + in_types : Dict[str, DataType] + A dictionary mapping function argument names to + their respective DataType. + The argument names will be used to generate + documentation for the function. The number of + arguments specified here determines the function + arity. + out_type : DataType + Output type of the function. + func_registry : FunctionRegistry + Optional function registry to use instead of the default global one. + + Examples + -------- + >>> import pyarrow as pa + >>> import pyarrow.compute as pc + >>> + >>> func_doc = {} + >>> func_doc["summary"] = "simple udf" + >>> func_doc["description"] = "add a constant to a scalar" + >>> + >>> def add_constant(ctx, array): + ... return pc.add(array, 1, memory_pool=ctx.memory_pool) + >>> + >>> func_name = "py_add_func" + >>> in_types = {"array": pa.int64()} + >>> out_type = pa.int64() + >>> pc.register_scalar_function(add_constant, func_name, func_doc, + ... in_types, out_type) + >>> + >>> func = pc.get_function(func_name) + >>> func.name + 'py_add_func' + >>> answer = pc.call_function(func_name, [pa.array([20])]) + >>> answer + + [ + 21 + ] + """ + return _register_user_defined_function(get_register_scalar_function(), + func, function_name, function_doc, in_types, + out_type, func_registry) + + +def register_vector_function(func, function_name, function_doc, in_types, out_type, + func_registry=None): + """ + Register a user-defined vector function. + + This API is EXPERIMENTAL. + + A vector function is a function that executes vector + operations on arrays. Vector function is often used + when compute doesn't fit other more specific types of + functions (e.g., scalar and aggregate). + + Parameters + ---------- + func : callable + A callable implementing the user-defined function. + The first argument is the context argument of type + UdfContext. + Then, it must take arguments equal to the number of + in_types defined. It must return an Array or Scalar + matching the out_type. It must return a Scalar if + all arguments are scalar, else it must return an Array. + + To define a varargs function, pass a callable that takes + *args. The last in_type will be the type of all varargs + arguments. + function_name : str + Name of the function. There should only be one function + registered with this name in the function registry. + function_doc : dict + A dictionary object with keys "summary" (str), + and "description" (str). + in_types : Dict[str, DataType] + A dictionary mapping function argument names to + their respective DataType. + The argument names will be used to generate + documentation for the function. The number of + arguments specified here determines the function + arity. + out_type : DataType + Output type of the function. + func_registry : FunctionRegistry + Optional function registry to use instead of the default global one. + + Examples + -------- + >>> import pyarrow as pa + >>> import pyarrow.compute as pc + >>> + >>> func_doc = {} + >>> func_doc["summary"] = "percent rank" + >>> func_doc["description"] = "compute percent rank" + >>> + >>> def list_flatten_udf(ctx, x): + ... return pc.list_flatten(x) + >>> + >>> func_name = "list_flatten_udf" + >>> in_types = {"array": pa.list_(pa.int64())} + >>> out_type = pa.int64() + >>> pc.register_vector_function(list_flatten_udf, func_name, func_doc, + ... in_types, out_type) + >>> + >>> answer = pc.call_function(func_name, [pa.array([[1, 2], [3, 4]])]) + >>> answer + + [ + 1, + 2, + 3, + 4 + ] + """ + return _register_user_defined_function(get_register_vector_function(), + func, function_name, function_doc, in_types, + out_type, func_registry) + + +def register_aggregate_function(func, function_name, function_doc, in_types, out_type, + func_registry=None): + """ + Register a user-defined non-decomposable aggregate function. + + This API is EXPERIMENTAL. + + A non-decomposable aggregation function is a function that executes + aggregate operations on the whole data that it is aggregating. + In other words, non-decomposable aggregate function cannot be + split into consume/merge/finalize steps. + + This is often used with ordered or segmented aggregation where groups + can be emit before accumulating all of the input data. + + Note that currently the size of any input column cannot exceed 2 GB + for a single segment (all groups combined). + + Parameters + ---------- + func : callable + A callable implementing the user-defined function. + The first argument is the context argument of type + UdfContext. + Then, it must take arguments equal to the number of + in_types defined. It must return a Scalar matching the + out_type. + To define a varargs function, pass a callable that takes + *args. The in_type needs to match in type of inputs when + the function gets called. + function_name : str + Name of the function. This name must be unique, i.e., + there should only be one function registered with + this name in the function registry. + function_doc : dict + A dictionary object with keys "summary" (str), + and "description" (str). + in_types : Dict[str, DataType] + A dictionary mapping function argument names to + their respective DataType. + The argument names will be used to generate + documentation for the function. The number of + arguments specified here determines the function + arity. + out_type : DataType + Output type of the function. + func_registry : FunctionRegistry + Optional function registry to use instead of the default global one. + + Examples + -------- + >>> import numpy as np + >>> import pyarrow as pa + >>> import pyarrow.compute as pc + >>> + >>> func_doc = {} + >>> func_doc["summary"] = "simple median udf" + >>> func_doc["description"] = "compute median" + >>> + >>> def compute_median(ctx, array): + ... return pa.scalar(np.median(array)) + >>> + >>> func_name = "py_compute_median" + >>> in_types = {"array": pa.int64()} + >>> out_type = pa.float64() + >>> pc.register_aggregate_function(compute_median, func_name, func_doc, + ... in_types, out_type) + >>> + >>> func = pc.get_function(func_name) + >>> func.name + 'py_compute_median' + >>> answer = pc.call_function(func_name, [pa.array([20, 40])]) + >>> answer + + >>> table = pa.table([pa.array([1, 1, 2, 2]), pa.array([10, 20, 30, 40])], names=['k', 'v']) + >>> result = table.group_by('k').aggregate([('v', 'py_compute_median')]) + >>> result + pyarrow.Table + k: int64 + v_py_compute_median: double + ---- + k: [[1,2]] + v_py_compute_median: [[15,35]] + """ + return _register_user_defined_function(get_register_aggregate_function(), + func, function_name, function_doc, in_types, + out_type, func_registry) + + +def register_tabular_function(func, function_name, function_doc, in_types, out_type, + func_registry=None): + """ + Register a user-defined tabular function. + + This API is EXPERIMENTAL. + + A tabular function is one accepting a context argument of type + UdfContext and returning a generator of struct arrays. + The in_types argument must be empty and the out_type argument + specifies a schema. Each struct array must have field types + corresponding to the schema. + + Parameters + ---------- + func : callable + A callable implementing the user-defined function. + The only argument is the context argument of type + UdfContext. It must return a callable that + returns on each invocation a StructArray matching + the out_type, where an empty array indicates end. + function_name : str + Name of the function. There should only be one function + registered with this name in the function registry. + function_doc : dict + A dictionary object with keys "summary" (str), + and "description" (str). + in_types : Dict[str, DataType] + Must be an empty dictionary (reserved for future use). + out_type : Union[Schema, DataType] + Schema of the function's output, or a corresponding flat struct type. + func_registry : FunctionRegistry + Optional function registry to use instead of the default global one. + """ + cdef: + shared_ptr[CSchema] c_schema + shared_ptr[CDataType] c_type + + if isinstance(out_type, Schema): + c_schema = pyarrow_unwrap_schema(out_type) + with nogil: + c_type = make_shared[CStructType](deref(c_schema).fields()) + out_type = pyarrow_wrap_data_type(c_type) + return _register_user_defined_function(get_register_tabular_function(), + func, function_name, function_doc, in_types, + out_type, func_registry) + + +def _register_user_defined_function(register_func, func, function_name, function_doc, in_types, + out_type, func_registry=None): + """ + Register a user-defined function. + + This method itself doesn't care about the type of the UDF + (i.e., scalar vs tabular vs aggregate) + + Parameters + ---------- + register_func: object + An object holding a CRegisterUdf in a "register_func" attribute. + func : callable + A callable implementing the user-defined function. + function_name : str + Name of the function. There should only be one function + registered with this name in the function registry. + function_doc : dict + A dictionary object with keys "summary" (str), + and "description" (str). + in_types : Dict[str, DataType] + A dictionary mapping function argument names to + their respective DataType. + out_type : DataType + Output type of the function. + func_registry : FunctionRegistry + Optional function registry to use instead of the default global one. + """ + cdef: + CRegisterUdf c_register_func + c_string c_func_name + CArity c_arity + CFunctionDoc c_func_doc + vector[shared_ptr[CDataType]] c_in_types + PyObject* c_function + shared_ptr[CDataType] c_out_type + CUdfOptions c_options + CFunctionRegistry* c_func_registry + + if callable(func): + c_function = func + else: + raise TypeError("func must be a callable") + + c_func_name = tobytes(function_name) + + func_spec = inspect.getfullargspec(func) + num_args = -1 + if isinstance(in_types, dict): + for in_type in in_types.values(): + c_in_types.push_back( + pyarrow_unwrap_data_type(ensure_type(in_type))) + function_doc["arg_names"] = in_types.keys() + num_args = len(in_types) + else: + raise TypeError( + "in_types must be a dictionary of DataType") + + c_arity = CArity( num_args, func_spec.varargs) + + if "summary" not in function_doc: + raise ValueError("Function doc must contain a summary") + + if "description" not in function_doc: + raise ValueError("Function doc must contain a description") + + if "arg_names" not in function_doc: + raise ValueError("Function doc must contain arg_names") + + c_func_doc = _make_function_doc(function_doc) + + c_out_type = pyarrow_unwrap_data_type(ensure_type(out_type)) + + c_options.func_name = c_func_name + c_options.arity = c_arity + c_options.func_doc = c_func_doc + c_options.input_types = c_in_types + c_options.output_type = c_out_type + + if func_registry is None: + c_func_registry = NULL + else: + c_func_registry = (func_registry).registry + + c_register_func = (register_func).register_func + + check_status(c_register_func(c_function, + &_udf_callback, + c_options, c_func_registry)) + + +def call_tabular_function(function_name, args=None, func_registry=None): + """ + Get a record batch iterator from a tabular function. + + Parameters + ---------- + function_name : str + Name of the function. + args : iterable + The arguments to pass to the function. Accepted types depend + on the specific function. Currently, only an empty args is supported. + func_registry : FunctionRegistry + Optional function registry to use instead of the default global one. + """ + cdef: + c_string c_func_name + vector[CDatum] c_args + CFunctionRegistry* c_func_registry + shared_ptr[CRecordBatchReader] c_reader + RecordBatchReader reader + + c_func_name = tobytes(function_name) + if func_registry is None: + c_func_registry = NULL + else: + c_func_registry = (func_registry).registry + if args is None: + args = [] + _pack_compute_args(args, &c_args) + + with nogil: + c_reader = GetResultValue(CallTabularFunction( + c_func_name, c_args, c_func_registry)) + reader = RecordBatchReader.__new__(RecordBatchReader) + reader.reader = c_reader + return RecordBatchReader.from_batches(pyarrow_wrap_schema(deref(c_reader).schema()), reader) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_csv.pxd b/vllm/lib/python3.10/site-packages/pyarrow/_csv.pxd new file mode 100644 index 0000000000000000000000000000000000000000..dcc562a41c795896d12fc7cdd3baebf0122bedc9 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_csv.pxd @@ -0,0 +1,55 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: language_level = 3 + +from pyarrow.includes.libarrow cimport * +from pyarrow.lib cimport _Weakrefable + + +cdef class ConvertOptions(_Weakrefable): + cdef: + unique_ptr[CCSVConvertOptions] options + + @staticmethod + cdef ConvertOptions wrap(CCSVConvertOptions options) + + +cdef class ParseOptions(_Weakrefable): + cdef: + unique_ptr[CCSVParseOptions] options + object _invalid_row_handler + + @staticmethod + cdef ParseOptions wrap(CCSVParseOptions options) + + +cdef class ReadOptions(_Weakrefable): + cdef: + unique_ptr[CCSVReadOptions] options + public object encoding + + @staticmethod + cdef ReadOptions wrap(CCSVReadOptions options) + + +cdef class WriteOptions(_Weakrefable): + cdef: + unique_ptr[CCSVWriteOptions] options + + @staticmethod + cdef WriteOptions wrap(CCSVWriteOptions options) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_csv.pyx b/vllm/lib/python3.10/site-packages/pyarrow/_csv.pyx new file mode 100644 index 0000000000000000000000000000000000000000..508488c0c3b3c3bcd2d2157f57f625b1e5b92c2e --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_csv.pyx @@ -0,0 +1,1542 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: profile=False +# distutils: language = c++ +# cython: language_level = 3 + +from cython.operator cimport dereference as deref + +from collections import namedtuple +from collections.abc import Mapping + +from pyarrow.includes.common cimport * +from pyarrow.includes.libarrow cimport * +from pyarrow.includes.libarrow_python cimport * +from pyarrow.lib cimport (check_status, Field, MemoryPool, Schema, + RecordBatchReader, ensure_type, + maybe_unbox_memory_pool, get_input_stream, + get_writer, native_transcoding_input_stream, + pyarrow_unwrap_batch, pyarrow_unwrap_schema, + pyarrow_unwrap_table, pyarrow_wrap_schema, + pyarrow_wrap_table, pyarrow_wrap_data_type, + pyarrow_unwrap_data_type, Table, RecordBatch, + StopToken, _CRecordBatchWriter) +from pyarrow.lib import frombytes, tobytes, SignalStopHandler + + +cdef unsigned char _single_char(s) except 0: + val = ord(s) + if val == 0 or val > 127: + raise ValueError("Expecting an ASCII character") + return val + + +_InvalidRow = namedtuple( + "_InvalidRow", ("expected_columns", "actual_columns", "number", "text"), + module=__name__) + + +class InvalidRow(_InvalidRow): + """ + Description of an invalid row in a CSV file. + + Parameters + ---------- + expected_columns : int + The expected number of columns in the row. + actual_columns : int + The actual number of columns in the row. + number : int or None + The physical row number if known, otherwise None. + text : str + The contents of the row. + """ + __slots__ = () + + +cdef CInvalidRowResult _handle_invalid_row( + handler, const CCSVInvalidRow& c_row) except CInvalidRowResult_Error: + # A negative row number means undetermined (because of parallel reading) + row_number = c_row.number if c_row.number >= 0 else None + row = InvalidRow(c_row.expected_columns, c_row.actual_columns, + row_number, frombytes( c_row.text)) + result = handler(row) + if result == 'error': + return CInvalidRowResult_Error + elif result == 'skip': + return CInvalidRowResult_Skip + else: + raise ValueError("Invalid return value for invalid row handler: " + f"expected 'error' or 'skip', got {result!r}") + + +cdef class ReadOptions(_Weakrefable): + """ + Options for reading CSV files. + + Parameters + ---------- + use_threads : bool, optional (default True) + Whether to use multiple threads to accelerate reading + block_size : int, optional + How much bytes to process at a time from the input stream. + This will determine multi-threading granularity as well as + the size of individual record batches or table chunks. + Minimum valid value for block size is 1 + skip_rows : int, optional (default 0) + The number of rows to skip before the column names (if any) + and the CSV data. + skip_rows_after_names : int, optional (default 0) + The number of rows to skip after the column names. + This number can be larger than the number of rows in one + block, and empty rows are counted. + The order of application is as follows: + - `skip_rows` is applied (if non-zero); + - column names are read (unless `column_names` is set); + - `skip_rows_after_names` is applied (if non-zero). + column_names : list, optional + The column names of the target table. If empty, fall back on + `autogenerate_column_names`. + autogenerate_column_names : bool, optional (default False) + Whether to autogenerate column names if `column_names` is empty. + If true, column names will be of the form "f0", "f1"... + If false, column names will be read from the first CSV row + after `skip_rows`. + encoding : str, optional (default 'utf8') + The character encoding of the CSV data. Columns that cannot + decode using this encoding can still be read as Binary. + + Examples + -------- + + Defining an example data: + + >>> import io + >>> s = "1,2,3\\nFlamingo,2,2022-03-01\\nHorse,4,2022-03-02\\nBrittle stars,5,2022-03-03\\nCentipede,100,2022-03-04" + >>> print(s) + 1,2,3 + Flamingo,2,2022-03-01 + Horse,4,2022-03-02 + Brittle stars,5,2022-03-03 + Centipede,100,2022-03-04 + + Ignore the first numbered row and substitute it with defined + or autogenerated column names: + + >>> from pyarrow import csv + >>> read_options = csv.ReadOptions( + ... column_names=["animals", "n_legs", "entry"], + ... skip_rows=1) + >>> csv.read_csv(io.BytesIO(s.encode()), read_options=read_options) + pyarrow.Table + animals: string + n_legs: int64 + entry: date32[day] + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede"]] + n_legs: [[2,4,5,100]] + entry: [[2022-03-01,2022-03-02,2022-03-03,2022-03-04]] + + >>> read_options = csv.ReadOptions(autogenerate_column_names=True, + ... skip_rows=1) + >>> csv.read_csv(io.BytesIO(s.encode()), read_options=read_options) + pyarrow.Table + f0: string + f1: int64 + f2: date32[day] + ---- + f0: [["Flamingo","Horse","Brittle stars","Centipede"]] + f1: [[2,4,5,100]] + f2: [[2022-03-01,2022-03-02,2022-03-03,2022-03-04]] + + Remove the first 2 rows of the data: + + >>> read_options = csv.ReadOptions(skip_rows_after_names=2) + >>> csv.read_csv(io.BytesIO(s.encode()), read_options=read_options) + pyarrow.Table + 1: string + 2: int64 + 3: date32[day] + ---- + 1: [["Brittle stars","Centipede"]] + 2: [[5,100]] + 3: [[2022-03-03,2022-03-04]] + """ + + # Avoid mistakingly creating attributes + __slots__ = () + + # __init__() is not called when unpickling, initialize storage here + def __cinit__(self, *argw, **kwargs): + self.options.reset(new CCSVReadOptions(CCSVReadOptions.Defaults())) + + def __init__(self, *, use_threads=None, block_size=None, skip_rows=None, + skip_rows_after_names=None, column_names=None, + autogenerate_column_names=None, encoding='utf8'): + if use_threads is not None: + self.use_threads = use_threads + if block_size is not None: + self.block_size = block_size + if skip_rows is not None: + self.skip_rows = skip_rows + if skip_rows_after_names is not None: + self.skip_rows_after_names = skip_rows_after_names + if column_names is not None: + self.column_names = column_names + if autogenerate_column_names is not None: + self.autogenerate_column_names= autogenerate_column_names + # Python-specific option + self.encoding = encoding + + @property + def use_threads(self): + """ + Whether to use multiple threads to accelerate reading. + """ + return deref(self.options).use_threads + + @use_threads.setter + def use_threads(self, value): + deref(self.options).use_threads = value + + @property + def block_size(self): + """ + How much bytes to process at a time from the input stream. + This will determine multi-threading granularity as well as + the size of individual record batches or table chunks. + """ + return deref(self.options).block_size + + @block_size.setter + def block_size(self, value): + deref(self.options).block_size = value + + @property + def skip_rows(self): + """ + The number of rows to skip before the column names (if any) + and the CSV data. + See `skip_rows_after_names` for interaction description + """ + return deref(self.options).skip_rows + + @skip_rows.setter + def skip_rows(self, value): + deref(self.options).skip_rows = value + + @property + def skip_rows_after_names(self): + """ + The number of rows to skip after the column names. + This number can be larger than the number of rows in one + block, and empty rows are counted. + The order of application is as follows: + - `skip_rows` is applied (if non-zero); + - column names are read (unless `column_names` is set); + - `skip_rows_after_names` is applied (if non-zero). + """ + return deref(self.options).skip_rows_after_names + + @skip_rows_after_names.setter + def skip_rows_after_names(self, value): + deref(self.options).skip_rows_after_names = value + + @property + def column_names(self): + """ + The column names of the target table. If empty, fall back on + `autogenerate_column_names`. + """ + return [frombytes(s) for s in deref(self.options).column_names] + + @column_names.setter + def column_names(self, value): + deref(self.options).column_names.clear() + for item in value: + deref(self.options).column_names.push_back(tobytes(item)) + + @property + def autogenerate_column_names(self): + """ + Whether to autogenerate column names if `column_names` is empty. + If true, column names will be of the form "f0", "f1"... + If false, column names will be read from the first CSV row + after `skip_rows`. + """ + return deref(self.options).autogenerate_column_names + + @autogenerate_column_names.setter + def autogenerate_column_names(self, value): + deref(self.options).autogenerate_column_names = value + + def validate(self): + check_status(deref(self.options).Validate()) + + def equals(self, ReadOptions other): + """ + Parameters + ---------- + other : pyarrow.csv.ReadOptions + + Returns + ------- + bool + """ + return ( + self.use_threads == other.use_threads and + self.block_size == other.block_size and + self.skip_rows == other.skip_rows and + self.skip_rows_after_names == other.skip_rows_after_names and + self.column_names == other.column_names and + self.autogenerate_column_names == + other.autogenerate_column_names and + self.encoding == other.encoding + ) + + @staticmethod + cdef ReadOptions wrap(CCSVReadOptions options): + out = ReadOptions() + out.options.reset(new CCSVReadOptions(move(options))) + out.encoding = 'utf8' # No way to know this + return out + + def __getstate__(self): + return (self.use_threads, self.block_size, self.skip_rows, + self.column_names, self.autogenerate_column_names, + self.encoding, self.skip_rows_after_names) + + def __setstate__(self, state): + (self.use_threads, self.block_size, self.skip_rows, + self.column_names, self.autogenerate_column_names, + self.encoding, self.skip_rows_after_names) = state + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return False + + +cdef class ParseOptions(_Weakrefable): + """ + Options for parsing CSV files. + + Parameters + ---------- + delimiter : 1-character string, optional (default ',') + The character delimiting individual cells in the CSV data. + quote_char : 1-character string or False, optional (default '"') + The character used optionally for quoting CSV values + (False if quoting is not allowed). + double_quote : bool, optional (default True) + Whether two quotes in a quoted CSV value denote a single quote + in the data. + escape_char : 1-character string or False, optional (default False) + The character used optionally for escaping special characters + (False if escaping is not allowed). + newlines_in_values : bool, optional (default False) + Whether newline characters are allowed in CSV values. + Setting this to True reduces the performance of multi-threaded + CSV reading. + ignore_empty_lines : bool, optional (default True) + Whether empty lines are ignored in CSV input. + If False, an empty line is interpreted as containing a single empty + value (assuming a one-column CSV file). + invalid_row_handler : callable, optional (default None) + If not None, this object is called for each CSV row that fails + parsing (because of a mismatching number of columns). + It should accept a single InvalidRow argument and return either + "skip" or "error" depending on the desired outcome. + + Examples + -------- + + Defining an example file from bytes object: + + >>> import io + >>> s = ( + ... "animals;n_legs;entry\\n" + ... "Flamingo;2;2022-03-01\\n" + ... "# Comment here:\\n" + ... "Horse;4;2022-03-02\\n" + ... "Brittle stars;5;2022-03-03\\n" + ... "Centipede;100;2022-03-04" + ... ) + >>> print(s) + animals;n_legs;entry + Flamingo;2;2022-03-01 + # Comment here: + Horse;4;2022-03-02 + Brittle stars;5;2022-03-03 + Centipede;100;2022-03-04 + >>> source = io.BytesIO(s.encode()) + + Read the data from a file skipping rows with comments + and defining the delimiter: + + >>> from pyarrow import csv + >>> def skip_comment(row): + ... if row.text.startswith("# "): + ... return 'skip' + ... else: + ... return 'error' + ... + >>> parse_options = csv.ParseOptions(delimiter=";", invalid_row_handler=skip_comment) + >>> csv.read_csv(source, parse_options=parse_options) + pyarrow.Table + animals: string + n_legs: int64 + entry: date32[day] + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede"]] + n_legs: [[2,4,5,100]] + entry: [[2022-03-01,2022-03-02,2022-03-03,2022-03-04]] + """ + __slots__ = () + + def __cinit__(self, *argw, **kwargs): + self._invalid_row_handler = None + self.options.reset(new CCSVParseOptions(CCSVParseOptions.Defaults())) + + def __init__(self, *, delimiter=None, quote_char=None, double_quote=None, + escape_char=None, newlines_in_values=None, + ignore_empty_lines=None, invalid_row_handler=None): + if delimiter is not None: + self.delimiter = delimiter + if quote_char is not None: + self.quote_char = quote_char + if double_quote is not None: + self.double_quote = double_quote + if escape_char is not None: + self.escape_char = escape_char + if newlines_in_values is not None: + self.newlines_in_values = newlines_in_values + if ignore_empty_lines is not None: + self.ignore_empty_lines = ignore_empty_lines + if invalid_row_handler is not None: + self.invalid_row_handler = invalid_row_handler + + @property + def delimiter(self): + """ + The character delimiting individual cells in the CSV data. + """ + return chr(deref(self.options).delimiter) + + @delimiter.setter + def delimiter(self, value): + deref(self.options).delimiter = _single_char(value) + + @property + def quote_char(self): + """ + The character used optionally for quoting CSV values + (False if quoting is not allowed). + """ + if deref(self.options).quoting: + return chr(deref(self.options).quote_char) + else: + return False + + @quote_char.setter + def quote_char(self, value): + if value is False: + deref(self.options).quoting = False + else: + deref(self.options).quote_char = _single_char(value) + deref(self.options).quoting = True + + @property + def double_quote(self): + """ + Whether two quotes in a quoted CSV value denote a single quote + in the data. + """ + return deref(self.options).double_quote + + @double_quote.setter + def double_quote(self, value): + deref(self.options).double_quote = value + + @property + def escape_char(self): + """ + The character used optionally for escaping special characters + (False if escaping is not allowed). + """ + if deref(self.options).escaping: + return chr(deref(self.options).escape_char) + else: + return False + + @escape_char.setter + def escape_char(self, value): + if value is False: + deref(self.options).escaping = False + else: + deref(self.options).escape_char = _single_char(value) + deref(self.options).escaping = True + + @property + def newlines_in_values(self): + """ + Whether newline characters are allowed in CSV values. + Setting this to True reduces the performance of multi-threaded + CSV reading. + """ + return deref(self.options).newlines_in_values + + @newlines_in_values.setter + def newlines_in_values(self, value): + deref(self.options).newlines_in_values = value + + @property + def ignore_empty_lines(self): + """ + Whether empty lines are ignored in CSV input. + If False, an empty line is interpreted as containing a single empty + value (assuming a one-column CSV file). + """ + return deref(self.options).ignore_empty_lines + + @property + def invalid_row_handler(self): + """ + Optional handler for invalid rows. + + If not None, this object is called for each CSV row that fails + parsing (because of a mismatching number of columns). + It should accept a single InvalidRow argument and return either + "skip" or "error" depending on the desired outcome. + """ + return self._invalid_row_handler + + @invalid_row_handler.setter + def invalid_row_handler(self, value): + if value is not None and not callable(value): + raise TypeError("Expected callable or None, " + f"got instance of {type(value)!r}") + self._invalid_row_handler = value + deref(self.options).invalid_row_handler = MakeInvalidRowHandler( + &_handle_invalid_row, value) + + @ignore_empty_lines.setter + def ignore_empty_lines(self, value): + deref(self.options).ignore_empty_lines = value + + def validate(self): + check_status(deref(self.options).Validate()) + + def equals(self, ParseOptions other): + """ + Parameters + ---------- + other : pyarrow.csv.ParseOptions + + Returns + ------- + bool + """ + return ( + self.delimiter == other.delimiter and + self.quote_char == other.quote_char and + self.double_quote == other.double_quote and + self.escape_char == other.escape_char and + self.newlines_in_values == other.newlines_in_values and + self.ignore_empty_lines == other.ignore_empty_lines and + self._invalid_row_handler == other._invalid_row_handler + ) + + @staticmethod + cdef ParseOptions wrap(CCSVParseOptions options): + out = ParseOptions() + out.options.reset(new CCSVParseOptions(move(options))) + return out + + def __getstate__(self): + return (self.delimiter, self.quote_char, self.double_quote, + self.escape_char, self.newlines_in_values, + self.ignore_empty_lines, self.invalid_row_handler) + + def __setstate__(self, state): + (self.delimiter, self.quote_char, self.double_quote, + self.escape_char, self.newlines_in_values, + self.ignore_empty_lines, self.invalid_row_handler) = state + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return False + + +cdef class _ISO8601(_Weakrefable): + """ + A special object indicating ISO-8601 parsing. + """ + __slots__ = () + + def __str__(self): + return 'ISO8601' + + def __eq__(self, other): + return isinstance(other, _ISO8601) + + +ISO8601 = _ISO8601() + + +cdef class ConvertOptions(_Weakrefable): + """ + Options for converting CSV data. + + Parameters + ---------- + check_utf8 : bool, optional (default True) + Whether to check UTF8 validity of string columns. + column_types : pyarrow.Schema or dict, optional + Explicitly map column names to column types. Passing this argument + disables type inference on the defined columns. + null_values : list, optional + A sequence of strings that denote nulls in the data + (defaults are appropriate in most cases). Note that by default, + string columns are not checked for null values. To enable + null checking for those, specify ``strings_can_be_null=True``. + true_values : list, optional + A sequence of strings that denote true booleans in the data + (defaults are appropriate in most cases). + false_values : list, optional + A sequence of strings that denote false booleans in the data + (defaults are appropriate in most cases). + decimal_point : 1-character string, optional (default '.') + The character used as decimal point in floating-point and decimal + data. + strings_can_be_null : bool, optional (default False) + Whether string / binary columns can have null values. + If true, then strings in null_values are considered null for + string columns. + If false, then all strings are valid string values. + quoted_strings_can_be_null : bool, optional (default True) + Whether quoted values can be null. + If true, then strings in "null_values" are also considered null + when they appear quoted in the CSV file. Otherwise, quoted values + are never considered null. + include_columns : list, optional + The names of columns to include in the Table. + If empty, the Table will include all columns from the CSV file. + If not empty, only these columns will be included, in this order. + include_missing_columns : bool, optional (default False) + If false, columns in `include_columns` but not in the CSV file will + error out. + If true, columns in `include_columns` but not in the CSV file will + produce a column of nulls (whose type is selected using + `column_types`, or null by default). + This option is ignored if `include_columns` is empty. + auto_dict_encode : bool, optional (default False) + Whether to try to automatically dict-encode string / binary data. + If true, then when type inference detects a string or binary column, + it it dict-encoded up to `auto_dict_max_cardinality` distinct values + (per chunk), after which it switches to regular encoding. + This setting is ignored for non-inferred columns (those in + `column_types`). + auto_dict_max_cardinality : int, optional + The maximum dictionary cardinality for `auto_dict_encode`. + This value is per chunk. + timestamp_parsers : list, optional + A sequence of strptime()-compatible format strings, tried in order + when attempting to infer or convert timestamp values (the special + value ISO8601() can also be given). By default, a fast built-in + ISO-8601 parser is used. + + Examples + -------- + + Defining an example data: + + >>> import io + >>> s = ( + ... "animals,n_legs,entry,fast\\n" + ... "Flamingo,2,01/03/2022,Yes\\n" + ... "Horse,4,02/03/2022,Yes\\n" + ... "Brittle stars,5,03/03/2022,No\\n" + ... "Centipede,100,04/03/2022,No\\n" + ... ",6,05/03/2022," + ... ) + >>> print(s) + animals,n_legs,entry,fast + Flamingo,2,01/03/2022,Yes + Horse,4,02/03/2022,Yes + Brittle stars,5,03/03/2022,No + Centipede,100,04/03/2022,No + ,6,05/03/2022, + + Change the type of a column: + + >>> import pyarrow as pa + >>> from pyarrow import csv + >>> convert_options = csv.ConvertOptions(column_types={"n_legs": pa.float64()}) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + animals: string + n_legs: double + entry: string + fast: string + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede",""]] + n_legs: [[2,4,5,100,6]] + entry: [["01/03/2022","02/03/2022","03/03/2022","04/03/2022","05/03/2022"]] + fast: [["Yes","Yes","No","No",""]] + + Define a date parsing format to get a timestamp type column + (in case dates are not in ISO format and not converted by default): + + >>> convert_options = csv.ConvertOptions( + ... timestamp_parsers=["%m/%d/%Y", "%m-%d-%Y"]) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + animals: string + n_legs: int64 + entry: timestamp[s] + fast: string + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede",""]] + n_legs: [[2,4,5,100,6]] + entry: [[2022-01-03 00:00:00,2022-02-03 00:00:00,2022-03-03 00:00:00,2022-04-03 00:00:00,2022-05-03 00:00:00]] + fast: [["Yes","Yes","No","No",""]] + + Specify a subset of columns to be read: + + >>> convert_options = csv.ConvertOptions( + ... include_columns=["animals", "n_legs"]) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + animals: string + n_legs: int64 + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede",""]] + n_legs: [[2,4,5,100,6]] + + List additional column to be included as a null typed column: + + >>> convert_options = csv.ConvertOptions( + ... include_columns=["animals", "n_legs", "location"], + ... include_missing_columns=True) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + animals: string + n_legs: int64 + location: null + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede",""]] + n_legs: [[2,4,5,100,6]] + location: [5 nulls] + + Define columns as dictionary type (by default only the + string/binary columns are dictionary encoded): + + >>> convert_options = csv.ConvertOptions( + ... timestamp_parsers=["%m/%d/%Y", "%m-%d-%Y"], + ... auto_dict_encode=True) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + animals: dictionary + n_legs: int64 + entry: timestamp[s] + fast: dictionary + ---- + animals: [ -- dictionary: + ["Flamingo","Horse","Brittle stars","Centipede",""] -- indices: + [0,1,2,3,4]] + n_legs: [[2,4,5,100,6]] + entry: [[2022-01-03 00:00:00,2022-02-03 00:00:00,2022-03-03 00:00:00,2022-04-03 00:00:00,2022-05-03 00:00:00]] + fast: [ -- dictionary: + ["Yes","No",""] -- indices: + [0,0,1,1,2]] + + Set upper limit for the number of categories. If the categories + is more than the limit, the conversion to dictionary will not + happen: + + >>> convert_options = csv.ConvertOptions( + ... include_columns=["animals"], + ... auto_dict_encode=True, + ... auto_dict_max_cardinality=2) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + animals: string + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede",""]] + + Set empty strings to missing values: + + >>> convert_options = csv.ConvertOptions(include_columns=["animals", "n_legs"], + ... strings_can_be_null=True) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + animals: string + n_legs: int64 + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede",null]] + n_legs: [[2,4,5,100,6]] + + Define values to be True and False when converting a column + into a bool type: + + >>> convert_options = csv.ConvertOptions( + ... include_columns=["fast"], + ... false_values=["No"], + ... true_values=["Yes"]) + >>> csv.read_csv(io.BytesIO(s.encode()), convert_options=convert_options) + pyarrow.Table + fast: bool + ---- + fast: [[true,true,false,false,null]] + """ + + # Avoid mistakingly creating attributes + __slots__ = () + + def __cinit__(self, *argw, **kwargs): + self.options.reset( + new CCSVConvertOptions(CCSVConvertOptions.Defaults())) + + def __init__(self, *, check_utf8=None, column_types=None, null_values=None, + true_values=None, false_values=None, decimal_point=None, + strings_can_be_null=None, quoted_strings_can_be_null=None, + include_columns=None, include_missing_columns=None, + auto_dict_encode=None, auto_dict_max_cardinality=None, + timestamp_parsers=None): + if check_utf8 is not None: + self.check_utf8 = check_utf8 + if column_types is not None: + self.column_types = column_types + if null_values is not None: + self.null_values = null_values + if true_values is not None: + self.true_values = true_values + if false_values is not None: + self.false_values = false_values + if decimal_point is not None: + self.decimal_point = decimal_point + if strings_can_be_null is not None: + self.strings_can_be_null = strings_can_be_null + if quoted_strings_can_be_null is not None: + self.quoted_strings_can_be_null = quoted_strings_can_be_null + if include_columns is not None: + self.include_columns = include_columns + if include_missing_columns is not None: + self.include_missing_columns = include_missing_columns + if auto_dict_encode is not None: + self.auto_dict_encode = auto_dict_encode + if auto_dict_max_cardinality is not None: + self.auto_dict_max_cardinality = auto_dict_max_cardinality + if timestamp_parsers is not None: + self.timestamp_parsers = timestamp_parsers + + @property + def check_utf8(self): + """ + Whether to check UTF8 validity of string columns. + """ + return deref(self.options).check_utf8 + + @check_utf8.setter + def check_utf8(self, value): + deref(self.options).check_utf8 = value + + @property + def strings_can_be_null(self): + """ + Whether string / binary columns can have null values. + """ + return deref(self.options).strings_can_be_null + + @strings_can_be_null.setter + def strings_can_be_null(self, value): + deref(self.options).strings_can_be_null = value + + @property + def quoted_strings_can_be_null(self): + """ + Whether quoted values can be null. + """ + return deref(self.options).quoted_strings_can_be_null + + @quoted_strings_can_be_null.setter + def quoted_strings_can_be_null(self, value): + deref(self.options).quoted_strings_can_be_null = value + + @property + def column_types(self): + """ + Explicitly map column names to column types. + """ + d = {frombytes(item.first): pyarrow_wrap_data_type(item.second) + for item in deref(self.options).column_types} + return d + + @column_types.setter + def column_types(self, value): + cdef: + shared_ptr[CDataType] typ + + if isinstance(value, Mapping): + value = value.items() + + deref(self.options).column_types.clear() + for item in value: + if isinstance(item, Field): + k = item.name + v = item.type + else: + k, v = item + typ = pyarrow_unwrap_data_type(ensure_type(v)) + assert typ != NULL + deref(self.options).column_types[tobytes(k)] = typ + + @property + def null_values(self): + """ + A sequence of strings that denote nulls in the data. + """ + return [frombytes(x) for x in deref(self.options).null_values] + + @null_values.setter + def null_values(self, value): + deref(self.options).null_values = [tobytes(x) for x in value] + + @property + def true_values(self): + """ + A sequence of strings that denote true booleans in the data. + """ + return [frombytes(x) for x in deref(self.options).true_values] + + @true_values.setter + def true_values(self, value): + deref(self.options).true_values = [tobytes(x) for x in value] + + @property + def false_values(self): + """ + A sequence of strings that denote false booleans in the data. + """ + return [frombytes(x) for x in deref(self.options).false_values] + + @false_values.setter + def false_values(self, value): + deref(self.options).false_values = [tobytes(x) for x in value] + + @property + def decimal_point(self): + """ + The character used as decimal point in floating-point and decimal + data. + """ + return chr(deref(self.options).decimal_point) + + @decimal_point.setter + def decimal_point(self, value): + deref(self.options).decimal_point = _single_char(value) + + @property + def auto_dict_encode(self): + """ + Whether to try to automatically dict-encode string / binary data. + """ + return deref(self.options).auto_dict_encode + + @auto_dict_encode.setter + def auto_dict_encode(self, value): + deref(self.options).auto_dict_encode = value + + @property + def auto_dict_max_cardinality(self): + """ + The maximum dictionary cardinality for `auto_dict_encode`. + + This value is per chunk. + """ + return deref(self.options).auto_dict_max_cardinality + + @auto_dict_max_cardinality.setter + def auto_dict_max_cardinality(self, value): + deref(self.options).auto_dict_max_cardinality = value + + @property + def include_columns(self): + """ + The names of columns to include in the Table. + + If empty, the Table will include all columns from the CSV file. + If not empty, only these columns will be included, in this order. + """ + return [frombytes(s) for s in deref(self.options).include_columns] + + @include_columns.setter + def include_columns(self, value): + deref(self.options).include_columns.clear() + for item in value: + deref(self.options).include_columns.push_back(tobytes(item)) + + @property + def include_missing_columns(self): + """ + If false, columns in `include_columns` but not in the CSV file will + error out. + If true, columns in `include_columns` but not in the CSV file will + produce a null column (whose type is selected using `column_types`, + or null by default). + This option is ignored if `include_columns` is empty. + """ + return deref(self.options).include_missing_columns + + @include_missing_columns.setter + def include_missing_columns(self, value): + deref(self.options).include_missing_columns = value + + @property + def timestamp_parsers(self): + """ + A sequence of strptime()-compatible format strings, tried in order + when attempting to infer or convert timestamp values (the special + value ISO8601() can also be given). By default, a fast built-in + ISO-8601 parser is used. + """ + cdef: + shared_ptr[CTimestampParser] c_parser + c_string kind + + parsers = [] + for c_parser in deref(self.options).timestamp_parsers: + kind = deref(c_parser).kind() + if kind == b'strptime': + parsers.append(frombytes(deref(c_parser).format())) + else: + assert kind == b'iso8601' + parsers.append(ISO8601) + + return parsers + + @timestamp_parsers.setter + def timestamp_parsers(self, value): + cdef: + vector[shared_ptr[CTimestampParser]] c_parsers + + for v in value: + if isinstance(v, str): + c_parsers.push_back(CTimestampParser.MakeStrptime(tobytes(v))) + elif v == ISO8601: + c_parsers.push_back(CTimestampParser.MakeISO8601()) + else: + raise TypeError("Expected list of str or ISO8601 objects") + + deref(self.options).timestamp_parsers = move(c_parsers) + + @staticmethod + cdef ConvertOptions wrap(CCSVConvertOptions options): + out = ConvertOptions() + out.options.reset(new CCSVConvertOptions(move(options))) + return out + + def validate(self): + check_status(deref(self.options).Validate()) + + def equals(self, ConvertOptions other): + """ + Parameters + ---------- + other : pyarrow.csv.ConvertOptions + + Returns + ------- + bool + """ + return ( + self.check_utf8 == other.check_utf8 and + self.column_types == other.column_types and + self.null_values == other.null_values and + self.true_values == other.true_values and + self.false_values == other.false_values and + self.decimal_point == other.decimal_point and + self.timestamp_parsers == other.timestamp_parsers and + self.strings_can_be_null == other.strings_can_be_null and + self.quoted_strings_can_be_null == + other.quoted_strings_can_be_null and + self.auto_dict_encode == other.auto_dict_encode and + self.auto_dict_max_cardinality == + other.auto_dict_max_cardinality and + self.include_columns == other.include_columns and + self.include_missing_columns == other.include_missing_columns + ) + + def __getstate__(self): + return (self.check_utf8, self.column_types, self.null_values, + self.true_values, self.false_values, self.decimal_point, + self.timestamp_parsers, self.strings_can_be_null, + self.quoted_strings_can_be_null, self.auto_dict_encode, + self.auto_dict_max_cardinality, self.include_columns, + self.include_missing_columns) + + def __setstate__(self, state): + (self.check_utf8, self.column_types, self.null_values, + self.true_values, self.false_values, self.decimal_point, + self.timestamp_parsers, self.strings_can_be_null, + self.quoted_strings_can_be_null, self.auto_dict_encode, + self.auto_dict_max_cardinality, self.include_columns, + self.include_missing_columns) = state + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return False + + +cdef _get_reader(input_file, ReadOptions read_options, + shared_ptr[CInputStream]* out): + use_memory_map = False + get_input_stream(input_file, use_memory_map, out) + if read_options is not None: + out[0] = native_transcoding_input_stream(out[0], + read_options.encoding, + 'utf8') + + +cdef _get_read_options(ReadOptions read_options, CCSVReadOptions* out): + if read_options is None: + out[0] = CCSVReadOptions.Defaults() + else: + out[0] = deref(read_options.options) + + +cdef _get_parse_options(ParseOptions parse_options, CCSVParseOptions* out): + if parse_options is None: + out[0] = CCSVParseOptions.Defaults() + else: + out[0] = deref(parse_options.options) + + +cdef _get_convert_options(ConvertOptions convert_options, + CCSVConvertOptions* out): + if convert_options is None: + out[0] = CCSVConvertOptions.Defaults() + else: + out[0] = deref(convert_options.options) + + +cdef class CSVStreamingReader(RecordBatchReader): + """An object that reads record batches incrementally from a CSV file. + + Should not be instantiated directly by user code. + """ + cdef readonly: + Schema schema + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly, " + "use pyarrow.csv.open_csv() instead." + .format(self.__class__.__name__)) + + # Note about cancellation: we cannot create a SignalStopHandler + # by default here, as several CSVStreamingReader instances may be + # created (including by the same thread). Handling cancellation + # would require having the user pass the SignalStopHandler. + # (in addition to solving ARROW-11853) + + cdef _open(self, shared_ptr[CInputStream] stream, + CCSVReadOptions c_read_options, + CCSVParseOptions c_parse_options, + CCSVConvertOptions c_convert_options, + MemoryPool memory_pool): + cdef: + shared_ptr[CSchema] c_schema + CIOContext io_context + + io_context = CIOContext(maybe_unbox_memory_pool(memory_pool)) + + with nogil: + self.reader = GetResultValue( + CCSVStreamingReader.Make( + io_context, stream, + move(c_read_options), move(c_parse_options), + move(c_convert_options))) + c_schema = self.reader.get().schema() + + self.schema = pyarrow_wrap_schema(c_schema) + + +def read_csv(input_file, read_options=None, parse_options=None, + convert_options=None, MemoryPool memory_pool=None): + """ + Read a Table from a stream of CSV data. + + Parameters + ---------- + input_file : string, path or file-like object + The location of CSV data. If a string or path, and if it ends + with a recognized compressed file extension (e.g. ".gz" or ".bz2"), + the data is automatically decompressed when reading. + read_options : pyarrow.csv.ReadOptions, optional + Options for the CSV reader (see pyarrow.csv.ReadOptions constructor + for defaults) + parse_options : pyarrow.csv.ParseOptions, optional + Options for the CSV parser + (see pyarrow.csv.ParseOptions constructor for defaults) + convert_options : pyarrow.csv.ConvertOptions, optional + Options for converting CSV data + (see pyarrow.csv.ConvertOptions constructor for defaults) + memory_pool : MemoryPool, optional + Pool to allocate Table memory from + + Returns + ------- + :class:`pyarrow.Table` + Contents of the CSV file as a in-memory table. + + Examples + -------- + + Defining an example file from bytes object: + + >>> import io + >>> s = ( + ... "animals,n_legs,entry\\n" + ... "Flamingo,2,2022-03-01\\n" + ... "Horse,4,2022-03-02\\n" + ... "Brittle stars,5,2022-03-03\\n" + ... "Centipede,100,2022-03-04" + ... ) + >>> print(s) + animals,n_legs,entry + Flamingo,2,2022-03-01 + Horse,4,2022-03-02 + Brittle stars,5,2022-03-03 + Centipede,100,2022-03-04 + >>> source = io.BytesIO(s.encode()) + + Reading from the file + + >>> from pyarrow import csv + >>> csv.read_csv(source) + pyarrow.Table + animals: string + n_legs: int64 + entry: date32[day] + ---- + animals: [["Flamingo","Horse","Brittle stars","Centipede"]] + n_legs: [[2,4,5,100]] + entry: [[2022-03-01,2022-03-02,2022-03-03,2022-03-04]] + """ + cdef: + shared_ptr[CInputStream] stream + CCSVReadOptions c_read_options + CCSVParseOptions c_parse_options + CCSVConvertOptions c_convert_options + CIOContext io_context + SharedPtrNoGIL[CCSVReader] reader + shared_ptr[CTable] table + + _get_reader(input_file, read_options, &stream) + _get_read_options(read_options, &c_read_options) + _get_parse_options(parse_options, &c_parse_options) + _get_convert_options(convert_options, &c_convert_options) + + with SignalStopHandler() as stop_handler: + io_context = CIOContext( + maybe_unbox_memory_pool(memory_pool), + ( stop_handler.stop_token).stop_token) + reader = GetResultValue(CCSVReader.Make( + io_context, stream, + c_read_options, c_parse_options, c_convert_options)) + + with nogil: + table = GetResultValue(reader.get().Read()) + + return pyarrow_wrap_table(table) + + +def open_csv(input_file, read_options=None, parse_options=None, + convert_options=None, MemoryPool memory_pool=None): + """ + Open a streaming reader of CSV data. + + Reading using this function is always single-threaded. + + Parameters + ---------- + input_file : string, path or file-like object + The location of CSV data. If a string or path, and if it ends + with a recognized compressed file extension (e.g. ".gz" or ".bz2"), + the data is automatically decompressed when reading. + read_options : pyarrow.csv.ReadOptions, optional + Options for the CSV reader (see pyarrow.csv.ReadOptions constructor + for defaults) + parse_options : pyarrow.csv.ParseOptions, optional + Options for the CSV parser + (see pyarrow.csv.ParseOptions constructor for defaults) + convert_options : pyarrow.csv.ConvertOptions, optional + Options for converting CSV data + (see pyarrow.csv.ConvertOptions constructor for defaults) + memory_pool : MemoryPool, optional + Pool to allocate Table memory from + + Returns + ------- + :class:`pyarrow.csv.CSVStreamingReader` + """ + cdef: + shared_ptr[CInputStream] stream + CCSVReadOptions c_read_options + CCSVParseOptions c_parse_options + CCSVConvertOptions c_convert_options + CSVStreamingReader reader + + _get_reader(input_file, read_options, &stream) + _get_read_options(read_options, &c_read_options) + _get_parse_options(parse_options, &c_parse_options) + _get_convert_options(convert_options, &c_convert_options) + + reader = CSVStreamingReader.__new__(CSVStreamingReader) + reader._open(stream, move(c_read_options), move(c_parse_options), + move(c_convert_options), memory_pool) + return reader + + +def _raise_invalid_function_option(value, description, *, + exception_class=ValueError): + raise exception_class(f"\"{value}\" is not a valid {description}") + + +cdef CQuotingStyle unwrap_quoting_style(quoting_style) except *: + if quoting_style == "needed": + return CQuotingStyle_Needed + elif quoting_style == "all_valid": + return CQuotingStyle_AllValid + elif quoting_style == "none": + return CQuotingStyle_None + _raise_invalid_function_option(quoting_style, "quoting style") + + +cdef wrap_quoting_style(quoting_style): + if quoting_style == CQuotingStyle_Needed: + return 'needed' + elif quoting_style == CQuotingStyle_AllValid: + return 'all_valid' + elif quoting_style == CQuotingStyle_None: + return 'none' + + +cdef class WriteOptions(_Weakrefable): + """ + Options for writing CSV files. + + Parameters + ---------- + include_header : bool, optional (default True) + Whether to write an initial header line with column names + batch_size : int, optional (default 1024) + How many rows to process together when converting and writing + CSV data + delimiter : 1-character string, optional (default ",") + The character delimiting individual cells in the CSV data. + quoting_style : str, optional (default "needed") + Whether to quote values, and if so, which quoting style to use. + The following values are accepted: + + - "needed" (default): only enclose values in quotes when needed. + - "all_valid": enclose all valid values in quotes; nulls are not quoted. + - "none": do not enclose any values in quotes; values containing + special characters (such as quotes, cell delimiters or line endings) + will raise an error. + """ + + # Avoid mistakingly creating attributes + __slots__ = () + + def __init__(self, *, include_header=None, batch_size=None, + delimiter=None, quoting_style=None): + self.options.reset(new CCSVWriteOptions(CCSVWriteOptions.Defaults())) + if include_header is not None: + self.include_header = include_header + if batch_size is not None: + self.batch_size = batch_size + if delimiter is not None: + self.delimiter = delimiter + if quoting_style is not None: + self.quoting_style = quoting_style + + @property + def include_header(self): + """ + Whether to write an initial header line with column names. + """ + return deref(self.options).include_header + + @include_header.setter + def include_header(self, value): + deref(self.options).include_header = value + + @property + def batch_size(self): + """ + How many rows to process together when converting and writing + CSV data. + """ + return deref(self.options).batch_size + + @batch_size.setter + def batch_size(self, value): + deref(self.options).batch_size = value + + @property + def delimiter(self): + """ + The character delimiting individual cells in the CSV data. + """ + return chr(deref(self.options).delimiter) + + @delimiter.setter + def delimiter(self, value): + deref(self.options).delimiter = _single_char(value) + + @property + def quoting_style(self): + """ + Whether to quote values, and if so, which quoting style to use. + The following values are accepted: + + - "needed" (default): only enclose values in quotes when needed. + - "all_valid": enclose all valid values in quotes; nulls are not quoted. + - "none": do not enclose any values in quotes; values containing + special characters (such as quotes, cell delimiters or line endings) + will raise an error. + """ + return wrap_quoting_style(deref(self.options).quoting_style) + + @quoting_style.setter + def quoting_style(self, value): + deref(self.options).quoting_style = unwrap_quoting_style(value) + + @staticmethod + cdef WriteOptions wrap(CCSVWriteOptions options): + out = WriteOptions() + out.options.reset(new CCSVWriteOptions(move(options))) + return out + + def validate(self): + check_status(self.options.get().Validate()) + + +cdef _get_write_options(WriteOptions write_options, CCSVWriteOptions* out): + if write_options is None: + out[0] = CCSVWriteOptions.Defaults() + else: + out[0] = deref(write_options.options) + + +def write_csv(data, output_file, write_options=None, + MemoryPool memory_pool=None): + """ + Write record batch or table to a CSV file. + + Parameters + ---------- + data : pyarrow.RecordBatch or pyarrow.Table + The data to write. + output_file : string, path, pyarrow.NativeFile, or file-like object + The location where to write the CSV data. + write_options : pyarrow.csv.WriteOptions + Options to configure writing the CSV data. + memory_pool : MemoryPool, optional + Pool for temporary allocations. + + Examples + -------- + + >>> import pyarrow as pa + >>> from pyarrow import csv + + >>> legs = pa.array([2, 4, 5, 100]) + >>> animals = pa.array(["Flamingo", "Horse", "Brittle stars", "Centipede"]) + >>> entry_date = pa.array(["01/03/2022", "02/03/2022", + ... "03/03/2022", "04/03/2022"]) + >>> table = pa.table([animals, legs, entry_date], + ... names=["animals", "n_legs", "entry"]) + + >>> csv.write_csv(table, "animals.csv") + + >>> write_options = csv.WriteOptions(include_header=False) + >>> csv.write_csv(table, "animals.csv", write_options=write_options) + + >>> write_options = csv.WriteOptions(delimiter=";") + >>> csv.write_csv(table, "animals.csv", write_options=write_options) + """ + cdef: + shared_ptr[COutputStream] stream + CCSVWriteOptions c_write_options + CMemoryPool* c_memory_pool + CRecordBatch* batch + CTable* table + _get_write_options(write_options, &c_write_options) + + get_writer(output_file, &stream) + c_memory_pool = maybe_unbox_memory_pool(memory_pool) + c_write_options.io_context = CIOContext(c_memory_pool) + if isinstance(data, RecordBatch): + batch = pyarrow_unwrap_batch(data).get() + with nogil: + check_status(WriteCSV(deref(batch), c_write_options, stream.get())) + elif isinstance(data, Table): + table = pyarrow_unwrap_table(data).get() + with nogil: + check_status(WriteCSV(deref(table), c_write_options, stream.get())) + else: + raise TypeError(f"Expected Table or RecordBatch, got '{type(data)}'") + + +cdef class CSVWriter(_CRecordBatchWriter): + """ + Writer to create a CSV file. + + Parameters + ---------- + sink : str, path, pyarrow.OutputStream or file-like object + The location where to write the CSV data. + schema : pyarrow.Schema + The schema of the data to be written. + write_options : pyarrow.csv.WriteOptions + Options to configure writing the CSV data. + memory_pool : MemoryPool, optional + Pool for temporary allocations. + """ + + def __init__(self, sink, Schema schema, *, + WriteOptions write_options=None, MemoryPool memory_pool=None): + cdef: + shared_ptr[COutputStream] c_stream + shared_ptr[CSchema] c_schema = pyarrow_unwrap_schema(schema) + CCSVWriteOptions c_write_options + CMemoryPool* c_memory_pool = maybe_unbox_memory_pool(memory_pool) + _get_write_options(write_options, &c_write_options) + c_write_options.io_context = CIOContext(c_memory_pool) + get_writer(sink, &c_stream) + with nogil: + self.writer = GetResultValue(MakeCSVWriter( + c_stream, c_schema, c_write_options)) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_dataset_orc.cpython-310-x86_64-linux-gnu.so b/vllm/lib/python3.10/site-packages/pyarrow/_dataset_orc.cpython-310-x86_64-linux-gnu.so new file mode 100644 index 0000000000000000000000000000000000000000..221f54787bbff62849ebf134afc9ec38c47a1380 Binary files /dev/null and b/vllm/lib/python3.10/site-packages/pyarrow/_dataset_orc.cpython-310-x86_64-linux-gnu.so differ diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_dataset_parquet_encryption.cpython-310-x86_64-linux-gnu.so b/vllm/lib/python3.10/site-packages/pyarrow/_dataset_parquet_encryption.cpython-310-x86_64-linux-gnu.so new file mode 100644 index 0000000000000000000000000000000000000000..31cfda0aee257007cbde610a366b9f25b31de953 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_dataset_parquet_encryption.cpython-310-x86_64-linux-gnu.so @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a7f92d2b69ed6924c97d577a0deb5a205612202ff4d305482875155c2e9429c +size 122184 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_flight.cpython-310-x86_64-linux-gnu.so b/vllm/lib/python3.10/site-packages/pyarrow/_flight.cpython-310-x86_64-linux-gnu.so new file mode 100644 index 0000000000000000000000000000000000000000..a66594b9305a3ee1678ce5ef8b889d3fbc6603b4 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_flight.cpython-310-x86_64-linux-gnu.so @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c37589d0c540141bced029f2902076bfc9724d60c7e0bbcea346d75d7a211c22 +size 1303088 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_flight.pyx b/vllm/lib/python3.10/site-packages/pyarrow/_flight.pyx new file mode 100644 index 0000000000000000000000000000000000000000..8289215de2e29c6cd7e09affd7ec5d377ee0fa9c --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_flight.pyx @@ -0,0 +1,3189 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: language_level = 3 + +import collections +import enum +import re +import time +import warnings +import weakref + +from cython.operator cimport dereference as deref +from cython.operator cimport postincrement +from libcpp cimport bool as c_bool + +from pyarrow.lib cimport * +from pyarrow.lib import (ArrowCancelled, ArrowException, ArrowInvalid, + SignalStopHandler) +from pyarrow.lib import as_buffer, frombytes, tobytes +from pyarrow.includes.libarrow_flight cimport * +from pyarrow.ipc import _get_legacy_format_default, _ReadPandasMixin +import pyarrow.lib as lib + + +cdef CFlightCallOptions DEFAULT_CALL_OPTIONS + + +cdef int check_flight_status(const CStatus& status) except -1 nogil: + cdef shared_ptr[FlightStatusDetail] detail + + if status.ok(): + return 0 + + detail = FlightStatusDetail.UnwrapStatus(status) + if detail: + with gil: + message = frombytes(status.message(), safe=True) + detail_msg = detail.get().extra_info() + if detail.get().code() == CFlightStatusInternal: + raise FlightInternalError(message, detail_msg) + elif detail.get().code() == CFlightStatusFailed: + message = _munge_grpc_python_error(message) + raise FlightServerError(message, detail_msg) + elif detail.get().code() == CFlightStatusTimedOut: + raise FlightTimedOutError(message, detail_msg) + elif detail.get().code() == CFlightStatusCancelled: + raise FlightCancelledError(message, detail_msg) + elif detail.get().code() == CFlightStatusUnauthenticated: + raise FlightUnauthenticatedError(message, detail_msg) + elif detail.get().code() == CFlightStatusUnauthorized: + raise FlightUnauthorizedError(message, detail_msg) + elif detail.get().code() == CFlightStatusUnavailable: + raise FlightUnavailableError(message, detail_msg) + + size_detail = FlightWriteSizeStatusDetail.UnwrapStatus(status) + if size_detail: + with gil: + message = frombytes(status.message(), safe=True) + raise FlightWriteSizeExceededError( + message, + size_detail.get().limit(), size_detail.get().actual()) + + return check_status(status) + + +_FLIGHT_SERVER_ERROR_REGEX = re.compile( + r'Flight RPC failed with message: (.*). Detail: ' + r'Python exception: (.*)', + re.DOTALL +) + + +def _munge_grpc_python_error(message): + m = _FLIGHT_SERVER_ERROR_REGEX.match(message) + if m: + return ('Flight RPC failed with Python exception \"{}: {}\"' + .format(m.group(2), m.group(1))) + else: + return message + + +cdef IpcWriteOptions _get_options(options): + return _get_legacy_format_default( + use_legacy_format=None, options=options) + + +cdef class FlightCallOptions(_Weakrefable): + """RPC-layer options for a Flight call.""" + + cdef: + CFlightCallOptions options + + def __init__(self, timeout=None, write_options=None, headers=None, + IpcReadOptions read_options=None): + """Create call options. + + Parameters + ---------- + timeout : float, None + A timeout for the call, in seconds. None means that the + timeout defaults to an implementation-specific value. + write_options : pyarrow.ipc.IpcWriteOptions, optional + IPC write options. The default options can be controlled + by environment variables (see pyarrow.ipc). + headers : List[Tuple[str, str]], optional + A list of arbitrary headers as key, value tuples + read_options : pyarrow.ipc.IpcReadOptions, optional + Serialization options for reading IPC format. + """ + cdef IpcWriteOptions c_write_options + + if timeout is not None: + self.options.timeout = CTimeoutDuration(timeout) + if write_options is not None: + c_write_options = _get_options(write_options) + self.options.write_options = c_write_options.c_options + if read_options is not None: + if not isinstance(read_options, IpcReadOptions): + raise TypeError("expected IpcReadOptions, got {}" + .format(type(read_options))) + self.options.read_options = read_options.c_options + if headers is not None: + self.options.headers = headers + + @staticmethod + cdef CFlightCallOptions* unwrap(obj): + if not obj: + return &DEFAULT_CALL_OPTIONS + elif isinstance(obj, FlightCallOptions): + return &(( obj).options) + raise TypeError("Expected a FlightCallOptions object, not " + "'{}'".format(type(obj))) + + +_CertKeyPair = collections.namedtuple('_CertKeyPair', ['cert', 'key']) + + +class CertKeyPair(_CertKeyPair): + """A TLS certificate and key for use in Flight.""" + + +cdef class FlightError(Exception): + """ + The base class for Flight-specific errors. + + A server may raise this class or one of its subclasses to provide + a more detailed error to clients. + + Parameters + ---------- + message : str, optional + The error message. + extra_info : bytes, optional + Extra binary error details that were provided by the + server/will be sent to the client. + + Attributes + ---------- + extra_info : bytes + Extra binary error details that were provided by the + server/will be sent to the client. + """ + + cdef dict __dict__ + + def __init__(self, message='', extra_info=b''): + super().__init__(message) + self.extra_info = tobytes(extra_info) + + cdef CStatus to_status(self): + message = tobytes("Flight error: {}".format(str(self))) + return CStatus_UnknownError(message) + + +cdef class FlightInternalError(FlightError, ArrowException): + """An error internal to the Flight server occurred.""" + + cdef CStatus to_status(self): + return MakeFlightError(CFlightStatusInternal, + tobytes(str(self)), self.extra_info) + + +cdef class FlightTimedOutError(FlightError, ArrowException): + """The Flight RPC call timed out.""" + + cdef CStatus to_status(self): + return MakeFlightError(CFlightStatusTimedOut, + tobytes(str(self)), self.extra_info) + + +cdef class FlightCancelledError(FlightError, ArrowCancelled): + """The operation was cancelled.""" + + cdef CStatus to_status(self): + return MakeFlightError(CFlightStatusCancelled, tobytes(str(self)), + self.extra_info) + + +cdef class FlightServerError(FlightError, ArrowException): + """A server error occurred.""" + + cdef CStatus to_status(self): + return MakeFlightError(CFlightStatusFailed, tobytes(str(self)), + self.extra_info) + + +cdef class FlightUnauthenticatedError(FlightError, ArrowException): + """The client is not authenticated.""" + + cdef CStatus to_status(self): + return MakeFlightError( + CFlightStatusUnauthenticated, tobytes(str(self)), self.extra_info) + + +cdef class FlightUnauthorizedError(FlightError, ArrowException): + """The client is not authorized to perform the given operation.""" + + cdef CStatus to_status(self): + return MakeFlightError(CFlightStatusUnauthorized, tobytes(str(self)), + self.extra_info) + + +cdef class FlightUnavailableError(FlightError, ArrowException): + """The server is not reachable or available.""" + + cdef CStatus to_status(self): + return MakeFlightError(CFlightStatusUnavailable, tobytes(str(self)), + self.extra_info) + + +class FlightWriteSizeExceededError(ArrowInvalid): + """A write operation exceeded the client-configured limit.""" + + def __init__(self, message, limit, actual): + super().__init__(message) + self.limit = limit + self.actual = actual + + +cdef class Action(_Weakrefable): + """An action executable on a Flight service.""" + cdef: + CAction action + + def __init__(self, action_type, buf): + """Create an action from a type and a buffer. + + Parameters + ---------- + action_type : bytes or str + buf : Buffer or bytes-like object + """ + self.action.type = tobytes(action_type) + self.action.body = pyarrow_unwrap_buffer(as_buffer(buf)) + + @property + def type(self): + """The action type.""" + return frombytes(self.action.type) + + @property + def body(self): + """The action body (arguments for the action).""" + return pyarrow_wrap_buffer(self.action.body) + + @staticmethod + cdef CAction unwrap(action) except *: + if not isinstance(action, Action): + raise TypeError("Must provide Action, not '{}'".format( + type(action))) + return ( action).action + + def serialize(self): + """Get the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + return GetResultValue(self.action.SerializeToString()) + + @classmethod + def deserialize(cls, serialized): + """Parse the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + cdef Action action = Action.__new__(Action) + action.action = GetResultValue( + CAction.Deserialize(tobytes(serialized))) + return action + + def __eq__(self, Action other): + return self.action == other.action + + def __repr__(self): + return (f"") + + +_ActionType = collections.namedtuple('_ActionType', ['type', 'description']) + + +class ActionType(_ActionType): + """A type of action that is executable on a Flight service.""" + + def make_action(self, buf): + """Create an Action with this type. + + Parameters + ---------- + buf : obj + An Arrow buffer or Python bytes or bytes-like object. + """ + return Action(self.type, buf) + + +cdef class Result(_Weakrefable): + """A result from executing an Action.""" + cdef: + unique_ptr[CFlightResult] result + + def __init__(self, buf): + """Create a new result. + + Parameters + ---------- + buf : Buffer or bytes-like object + """ + self.result.reset(new CFlightResult()) + self.result.get().body = pyarrow_unwrap_buffer(as_buffer(buf)) + + @property + def body(self): + """Get the Buffer containing the result.""" + return pyarrow_wrap_buffer(self.result.get().body) + + def serialize(self): + """Get the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + return GetResultValue(self.result.get().SerializeToString()) + + @classmethod + def deserialize(cls, serialized): + """Parse the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + cdef Result result = Result.__new__(Result) + result.result.reset(new CFlightResult(GetResultValue( + CFlightResult.Deserialize(tobytes(serialized))))) + return result + + def __eq__(self, Result other): + return deref(self.result.get()) == deref(other.result.get()) + + def __repr__(self): + return f"" + + +cdef class BasicAuth(_Weakrefable): + """A container for basic auth.""" + cdef: + unique_ptr[CBasicAuth] basic_auth + + def __init__(self, username=None, password=None): + """Create a new basic auth object. + + Parameters + ---------- + username : string + password : string + """ + self.basic_auth.reset(new CBasicAuth()) + if username: + self.basic_auth.get().username = tobytes(username) + if password: + self.basic_auth.get().password = tobytes(password) + + @property + def username(self): + """Get the username.""" + return self.basic_auth.get().username + + @property + def password(self): + """Get the password.""" + return self.basic_auth.get().password + + @staticmethod + def deserialize(serialized): + auth = BasicAuth() + auth.basic_auth.reset(new CBasicAuth(GetResultValue( + CBasicAuth.Deserialize(tobytes(serialized))))) + return auth + + def serialize(self): + return GetResultValue(self.basic_auth.get().SerializeToString()) + + def __eq__(self, BasicAuth other): + return deref(self.basic_auth.get()) == deref(other.basic_auth.get()) + + def __repr__(self): + return (f"") + + +class DescriptorType(enum.Enum): + """ + The type of a FlightDescriptor. + + Attributes + ---------- + + UNKNOWN + An unknown descriptor type. + + PATH + A Flight stream represented by a path. + + CMD + A Flight stream represented by an application-defined command. + + """ + + UNKNOWN = 0 + PATH = 1 + CMD = 2 + + +class FlightMethod(enum.Enum): + """The implemented methods in Flight.""" + + INVALID = 0 + HANDSHAKE = 1 + LIST_FLIGHTS = 2 + GET_FLIGHT_INFO = 3 + GET_SCHEMA = 4 + DO_GET = 5 + DO_PUT = 6 + DO_ACTION = 7 + LIST_ACTIONS = 8 + DO_EXCHANGE = 9 + + +cdef wrap_flight_method(CFlightMethod method): + if method == CFlightMethodHandshake: + return FlightMethod.HANDSHAKE + elif method == CFlightMethodListFlights: + return FlightMethod.LIST_FLIGHTS + elif method == CFlightMethodGetFlightInfo: + return FlightMethod.GET_FLIGHT_INFO + elif method == CFlightMethodGetSchema: + return FlightMethod.GET_SCHEMA + elif method == CFlightMethodDoGet: + return FlightMethod.DO_GET + elif method == CFlightMethodDoPut: + return FlightMethod.DO_PUT + elif method == CFlightMethodDoAction: + return FlightMethod.DO_ACTION + elif method == CFlightMethodListActions: + return FlightMethod.LIST_ACTIONS + elif method == CFlightMethodDoExchange: + return FlightMethod.DO_EXCHANGE + return FlightMethod.INVALID + + +cdef class FlightDescriptor(_Weakrefable): + """A description of a data stream available from a Flight service.""" + cdef: + CFlightDescriptor descriptor + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly, use " + "`pyarrow.flight.FlightDescriptor.for_{path,command}` " + "function instead." + .format(self.__class__.__name__)) + + @staticmethod + def for_path(*path): + """Create a FlightDescriptor for a resource path.""" + cdef FlightDescriptor result = \ + FlightDescriptor.__new__(FlightDescriptor) + result.descriptor.type = CDescriptorTypePath + result.descriptor.path = [tobytes(p) for p in path] + return result + + @staticmethod + def for_command(command): + """Create a FlightDescriptor for an opaque command.""" + cdef FlightDescriptor result = \ + FlightDescriptor.__new__(FlightDescriptor) + result.descriptor.type = CDescriptorTypeCmd + result.descriptor.cmd = tobytes(command) + return result + + @property + def descriptor_type(self): + """Get the type of this descriptor.""" + if self.descriptor.type == CDescriptorTypeUnknown: + return DescriptorType.UNKNOWN + elif self.descriptor.type == CDescriptorTypePath: + return DescriptorType.PATH + elif self.descriptor.type == CDescriptorTypeCmd: + return DescriptorType.CMD + raise RuntimeError("Invalid descriptor type!") + + @property + def command(self): + """Get the command for this descriptor.""" + if self.descriptor_type != DescriptorType.CMD: + return None + return self.descriptor.cmd + + @property + def path(self): + """Get the path for this descriptor.""" + if self.descriptor_type != DescriptorType.PATH: + return None + return self.descriptor.path + + def __repr__(self): + if self.descriptor_type == DescriptorType.PATH: + return f"" + elif self.descriptor_type == DescriptorType.CMD: + return f"" + else: + return "" + + @staticmethod + cdef CFlightDescriptor unwrap(descriptor) except *: + if not isinstance(descriptor, FlightDescriptor): + raise TypeError("Must provide a FlightDescriptor, not '{}'".format( + type(descriptor))) + return ( descriptor).descriptor + + def serialize(self): + """Get the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + return GetResultValue(self.descriptor.SerializeToString()) + + @classmethod + def deserialize(cls, serialized): + """Parse the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + cdef FlightDescriptor descriptor = \ + FlightDescriptor.__new__(FlightDescriptor) + descriptor.descriptor = GetResultValue( + CFlightDescriptor.Deserialize(tobytes(serialized))) + return descriptor + + def __eq__(self, FlightDescriptor other): + return self.descriptor == other.descriptor + + +cdef class Ticket(_Weakrefable): + """A ticket for requesting a Flight stream.""" + + cdef: + CTicket c_ticket + + def __init__(self, ticket): + self.c_ticket.ticket = tobytes(ticket) + + @property + def ticket(self): + return self.c_ticket.ticket + + def serialize(self): + """Get the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + return GetResultValue(self.c_ticket.SerializeToString()) + + @classmethod + def deserialize(cls, serialized): + """Parse the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + cdef Ticket ticket = Ticket.__new__(Ticket) + ticket.c_ticket = GetResultValue( + CTicket.Deserialize(tobytes(serialized))) + return ticket + + def __eq__(self, Ticket other): + return self.c_ticket == other.c_ticket + + def __repr__(self): + return f"" + + +cdef class Location(_Weakrefable): + """The location of a Flight service.""" + cdef: + CLocation location + + def __init__(self, uri): + check_flight_status(CLocation.Parse(tobytes(uri)).Value(&self.location)) + + def __repr__(self): + return f'' + + @property + def uri(self): + return self.location.ToString() + + def equals(self, Location other): + return self == other + + def __eq__(self, other): + if not isinstance(other, Location): + return NotImplemented + return self.location.Equals(( other).location) + + @staticmethod + def for_grpc_tcp(host, port): + """Create a Location for a TCP-based gRPC service.""" + cdef: + c_string c_host = tobytes(host) + int c_port = port + Location result = Location.__new__(Location) + check_flight_status( + CLocation.ForGrpcTcp(c_host, c_port).Value(&result.location)) + return result + + @staticmethod + def for_grpc_tls(host, port): + """Create a Location for a TLS-based gRPC service.""" + cdef: + c_string c_host = tobytes(host) + int c_port = port + Location result = Location.__new__(Location) + check_flight_status( + CLocation.ForGrpcTls(c_host, c_port).Value(&result.location)) + return result + + @staticmethod + def for_grpc_unix(path): + """Create a Location for a domain socket-based gRPC service.""" + cdef: + c_string c_path = tobytes(path) + Location result = Location.__new__(Location) + check_flight_status(CLocation.ForGrpcUnix(c_path).Value(&result.location)) + return result + + @staticmethod + cdef Location wrap(CLocation location): + cdef Location result = Location.__new__(Location) + result.location = location + return result + + @staticmethod + cdef CLocation unwrap(object location) except *: + cdef CLocation c_location + if isinstance(location, str): + check_flight_status( + CLocation.Parse(tobytes(location)).Value(&c_location)) + return c_location + elif not isinstance(location, Location): + raise TypeError("Must provide a Location, not '{}'".format( + type(location))) + return ( location).location + + +cdef class FlightEndpoint(_Weakrefable): + """A Flight stream, along with the ticket and locations to access it.""" + cdef: + CFlightEndpoint endpoint + + def __init__(self, ticket, locations): + """Create a FlightEndpoint from a ticket and list of locations. + + Parameters + ---------- + ticket : Ticket or bytes + the ticket needed to access this flight + locations : list of string URIs + locations where this flight is available + + Raises + ------ + ArrowException + If one of the location URIs is not a valid URI. + """ + cdef: + CLocation c_location + + if isinstance(ticket, Ticket): + self.endpoint.ticket.ticket = tobytes(ticket.ticket) + else: + self.endpoint.ticket.ticket = tobytes(ticket) + + for location in locations: + if isinstance(location, Location): + c_location = ( location).location + else: + c_location = CLocation() + check_flight_status( + CLocation.Parse(tobytes(location)).Value(&c_location)) + self.endpoint.locations.push_back(c_location) + + @property + def ticket(self): + """Get the ticket in this endpoint.""" + return Ticket(self.endpoint.ticket.ticket) + + @property + def locations(self): + return [Location.wrap(location) + for location in self.endpoint.locations] + + def serialize(self): + """Get the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + return GetResultValue(self.endpoint.SerializeToString()) + + @classmethod + def deserialize(cls, serialized): + """Parse the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + cdef FlightEndpoint endpoint = FlightEndpoint.__new__(FlightEndpoint) + endpoint.endpoint = GetResultValue( + CFlightEndpoint.Deserialize(tobytes(serialized))) + return endpoint + + def __repr__(self): + return (f"") + + def __eq__(self, FlightEndpoint other): + return self.endpoint == other.endpoint + + +cdef class SchemaResult(_Weakrefable): + """The serialized schema returned from a GetSchema request.""" + cdef: + unique_ptr[CSchemaResult] result + + def __init__(self, Schema schema): + """Create a SchemaResult from a schema. + + Parameters + ---------- + schema: Schema + the schema of the data in this flight. + """ + cdef: + shared_ptr[CSchema] c_schema = pyarrow_unwrap_schema(schema) + check_flight_status(CreateSchemaResult(c_schema, &self.result)) + + @property + def schema(self): + """The schema of the data in this flight.""" + cdef: + shared_ptr[CSchema] schema + CDictionaryMemo dummy_memo + + check_flight_status(self.result.get().GetSchema(&dummy_memo).Value(&schema)) + return pyarrow_wrap_schema(schema) + + def serialize(self): + """Get the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + return GetResultValue(self.result.get().SerializeToString()) + + @classmethod + def deserialize(cls, serialized): + """Parse the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + cdef SchemaResult result = SchemaResult.__new__(SchemaResult) + result.result.reset(new CSchemaResult(GetResultValue( + CSchemaResult.Deserialize(tobytes(serialized))))) + return result + + def __eq__(self, SchemaResult other): + return deref(self.result.get()) == deref(other.result.get()) + + def __repr__(self): + return f"" + + +cdef class FlightInfo(_Weakrefable): + """A description of a Flight stream.""" + cdef: + unique_ptr[CFlightInfo] info + + @staticmethod + cdef wrap(CFlightInfo c_info): + cdef FlightInfo obj = FlightInfo.__new__(FlightInfo) + obj.info.reset(new CFlightInfo(move(c_info))) + return obj + + def __init__(self, Schema schema, FlightDescriptor descriptor, endpoints, + total_records, total_bytes): + """Create a FlightInfo object from a schema, descriptor, and endpoints. + + Parameters + ---------- + schema : Schema + the schema of the data in this flight. + descriptor : FlightDescriptor + the descriptor for this flight. + endpoints : list of FlightEndpoint + a list of endpoints where this flight is available. + total_records : int + the total records in this flight, or -1 if unknown + total_bytes : int + the total bytes in this flight, or -1 if unknown + """ + cdef: + shared_ptr[CSchema] c_schema = pyarrow_unwrap_schema(schema) + vector[CFlightEndpoint] c_endpoints + + for endpoint in endpoints: + if isinstance(endpoint, FlightEndpoint): + c_endpoints.push_back(( endpoint).endpoint) + else: + raise TypeError('Endpoint {} is not instance of' + ' FlightEndpoint'.format(endpoint)) + + check_flight_status(CreateFlightInfo(c_schema, + descriptor.descriptor, + c_endpoints, + total_records, + total_bytes, &self.info)) + + @property + def total_records(self): + """The total record count of this flight, or -1 if unknown.""" + return self.info.get().total_records() + + @property + def total_bytes(self): + """The size in bytes of the data in this flight, or -1 if unknown.""" + return self.info.get().total_bytes() + + @property + def schema(self): + """The schema of the data in this flight.""" + cdef: + shared_ptr[CSchema] schema + CDictionaryMemo dummy_memo + + check_flight_status(self.info.get().GetSchema(&dummy_memo).Value(&schema)) + return pyarrow_wrap_schema(schema) + + @property + def descriptor(self): + """The descriptor of the data in this flight.""" + cdef FlightDescriptor result = \ + FlightDescriptor.__new__(FlightDescriptor) + result.descriptor = self.info.get().descriptor() + return result + + @property + def endpoints(self): + """The endpoints where this flight is available.""" + # TODO: get Cython to iterate over reference directly + cdef: + vector[CFlightEndpoint] endpoints = self.info.get().endpoints() + FlightEndpoint py_endpoint + + result = [] + for endpoint in endpoints: + py_endpoint = FlightEndpoint.__new__(FlightEndpoint) + py_endpoint.endpoint = endpoint + result.append(py_endpoint) + return result + + def serialize(self): + """Get the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + return GetResultValue(self.info.get().SerializeToString()) + + @classmethod + def deserialize(cls, serialized): + """Parse the wire-format representation of this type. + + Useful when interoperating with non-Flight systems (e.g. REST + services) that may want to return Flight types. + + """ + cdef FlightInfo info = FlightInfo.__new__(FlightInfo) + info.info = move(GetResultValue( + CFlightInfo.Deserialize(tobytes(serialized)))) + return info + + def __eq__(self, FlightInfo other): + return deref(self.info.get()) == deref(other.info.get()) + + def __repr__(self): + return (f"") + + +cdef class FlightStreamChunk(_Weakrefable): + """A RecordBatch with application metadata on the side.""" + cdef: + CFlightStreamChunk chunk + + @property + def data(self): + if self.chunk.data == NULL: + return None + return pyarrow_wrap_batch(self.chunk.data) + + @property + def app_metadata(self): + if self.chunk.app_metadata == NULL: + return None + return pyarrow_wrap_buffer(self.chunk.app_metadata) + + def __iter__(self): + return iter((self.data, self.app_metadata)) + + def __repr__(self): + return "".format( + self.chunk.data != NULL, self.chunk.app_metadata != NULL) + + +cdef class _MetadataRecordBatchReader(_Weakrefable, _ReadPandasMixin): + """A reader for Flight streams.""" + + # Needs to be separate class so the "real" class can subclass the + # pure-Python mixin class + + cdef dict __dict__ + cdef shared_ptr[CMetadataRecordBatchReader] reader + + def __iter__(self): + return self + + def __next__(self): + return self.read_chunk() + + @property + def schema(self): + """Get the schema for this reader.""" + cdef shared_ptr[CSchema] c_schema + with nogil: + check_flight_status(self.reader.get().GetSchema().Value(&c_schema)) + return pyarrow_wrap_schema(c_schema) + + def read_all(self): + """Read the entire contents of the stream as a Table.""" + cdef: + shared_ptr[CTable] c_table + with nogil: + check_flight_status(self.reader.get().ToTable().Value(&c_table)) + return pyarrow_wrap_table(c_table) + + def read_chunk(self): + """Read the next FlightStreamChunk along with any metadata. + + Returns + ------- + chunk : FlightStreamChunk + The next FlightStreamChunk in the stream. + + Raises + ------ + StopIteration + when the stream is finished + """ + cdef: + FlightStreamChunk chunk = FlightStreamChunk() + + with nogil: + check_flight_status(self.reader.get().Next().Value(&chunk.chunk)) + + if chunk.chunk.data == NULL and chunk.chunk.app_metadata == NULL: + raise StopIteration + + return chunk + + def to_reader(self): + """Convert this reader into a regular RecordBatchReader. + + This may fail if the schema cannot be read from the remote end. + + Returns + ------- + RecordBatchReader + """ + cdef RecordBatchReader reader + reader = RecordBatchReader.__new__(RecordBatchReader) + with nogil: + reader.reader = GetResultValue(MakeRecordBatchReader(self.reader)) + + return reader + + +cdef class MetadataRecordBatchReader(_MetadataRecordBatchReader): + """The base class for readers for Flight streams. + + See Also + -------- + FlightStreamReader + """ + + +cdef class FlightStreamReader(MetadataRecordBatchReader): + """A reader that can also be canceled.""" + + def cancel(self): + """Cancel the read operation.""" + with nogil: + ( self.reader.get()).Cancel() + + def read_all(self): + """Read the entire contents of the stream as a Table.""" + cdef: + shared_ptr[CTable] c_table + CStopToken stop_token + with SignalStopHandler() as stop_handler: + stop_token = ( stop_handler.stop_token).stop_token + with nogil: + check_flight_status( + ( self.reader.get()) + .ToTableWithStopToken(stop_token).Value(&c_table)) + return pyarrow_wrap_table(c_table) + + +cdef class MetadataRecordBatchWriter(_CRecordBatchWriter): + """A RecordBatchWriter that also allows writing application metadata. + + This class is a context manager; on exit, close() will be called. + """ + + cdef CMetadataRecordBatchWriter* _writer(self) nogil: + return self.writer.get() + + def begin(self, schema: Schema, options=None): + """Prepare to write data to this stream with the given schema.""" + cdef: + shared_ptr[CSchema] c_schema = pyarrow_unwrap_schema(schema) + CIpcWriteOptions c_options = _get_options(options).c_options + with nogil: + check_flight_status(self._writer().Begin(c_schema, c_options)) + + def write_metadata(self, buf): + """Write Flight metadata by itself.""" + cdef shared_ptr[CBuffer] c_buf = pyarrow_unwrap_buffer(as_buffer(buf)) + with nogil: + check_flight_status( + self._writer().WriteMetadata(c_buf)) + + def write_batch(self, RecordBatch batch): + """ + Write RecordBatch to stream. + + Parameters + ---------- + batch : RecordBatch + """ + cdef: + shared_ptr[const CKeyValueMetadata] custom_metadata + + # Override superclass method to use check_flight_status so we + # can generate FlightWriteSizeExceededError. We don't do this + # for write_table as callers who intend to handle the error + # and retry with a smaller batch should be working with + # individual batches to have control. + + with nogil: + check_flight_status( + self._writer().WriteRecordBatch(deref(batch.batch), custom_metadata)) + + def write_table(self, Table table, max_chunksize=None, **kwargs): + """ + Write Table to stream in (contiguous) RecordBatch objects. + + Parameters + ---------- + table : Table + max_chunksize : int, default None + Maximum number of rows for RecordBatch chunks. Individual chunks may + be smaller depending on the chunk layout of individual columns. + """ + cdef: + # max_chunksize must be > 0 to have any impact + int64_t c_max_chunksize = -1 + + if 'chunksize' in kwargs: + max_chunksize = kwargs['chunksize'] + msg = ('The parameter chunksize is deprecated for the write_table ' + 'methods as of 0.15, please use parameter ' + 'max_chunksize instead') + warnings.warn(msg, FutureWarning) + + if max_chunksize is not None: + c_max_chunksize = max_chunksize + + with nogil: + check_flight_status( + self._writer().WriteTable(table.table[0], c_max_chunksize)) + + def close(self): + """ + Close stream and write end-of-stream 0 marker. + """ + with nogil: + check_flight_status(self._writer().Close()) + + def write_with_metadata(self, RecordBatch batch, buf): + """Write a RecordBatch along with Flight metadata. + + Parameters + ---------- + batch : RecordBatch + The next RecordBatch in the stream. + buf : Buffer + Application-specific metadata for the batch as defined by + Flight. + """ + cdef shared_ptr[CBuffer] c_buf = pyarrow_unwrap_buffer(as_buffer(buf)) + with nogil: + check_flight_status( + self._writer().WriteWithMetadata(deref(batch.batch), c_buf)) + + +cdef class FlightStreamWriter(MetadataRecordBatchWriter): + """A writer that also allows closing the write side of a stream.""" + + def done_writing(self): + """Indicate that the client is done writing, but not done reading.""" + with nogil: + check_flight_status( + ( self.writer.get()).DoneWriting()) + + +cdef class FlightMetadataReader(_Weakrefable): + """A reader for Flight metadata messages sent during a DoPut.""" + + cdef: + unique_ptr[CFlightMetadataReader] reader + + def read(self): + """Read the next metadata message.""" + cdef shared_ptr[CBuffer] buf + with nogil: + check_flight_status(self.reader.get().ReadMetadata(&buf)) + if buf == NULL: + return None + return pyarrow_wrap_buffer(buf) + + +cdef class FlightMetadataWriter(_Weakrefable): + """A sender for Flight metadata messages during a DoPut.""" + + cdef: + unique_ptr[CFlightMetadataWriter] writer + + def write(self, message): + """Write the next metadata message. + + Parameters + ---------- + message : Buffer + """ + cdef shared_ptr[CBuffer] buf = \ + pyarrow_unwrap_buffer(as_buffer(message)) + with nogil: + check_flight_status(self.writer.get().WriteMetadata(deref(buf))) + + +class AsyncioCall: + """State for an async RPC using asyncio.""" + + def __init__(self) -> None: + import asyncio + self._future = asyncio.get_running_loop().create_future() + + def as_awaitable(self) -> object: + return self._future + + def wakeup(self, result_or_exception) -> None: + # Mark the Future done from within its loop (asyncio + # objects are generally not thread-safe) + loop = self._future.get_loop() + if isinstance(result_or_exception, BaseException): + loop.call_soon_threadsafe( + self._future.set_exception, result_or_exception) + else: + loop.call_soon_threadsafe( + self._future.set_result, result_or_exception) + + +cdef class AsyncioFlightClient: + """ + A FlightClient with an asyncio-based async interface. + + This interface is EXPERIMENTAL. + """ + + cdef: + FlightClient _client + + def __init__(self, FlightClient client) -> None: + self._client = client + + async def get_flight_info( + self, + descriptor: FlightDescriptor, + *, + options: FlightCallOptions = None, + ): + call = AsyncioCall() + self._get_flight_info(call, descriptor, options) + return await call.as_awaitable() + + cdef _get_flight_info(self, call, descriptor, options): + cdef: + CFlightCallOptions* c_options = \ + FlightCallOptions.unwrap(options) + CFlightDescriptor c_descriptor = \ + FlightDescriptor.unwrap(descriptor) + CFuture[CFlightInfo] c_future + + with nogil: + c_future = self._client.client.get().GetFlightInfoAsync( + deref(c_options), c_descriptor) + + BindFuture(move(c_future), call.wakeup, FlightInfo.wrap) + + +cdef class FlightClient(_Weakrefable): + """A client to a Flight service. + + Connect to a Flight service on the given host and port. + + Parameters + ---------- + location : str, tuple or Location + Location to connect to. Either a gRPC URI like `grpc://localhost:port`, + a tuple of (host, port) pair, or a Location instance. + tls_root_certs : bytes or None + PEM-encoded + cert_chain: bytes or None + Client certificate if using mutual TLS + private_key: bytes or None + Client private key for cert_chain is using mutual TLS + override_hostname : str or None + Override the hostname checked by TLS. Insecure, use with caution. + middleware : list optional, default None + A list of ClientMiddlewareFactory instances. + write_size_limit_bytes : int optional, default None + A soft limit on the size of a data payload sent to the + server. Enabled if positive. If enabled, writing a record + batch that (when serialized) exceeds this limit will raise an + exception; the client can retry the write with a smaller + batch. + disable_server_verification : boolean optional, default False + A flag that indicates that, if the client is connecting + with TLS, that it skips server verification. If this is + enabled, all other TLS settings are overridden. + generic_options : list optional, default None + A list of generic (string, int or string) option tuples passed + to the underlying transport. Effect is implementation + dependent. + """ + cdef: + unique_ptr[CFlightClient] client + + def __init__(self, location, *, tls_root_certs=None, cert_chain=None, + private_key=None, override_hostname=None, middleware=None, + write_size_limit_bytes=None, + disable_server_verification=None, generic_options=None): + if isinstance(location, (bytes, str)): + location = Location(location) + elif isinstance(location, tuple): + host, port = location + if tls_root_certs or disable_server_verification is not None: + location = Location.for_grpc_tls(host, port) + else: + location = Location.for_grpc_tcp(host, port) + elif not isinstance(location, Location): + raise TypeError('`location` argument must be a string, tuple or a ' + 'Location instance') + self.init(location, tls_root_certs, cert_chain, private_key, + override_hostname, middleware, write_size_limit_bytes, + disable_server_verification, generic_options) + + cdef init(self, Location location, tls_root_certs, cert_chain, + private_key, override_hostname, middleware, + write_size_limit_bytes, disable_server_verification, + generic_options): + cdef: + CLocation c_location = Location.unwrap(location) + CFlightClientOptions c_options = CFlightClientOptions.Defaults() + function[cb_client_middleware_start_call] start_call = \ + &_client_middleware_start_call + CIntStringVariant variant + + if tls_root_certs: + c_options.tls_root_certs = tobytes(tls_root_certs) + if cert_chain: + c_options.cert_chain = tobytes(cert_chain) + if private_key: + c_options.private_key = tobytes(private_key) + if override_hostname: + c_options.override_hostname = tobytes(override_hostname) + if disable_server_verification is not None: + c_options.disable_server_verification = disable_server_verification + if middleware: + for factory in middleware: + c_options.middleware.push_back( + + make_shared[CPyClientMiddlewareFactory]( + factory, start_call)) + if write_size_limit_bytes is not None: + c_options.write_size_limit_bytes = write_size_limit_bytes + else: + c_options.write_size_limit_bytes = 0 + if generic_options: + for key, value in generic_options: + if isinstance(value, (str, bytes)): + variant = CIntStringVariant( tobytes(value)) + else: + variant = CIntStringVariant( value) + c_options.generic_options.push_back( + pair[c_string, CIntStringVariant](tobytes(key), variant)) + + with nogil: + check_flight_status(CFlightClient.Connect(c_location, c_options + ).Value(&self.client)) + + @property + def supports_async(self): + return self.client.get().supports_async() + + def as_async(self) -> None: + check_status(self.client.get().CheckAsyncSupport()) + return AsyncioFlightClient(self) + + def wait_for_available(self, timeout=5): + """Block until the server can be contacted. + + Parameters + ---------- + timeout : int, default 5 + The maximum seconds to wait. + """ + deadline = time.time() + timeout + while True: + try: + list(self.list_flights()) + except FlightUnavailableError: + if time.time() < deadline: + time.sleep(0.025) + continue + else: + raise + except NotImplementedError: + # allow if list_flights is not implemented, because + # the server can be contacted nonetheless + break + else: + break + + @classmethod + def connect(cls, location, tls_root_certs=None, cert_chain=None, + private_key=None, override_hostname=None, + disable_server_verification=None): + """Connect to a Flight server. + + .. deprecated:: 0.15.0 + Use the ``FlightClient`` constructor or ``pyarrow.flight.connect`` function instead. + """ + warnings.warn("The 'FlightClient.connect' method is deprecated, use " + "FlightClient constructor or pyarrow.flight.connect " + "function instead") + return FlightClient( + location, tls_root_certs=tls_root_certs, + cert_chain=cert_chain, private_key=private_key, + override_hostname=override_hostname, + disable_server_verification=disable_server_verification + ) + + def authenticate(self, auth_handler, options: FlightCallOptions = None): + """Authenticate to the server. + + Parameters + ---------- + auth_handler : ClientAuthHandler + The authentication mechanism to use. + options : FlightCallOptions + Options for this call. + """ + cdef: + unique_ptr[CClientAuthHandler] handler + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + + if not isinstance(auth_handler, ClientAuthHandler): + raise TypeError( + "FlightClient.authenticate takes a ClientAuthHandler, " + "not '{}'".format(type(auth_handler))) + handler.reset(( auth_handler).to_handler()) + with nogil: + check_flight_status( + self.client.get().Authenticate(deref(c_options), + move(handler))) + + def authenticate_basic_token(self, username, password, + options: FlightCallOptions = None): + """Authenticate to the server with HTTP basic authentication. + + Parameters + ---------- + username : string + Username to authenticate with + password : string + Password to authenticate with + options : FlightCallOptions + Options for this call + + Returns + ------- + tuple : Tuple[str, str] + A tuple representing the FlightCallOptions authorization + header entry of a bearer token. + """ + cdef: + CResult[pair[c_string, c_string]] result + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + c_string user = tobytes(username) + c_string pw = tobytes(password) + + with nogil: + result = self.client.get().AuthenticateBasicToken(deref(c_options), + user, pw) + check_flight_status(result.status()) + + return GetResultValue(result) + + def list_actions(self, options: FlightCallOptions = None): + """List the actions available on a service.""" + cdef: + vector[CActionType] results + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + + with SignalStopHandler() as stop_handler: + c_options.stop_token = \ + ( stop_handler.stop_token).stop_token + with nogil: + check_flight_status( + self.client.get().ListActions(deref(c_options)).Value(&results)) + + result = [] + for action_type in results: + py_action = ActionType(frombytes(action_type.type), + frombytes(action_type.description)) + result.append(py_action) + + return result + + def do_action(self, action, options: FlightCallOptions = None): + """ + Execute an action on a service. + + Parameters + ---------- + action : str, tuple, or Action + Can be action type name (no body), type and body, or any Action + object + options : FlightCallOptions + RPC options + + Returns + ------- + results : iterator of Result values + """ + cdef: + unique_ptr[CResultStream] results + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + + if isinstance(action, (str, bytes)): + action = Action(action, b'') + elif isinstance(action, tuple): + action = Action(*action) + elif not isinstance(action, Action): + raise TypeError("Action must be Action instance, string, or tuple") + + cdef CAction c_action = Action.unwrap( action) + with nogil: + check_flight_status( + self.client.get().DoAction( + deref(c_options), c_action).Value(&results)) + + def _do_action_response(): + cdef: + Result result + while True: + result = Result.__new__(Result) + with nogil: + check_flight_status(results.get().Next().Value(&result.result)) + if result.result == NULL: + break + yield result + return _do_action_response() + + def list_flights(self, criteria: bytes = None, + options: FlightCallOptions = None): + """List the flights available on a service.""" + cdef: + unique_ptr[CFlightListing] listing + FlightInfo result + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + CCriteria c_criteria + + if criteria: + c_criteria.expression = tobytes(criteria) + + with SignalStopHandler() as stop_handler: + c_options.stop_token = \ + ( stop_handler.stop_token).stop_token + with nogil: + check_flight_status( + self.client.get().ListFlights(deref(c_options), + c_criteria).Value(&listing)) + + while True: + result = FlightInfo.__new__(FlightInfo) + with nogil: + check_flight_status(listing.get().Next().Value(&result.info)) + if result.info == NULL: + break + yield result + + def get_flight_info(self, descriptor: FlightDescriptor, + options: FlightCallOptions = None): + """Request information about an available flight.""" + cdef: + FlightInfo result = FlightInfo.__new__(FlightInfo) + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + CFlightDescriptor c_descriptor = \ + FlightDescriptor.unwrap(descriptor) + + with nogil: + check_flight_status(self.client.get().GetFlightInfo( + deref(c_options), c_descriptor).Value(&result.info)) + + return result + + def get_schema(self, descriptor: FlightDescriptor, + options: FlightCallOptions = None): + """Request schema for an available flight.""" + cdef: + SchemaResult result = SchemaResult.__new__(SchemaResult) + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + CFlightDescriptor c_descriptor = \ + FlightDescriptor.unwrap(descriptor) + with nogil: + check_status( + self.client.get() + .GetSchema(deref(c_options), c_descriptor).Value(&result.result) + ) + + return result + + def do_get(self, ticket: Ticket, options: FlightCallOptions = None): + """Request the data for a flight. + + Returns + ------- + reader : FlightStreamReader + """ + cdef: + unique_ptr[CFlightStreamReader] reader + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + + with nogil: + check_flight_status( + self.client.get().DoGet( + deref(c_options), ticket.c_ticket).Value(&reader)) + result = FlightStreamReader() + result.reader.reset(reader.release()) + return result + + def do_put(self, descriptor: FlightDescriptor, Schema schema not None, + options: FlightCallOptions = None): + """Upload data to a flight. + + Returns + ------- + writer : FlightStreamWriter + reader : FlightMetadataReader + """ + cdef: + shared_ptr[CSchema] c_schema = pyarrow_unwrap_schema(schema) + CDoPutResult c_do_put_result + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + CFlightDescriptor c_descriptor = \ + FlightDescriptor.unwrap(descriptor) + + with nogil: + check_flight_status(self.client.get().DoPut( + deref(c_options), + c_descriptor, + c_schema).Value(&c_do_put_result)) + py_writer = FlightStreamWriter() + py_writer.writer.reset(c_do_put_result.writer.release()) + py_reader = FlightMetadataReader() + py_reader.reader.reset(c_do_put_result.reader.release()) + return py_writer, py_reader + + def do_exchange(self, descriptor: FlightDescriptor, + options: FlightCallOptions = None): + """Start a bidirectional data exchange with a server. + + Parameters + ---------- + descriptor : FlightDescriptor + A descriptor for the flight. + options : FlightCallOptions + RPC options. + + Returns + ------- + writer : FlightStreamWriter + reader : FlightStreamReader + """ + cdef: + CDoExchangeResult c_do_exchange_result + CFlightCallOptions* c_options = FlightCallOptions.unwrap(options) + CFlightDescriptor c_descriptor = \ + FlightDescriptor.unwrap(descriptor) + + with nogil: + check_flight_status(self.client.get().DoExchange( + deref(c_options), + c_descriptor).Value(&c_do_exchange_result)) + py_writer = FlightStreamWriter() + py_writer.writer.reset(c_do_exchange_result.writer.release()) + py_reader = FlightStreamReader() + py_reader.reader.reset(c_do_exchange_result.reader.release()) + return py_writer, py_reader + + def close(self): + """Close the client and disconnect.""" + client = self.client.get() + if client != NULL: + check_flight_status(client.Close()) + + def __del__(self): + # Not ideal, but close() wasn't originally present so + # applications may not be calling it + self.close() + + def __enter__(self): + return self + + def __exit__(self, exc_type, exc_value, traceback): + self.close() + + +cdef class FlightDataStream(_Weakrefable): + """ + Abstract base class for Flight data streams. + + See Also + -------- + RecordBatchStream + GeneratorStream + """ + + cdef CFlightDataStream* to_stream(self) except *: + """Create the C++ data stream for the backing Python object. + + We don't expose the C++ object to Python, so we can manage its + lifetime from the Cython/C++ side. + """ + raise NotImplementedError + + +cdef class RecordBatchStream(FlightDataStream): + """A Flight data stream backed by RecordBatches. + + The remainder of this DoGet request will be handled in C++, + without having to acquire the GIL. + + """ + cdef: + object data_source + CIpcWriteOptions write_options + + def __init__(self, data_source, options=None): + """Create a RecordBatchStream from a data source. + + Parameters + ---------- + data_source : RecordBatchReader or Table + The data to stream to the client. + options : pyarrow.ipc.IpcWriteOptions, optional + Optional IPC options to control how to write the data. + """ + if (not isinstance(data_source, RecordBatchReader) and + not isinstance(data_source, lib.Table)): + raise TypeError("Expected RecordBatchReader or Table, " + "but got: {}".format(type(data_source))) + self.data_source = data_source + self.write_options = _get_options(options).c_options + + cdef CFlightDataStream* to_stream(self) except *: + cdef: + shared_ptr[CRecordBatchReader] reader + if isinstance(self.data_source, RecordBatchReader): + reader = ( self.data_source).reader + elif isinstance(self.data_source, lib.Table): + table = (
self.data_source).table + reader.reset(new TableBatchReader(deref(table))) + else: + raise RuntimeError("Can't construct RecordBatchStream " + "from type {}".format(type(self.data_source))) + return new CRecordBatchStream(reader, self.write_options) + + +cdef class GeneratorStream(FlightDataStream): + """A Flight data stream backed by a Python generator.""" + cdef: + shared_ptr[CSchema] schema + object generator + # A substream currently being consumed by the client, if + # present. Produced by the generator. + unique_ptr[CFlightDataStream] current_stream + CIpcWriteOptions c_options + + def __init__(self, schema, generator, options=None): + """Create a GeneratorStream from a Python generator. + + Parameters + ---------- + schema : Schema + The schema for the data to be returned. + + generator : iterator or iterable + The generator should yield other FlightDataStream objects, + Tables, RecordBatches, or RecordBatchReaders. + + options : pyarrow.ipc.IpcWriteOptions, optional + """ + self.schema = pyarrow_unwrap_schema(schema) + self.generator = iter(generator) + self.c_options = _get_options(options).c_options + + cdef CFlightDataStream* to_stream(self) except *: + cdef: + function[cb_data_stream_next] callback = &_data_stream_next + return new CPyGeneratorFlightDataStream(self, self.schema, callback, + self.c_options) + + +cdef class ServerCallContext(_Weakrefable): + """Per-call state/context.""" + cdef: + const CServerCallContext* context + + def peer_identity(self): + """Get the identity of the authenticated peer. + + May be the empty string. + """ + return tobytes(self.context.peer_identity()) + + def peer(self): + """Get the address of the peer.""" + # Set safe=True as gRPC on Windows sometimes gives garbage bytes + return frombytes(self.context.peer(), safe=True) + + def is_cancelled(self): + """Check if the current RPC call has been canceled by the client.""" + return self.context.is_cancelled() + + def add_header(self, key, value): + """Add a response header.""" + self.context.AddHeader(tobytes(key), tobytes(value)) + + def add_trailer(self, key, value): + """Add a response trailer.""" + self.context.AddTrailer(tobytes(key), tobytes(value)) + + def get_middleware(self, key): + """ + Get a middleware instance by key. + + Returns None if the middleware was not found. + """ + cdef: + CServerMiddleware* c_middleware = \ + self.context.GetMiddleware(CPyServerMiddlewareName) + CPyServerMiddleware* middleware + vector[CTracingServerMiddlewareTraceKey] c_trace_context + if c_middleware == NULL: + c_middleware = self.context.GetMiddleware(tobytes(key)) + + if c_middleware == NULL: + return None + elif c_middleware.name() == CPyServerMiddlewareName: + middleware = c_middleware + py_middleware = <_ServerMiddlewareWrapper> middleware.py_object() + return py_middleware.middleware.get(key) + elif c_middleware.name() == CTracingServerMiddlewareName: + c_trace_context = ( c_middleware + ).GetTraceContext() + trace_context = {pair.key: pair.value for pair in c_trace_context} + return TracingServerMiddleware(trace_context) + return None + + @staticmethod + cdef ServerCallContext wrap(const CServerCallContext& context): + cdef ServerCallContext result = \ + ServerCallContext.__new__(ServerCallContext) + result.context = &context + return result + + +cdef class ServerAuthReader(_Weakrefable): + """A reader for messages from the client during an auth handshake.""" + cdef: + CServerAuthReader* reader + + def read(self): + cdef c_string token + if not self.reader: + raise ValueError("Cannot use ServerAuthReader outside " + "ServerAuthHandler.authenticate") + with nogil: + check_flight_status(self.reader.Read(&token)) + return token + + cdef void poison(self): + """Prevent further usage of this object. + + This object is constructed by taking a pointer to a reference, + so we want to make sure Python users do not access this after + the reference goes away. + """ + self.reader = NULL + + @staticmethod + cdef ServerAuthReader wrap(CServerAuthReader* reader): + cdef ServerAuthReader result = \ + ServerAuthReader.__new__(ServerAuthReader) + result.reader = reader + return result + + +cdef class ServerAuthSender(_Weakrefable): + """A writer for messages to the client during an auth handshake.""" + cdef: + CServerAuthSender* sender + + def write(self, message): + cdef c_string c_message = tobytes(message) + if not self.sender: + raise ValueError("Cannot use ServerAuthSender outside " + "ServerAuthHandler.authenticate") + with nogil: + check_flight_status(self.sender.Write(c_message)) + + cdef void poison(self): + """Prevent further usage of this object. + + This object is constructed by taking a pointer to a reference, + so we want to make sure Python users do not access this after + the reference goes away. + """ + self.sender = NULL + + @staticmethod + cdef ServerAuthSender wrap(CServerAuthSender* sender): + cdef ServerAuthSender result = \ + ServerAuthSender.__new__(ServerAuthSender) + result.sender = sender + return result + + +cdef class ClientAuthReader(_Weakrefable): + """A reader for messages from the server during an auth handshake.""" + cdef: + CClientAuthReader* reader + + def read(self): + cdef c_string token + if not self.reader: + raise ValueError("Cannot use ClientAuthReader outside " + "ClientAuthHandler.authenticate") + with nogil: + check_flight_status(self.reader.Read(&token)) + return token + + cdef void poison(self): + """Prevent further usage of this object. + + This object is constructed by taking a pointer to a reference, + so we want to make sure Python users do not access this after + the reference goes away. + """ + self.reader = NULL + + @staticmethod + cdef ClientAuthReader wrap(CClientAuthReader* reader): + cdef ClientAuthReader result = \ + ClientAuthReader.__new__(ClientAuthReader) + result.reader = reader + return result + + +cdef class ClientAuthSender(_Weakrefable): + """A writer for messages to the server during an auth handshake.""" + cdef: + CClientAuthSender* sender + + def write(self, message): + cdef c_string c_message = tobytes(message) + if not self.sender: + raise ValueError("Cannot use ClientAuthSender outside " + "ClientAuthHandler.authenticate") + with nogil: + check_flight_status(self.sender.Write(c_message)) + + cdef void poison(self): + """Prevent further usage of this object. + + This object is constructed by taking a pointer to a reference, + so we want to make sure Python users do not access this after + the reference goes away. + """ + self.sender = NULL + + @staticmethod + cdef ClientAuthSender wrap(CClientAuthSender* sender): + cdef ClientAuthSender result = \ + ClientAuthSender.__new__(ClientAuthSender) + result.sender = sender + return result + + +cdef CStatus _data_stream_next(void* self, CFlightPayload* payload) except *: + """Callback for implementing FlightDataStream in Python.""" + cdef: + unique_ptr[CFlightDataStream] data_stream + + py_stream = self + if not isinstance(py_stream, GeneratorStream): + raise RuntimeError("self object in callback is not GeneratorStream") + stream = py_stream + + # The generator is allowed to yield a reader or table which we + # yield from; if that sub-generator is empty, we need to reset and + # try again. However, limit the number of attempts so that we + # don't just spin forever. + max_attempts = 128 + for _ in range(max_attempts): + if stream.current_stream != nullptr: + with nogil: + check_flight_status( + stream.current_stream.get().Next().Value(payload)) + # If the stream ended, see if there's another stream from the + # generator + if payload.ipc_message.metadata != nullptr: + return CStatus_OK() + stream.current_stream.reset(nullptr) + + try: + result = next(stream.generator) + except StopIteration: + payload.ipc_message.metadata.reset( nullptr) + return CStatus_OK() + except FlightError as flight_error: + return ( flight_error).to_status() + + if isinstance(result, (list, tuple)): + result, metadata = result + else: + result, metadata = result, None + + if isinstance(result, (Table, RecordBatchReader)): + if metadata: + raise ValueError("Can only return metadata alongside a " + "RecordBatch.") + result = RecordBatchStream(result) + + stream_schema = pyarrow_wrap_schema(stream.schema) + if isinstance(result, FlightDataStream): + if metadata: + raise ValueError("Can only return metadata alongside a " + "RecordBatch.") + data_stream = unique_ptr[CFlightDataStream]( + ( result).to_stream()) + substream_schema = pyarrow_wrap_schema(data_stream.get().schema()) + if substream_schema != stream_schema: + raise ValueError("Got a FlightDataStream whose schema " + "does not match the declared schema of this " + "GeneratorStream. " + "Got: {}\nExpected: {}".format( + substream_schema, stream_schema)) + stream.current_stream.reset( + new CPyFlightDataStream(result, move(data_stream))) + # Loop around and try again + continue + elif isinstance(result, RecordBatch): + batch = result + if batch.schema != stream_schema: + raise ValueError("Got a RecordBatch whose schema does not " + "match the declared schema of this " + "GeneratorStream. " + "Got: {}\nExpected: {}".format(batch.schema, + stream_schema)) + check_flight_status(GetRecordBatchPayload( + deref(batch.batch), + stream.c_options, + &payload.ipc_message)) + if metadata: + payload.app_metadata = pyarrow_unwrap_buffer( + as_buffer(metadata)) + else: + raise TypeError("GeneratorStream must be initialized with " + "an iterator of FlightDataStream, Table, " + "RecordBatch, or RecordBatchStreamReader objects, " + "not {}.".format(type(result))) + # Don't loop around + return CStatus_OK() + # Ran out of attempts (the RPC handler kept yielding empty tables/readers) + raise RuntimeError("While getting next payload, ran out of attempts to " + "get something to send " + "(application server implementation error)") + + +cdef CStatus _list_flights(void* self, const CServerCallContext& context, + const CCriteria* c_criteria, + unique_ptr[CFlightListing]* listing) except *: + """Callback for implementing ListFlights in Python.""" + cdef: + vector[CFlightInfo] flights + + try: + result = ( self).list_flights(ServerCallContext.wrap(context), + c_criteria.expression) + for info in result: + if not isinstance(info, FlightInfo): + raise TypeError("FlightServerBase.list_flights must return " + "FlightInfo instances, but got {}".format( + type(info))) + flights.push_back(deref(( info).info.get())) + listing.reset(new CSimpleFlightListing(flights)) + except FlightError as flight_error: + return ( flight_error).to_status() + return CStatus_OK() + + +cdef CStatus _get_flight_info(void* self, const CServerCallContext& context, + CFlightDescriptor c_descriptor, + unique_ptr[CFlightInfo]* info) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + FlightDescriptor py_descriptor = \ + FlightDescriptor.__new__(FlightDescriptor) + py_descriptor.descriptor = c_descriptor + try: + result = ( self).get_flight_info( + ServerCallContext.wrap(context), + py_descriptor) + except FlightError as flight_error: + return ( flight_error).to_status() + if not isinstance(result, FlightInfo): + raise TypeError("FlightServerBase.get_flight_info must return " + "a FlightInfo instance, but got {}".format( + type(result))) + info.reset(new CFlightInfo(deref(( result).info.get()))) + return CStatus_OK() + +cdef CStatus _get_schema(void* self, const CServerCallContext& context, + CFlightDescriptor c_descriptor, + unique_ptr[CSchemaResult]* info) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + FlightDescriptor py_descriptor = \ + FlightDescriptor.__new__(FlightDescriptor) + py_descriptor.descriptor = c_descriptor + result = ( self).get_schema(ServerCallContext.wrap(context), + py_descriptor) + if not isinstance(result, SchemaResult): + raise TypeError("FlightServerBase.get_schema_info must return " + "a SchemaResult instance, but got {}".format( + type(result))) + info.reset(new CSchemaResult(deref(( result).result.get()))) + return CStatus_OK() + +cdef CStatus _do_put(void* self, const CServerCallContext& context, + unique_ptr[CFlightMessageReader] reader, + unique_ptr[CFlightMetadataWriter] writer) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + MetadataRecordBatchReader py_reader = MetadataRecordBatchReader() + FlightMetadataWriter py_writer = FlightMetadataWriter() + FlightDescriptor descriptor = \ + FlightDescriptor.__new__(FlightDescriptor) + + descriptor.descriptor = reader.get().descriptor() + py_reader.reader.reset(reader.release()) + py_writer.writer.reset(writer.release()) + try: + ( self).do_put(ServerCallContext.wrap(context), descriptor, + py_reader, py_writer) + return CStatus_OK() + except FlightError as flight_error: + return ( flight_error).to_status() + + +cdef CStatus _do_get(void* self, const CServerCallContext& context, + CTicket ticket, + unique_ptr[CFlightDataStream]* stream) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + unique_ptr[CFlightDataStream] data_stream + + py_ticket = Ticket(ticket.ticket) + try: + result = ( self).do_get(ServerCallContext.wrap(context), + py_ticket) + except FlightError as flight_error: + return ( flight_error).to_status() + if not isinstance(result, FlightDataStream): + raise TypeError("FlightServerBase.do_get must return " + "a FlightDataStream") + data_stream = unique_ptr[CFlightDataStream]( + ( result).to_stream()) + stream[0] = unique_ptr[CFlightDataStream]( + new CPyFlightDataStream(result, move(data_stream))) + return CStatus_OK() + + +cdef CStatus _do_exchange(void* self, const CServerCallContext& context, + unique_ptr[CFlightMessageReader] reader, + unique_ptr[CFlightMessageWriter] writer) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + MetadataRecordBatchReader py_reader = MetadataRecordBatchReader() + MetadataRecordBatchWriter py_writer = MetadataRecordBatchWriter() + FlightDescriptor descriptor = \ + FlightDescriptor.__new__(FlightDescriptor) + + descriptor.descriptor = reader.get().descriptor() + py_reader.reader.reset(reader.release()) + py_writer.writer.reset(writer.release()) + try: + ( self).do_exchange(ServerCallContext.wrap(context), + descriptor, py_reader, py_writer) + return CStatus_OK() + except FlightError as flight_error: + return ( flight_error).to_status() + + +cdef CStatus _do_action_result_next( + void* self, + unique_ptr[CFlightResult]* result +) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + CFlightResult* c_result + + try: + action_result = next( self) + if not isinstance(action_result, Result): + action_result = Result(action_result) + c_result = ( action_result).result.get() + result.reset(new CFlightResult(deref(c_result))) + except StopIteration: + result.reset(nullptr) + except FlightError as flight_error: + return ( flight_error).to_status() + return CStatus_OK() + + +cdef CStatus _do_action(void* self, const CServerCallContext& context, + const CAction& action, + unique_ptr[CResultStream]* result) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + function[cb_result_next] ptr = &_do_action_result_next + py_action = Action(action.type, pyarrow_wrap_buffer(action.body)) + try: + responses = ( self).do_action(ServerCallContext.wrap(context), + py_action) + except FlightError as flight_error: + return ( flight_error).to_status() + # Let the application return an iterator or anything convertible + # into one + if responses is None: + # Server didn't return anything + responses = [] + result.reset(new CPyFlightResultStream(iter(responses), ptr)) + return CStatus_OK() + + +cdef CStatus _list_actions(void* self, const CServerCallContext& context, + vector[CActionType]* actions) except *: + """Callback for implementing Flight servers in Python.""" + cdef: + CActionType action_type + # Method should return a list of ActionTypes or similar tuple + try: + result = ( self).list_actions(ServerCallContext.wrap(context)) + for action in result: + if not isinstance(action, tuple): + raise TypeError( + "Results of list_actions must be ActionType or tuple") + action_type.type = tobytes(action[0]) + action_type.description = tobytes(action[1]) + actions.push_back(action_type) + except FlightError as flight_error: + return ( flight_error).to_status() + return CStatus_OK() + + +cdef CStatus _server_authenticate(void* self, CServerAuthSender* outgoing, + CServerAuthReader* incoming) except *: + """Callback for implementing authentication in Python.""" + sender = ServerAuthSender.wrap(outgoing) + reader = ServerAuthReader.wrap(incoming) + try: + ( self).authenticate(sender, reader) + except FlightError as flight_error: + return ( flight_error).to_status() + finally: + sender.poison() + reader.poison() + return CStatus_OK() + +cdef CStatus _is_valid(void* self, const c_string& token, + c_string* peer_identity) except *: + """Callback for implementing authentication in Python.""" + cdef c_string c_result + try: + c_result = tobytes(( self).is_valid(token)) + peer_identity[0] = c_result + except FlightError as flight_error: + return ( flight_error).to_status() + return CStatus_OK() + + +cdef CStatus _client_authenticate(void* self, CClientAuthSender* outgoing, + CClientAuthReader* incoming) except *: + """Callback for implementing authentication in Python.""" + sender = ClientAuthSender.wrap(outgoing) + reader = ClientAuthReader.wrap(incoming) + try: + ( self).authenticate(sender, reader) + except FlightError as flight_error: + return ( flight_error).to_status() + finally: + sender.poison() + reader.poison() + return CStatus_OK() + + +cdef CStatus _get_token(void* self, c_string* token) except *: + """Callback for implementing authentication in Python.""" + cdef c_string c_result + try: + c_result = tobytes(( self).get_token()) + token[0] = c_result + except FlightError as flight_error: + return ( flight_error).to_status() + return CStatus_OK() + + +cdef CStatus _middleware_sending_headers( + void* self, CAddCallHeaders* add_headers) except *: + """Callback for implementing middleware.""" + try: + headers = ( self).sending_headers() + except FlightError as flight_error: + return ( flight_error).to_status() + + if headers: + for header, values in headers.items(): + if isinstance(values, (str, bytes)): + values = (values,) + # Headers in gRPC (and HTTP/1, HTTP/2) are required to be + # valid, lowercase ASCII. + header = header.lower() + if isinstance(header, str): + header = header.encode("ascii") + for value in values: + if isinstance(value, str): + value = value.encode("ascii") + # Allow bytes values to pass through. + add_headers.AddHeader(header, value) + + return CStatus_OK() + + +cdef CStatus _middleware_call_completed( + void* self, + const CStatus& call_status) except *: + """Callback for implementing middleware.""" + try: + try: + check_flight_status(call_status) + except Exception as e: + ( self).call_completed(e) + else: + ( self).call_completed(None) + except FlightError as flight_error: + return ( flight_error).to_status() + return CStatus_OK() + + +cdef CStatus _middleware_received_headers( + void* self, + const CCallHeaders& c_headers) except *: + """Callback for implementing middleware.""" + try: + headers = convert_headers(c_headers) + ( self).received_headers(headers) + except FlightError as flight_error: + return ( flight_error).to_status() + return CStatus_OK() + + +cdef dict convert_headers(const CCallHeaders& c_headers): + cdef: + CCallHeaders.const_iterator header_iter = c_headers.cbegin() + headers = {} + while header_iter != c_headers.cend(): + header = c_string(deref(header_iter).first).decode("ascii") + value = c_string(deref(header_iter).second) + if not header.endswith("-bin"): + # Text header values in gRPC (and HTTP/1, HTTP/2) are + # required to be valid ASCII. Binary header values are + # exposed as bytes. + value = value.decode("ascii") + headers.setdefault(header, []).append(value) + postincrement(header_iter) + return headers + + +cdef CStatus _server_middleware_start_call( + void* self, + const CCallInfo& c_info, + const CCallHeaders& c_headers, + shared_ptr[CServerMiddleware]* c_instance) except *: + """Callback for implementing server middleware.""" + instance = None + try: + call_info = wrap_call_info(c_info) + headers = convert_headers(c_headers) + instance = ( self).start_call(call_info, headers) + except FlightError as flight_error: + return ( flight_error).to_status() + + if instance: + ServerMiddleware.wrap(instance, c_instance) + + return CStatus_OK() + + +cdef CStatus _client_middleware_start_call( + void* self, + const CCallInfo& c_info, + unique_ptr[CClientMiddleware]* c_instance) except *: + """Callback for implementing client middleware.""" + instance = None + try: + call_info = wrap_call_info(c_info) + instance = ( self).start_call(call_info) + except FlightError as flight_error: + return ( flight_error).to_status() + + if instance: + ClientMiddleware.wrap(instance, c_instance) + + return CStatus_OK() + + +cdef class ServerAuthHandler(_Weakrefable): + """Authentication middleware for a server. + + To implement an authentication mechanism, subclass this class and + override its methods. + + """ + + def authenticate(self, outgoing, incoming): + """Conduct the handshake with the client. + + May raise an error if the client cannot authenticate. + + Parameters + ---------- + outgoing : ServerAuthSender + A channel to send messages to the client. + incoming : ServerAuthReader + A channel to read messages from the client. + """ + raise NotImplementedError + + def is_valid(self, token): + """Validate a client token, returning their identity. + + May return an empty string (if the auth mechanism does not + name the peer) or raise an exception (if the token is + invalid). + + Parameters + ---------- + token : bytes + The authentication token from the client. + + """ + raise NotImplementedError + + cdef PyServerAuthHandler* to_handler(self): + cdef PyServerAuthHandlerVtable vtable + vtable.authenticate = _server_authenticate + vtable.is_valid = _is_valid + return new PyServerAuthHandler(self, vtable) + + +cdef class ClientAuthHandler(_Weakrefable): + """Authentication plugin for a client.""" + + def authenticate(self, outgoing, incoming): + """Conduct the handshake with the server. + + Parameters + ---------- + outgoing : ClientAuthSender + A channel to send messages to the server. + incoming : ClientAuthReader + A channel to read messages from the server. + """ + raise NotImplementedError + + def get_token(self): + """Get the auth token for a call.""" + raise NotImplementedError + + cdef PyClientAuthHandler* to_handler(self): + cdef PyClientAuthHandlerVtable vtable + vtable.authenticate = _client_authenticate + vtable.get_token = _get_token + return new PyClientAuthHandler(self, vtable) + + +_CallInfo = collections.namedtuple("_CallInfo", ["method"]) + + +class CallInfo(_CallInfo): + """Information about a particular RPC for Flight middleware.""" + + +cdef wrap_call_info(const CCallInfo& c_info): + method = wrap_flight_method(c_info.method) + return CallInfo(method=method) + + +cdef class ClientMiddlewareFactory(_Weakrefable): + """A factory for new middleware instances. + + All middleware methods will be called from the same thread as the + RPC method implementation. That is, thread-locals set in the + client are accessible from the middleware itself. + + """ + + def start_call(self, info): + """Called at the start of an RPC. + + This must be thread-safe and must not raise exceptions. + + Parameters + ---------- + info : CallInfo + Information about the call. + + Returns + ------- + instance : ClientMiddleware + An instance of ClientMiddleware (the instance to use for + the call), or None if this call is not intercepted. + + """ + + +cdef class ClientMiddleware(_Weakrefable): + """Client-side middleware for a call, instantiated per RPC. + + Methods here should be fast and must be infallible: they should + not raise exceptions or stall indefinitely. + + """ + + def sending_headers(self): + """A callback before headers are sent. + + Returns + ------- + headers : dict + A dictionary of header values to add to the request, or + None if no headers are to be added. The dictionary should + have string keys and string or list-of-string values. + + Bytes values are allowed, but the underlying transport may + not support them or may restrict them. For gRPC, binary + values are only allowed on headers ending in "-bin". + + Header names must be lowercase ASCII. + + """ + + def received_headers(self, headers): + """A callback when headers are received. + + The default implementation does nothing. + + Parameters + ---------- + headers : dict + A dictionary of headers from the server. Keys are strings + and values are lists of strings (for text headers) or + bytes (for binary headers). + + """ + + def call_completed(self, exception): + """A callback when the call finishes. + + The default implementation does nothing. + + Parameters + ---------- + exception : ArrowException + If the call errored, this is the equivalent + exception. Will be None if the call succeeded. + + """ + + @staticmethod + cdef void wrap(object py_middleware, + unique_ptr[CClientMiddleware]* c_instance): + cdef PyClientMiddlewareVtable vtable + vtable.sending_headers = _middleware_sending_headers + vtable.received_headers = _middleware_received_headers + vtable.call_completed = _middleware_call_completed + c_instance[0].reset(new CPyClientMiddleware(py_middleware, vtable)) + + +cdef class ServerMiddlewareFactory(_Weakrefable): + """A factory for new middleware instances. + + All middleware methods will be called from the same thread as the + RPC method implementation. That is, thread-locals set in the + middleware are accessible from the method itself. + + """ + + def start_call(self, info, headers): + """Called at the start of an RPC. + + This must be thread-safe. + + Parameters + ---------- + info : CallInfo + Information about the call. + headers : dict + A dictionary of headers from the client. Keys are strings + and values are lists of strings (for text headers) or + bytes (for binary headers). + + Returns + ------- + instance : ServerMiddleware + An instance of ServerMiddleware (the instance to use for + the call), or None if this call is not intercepted. + + Raises + ------ + exception : pyarrow.ArrowException + If an exception is raised, the call will be rejected with + the given error. + + """ + + +cdef class TracingServerMiddlewareFactory(ServerMiddlewareFactory): + """A factory for tracing middleware instances. + + This enables OpenTelemetry support in Arrow (if Arrow was compiled + with OpenTelemetry support enabled). A new span will be started on + each RPC call. The TracingServerMiddleware instance can then be + retrieved within an RPC handler to get the propagated context, + which can be used to start a new span on the Python side. + + Because the Python/C++ OpenTelemetry libraries do not + interoperate, spans on the C++ side are not directly visible to + the Python side and vice versa. + + """ + + +cdef class ServerMiddleware(_Weakrefable): + """Server-side middleware for a call, instantiated per RPC. + + Methods here should be fast and must be infallible: they should + not raise exceptions or stall indefinitely. + + """ + + def sending_headers(self): + """A callback before headers are sent. + + Returns + ------- + headers : dict + A dictionary of header values to add to the response, or + None if no headers are to be added. The dictionary should + have string keys and string or list-of-string values. + + Bytes values are allowed, but the underlying transport may + not support them or may restrict them. For gRPC, binary + values are only allowed on headers ending in "-bin". + + Header names must be lowercase ASCII. + + """ + + def call_completed(self, exception): + """A callback when the call finishes. + + Parameters + ---------- + exception : pyarrow.ArrowException + If the call errored, this is the equivalent + exception. Will be None if the call succeeded. + + """ + + @staticmethod + cdef void wrap(object py_middleware, + shared_ptr[CServerMiddleware]* c_instance): + cdef PyServerMiddlewareVtable vtable + vtable.sending_headers = _middleware_sending_headers + vtable.call_completed = _middleware_call_completed + c_instance[0].reset(new CPyServerMiddleware(py_middleware, vtable)) + + +class TracingServerMiddleware(ServerMiddleware): + __slots__ = ["trace_context"] + + def __init__(self, trace_context): + self.trace_context = trace_context + + +cdef class _ServerMiddlewareFactoryWrapper(ServerMiddlewareFactory): + """Wrapper to bundle server middleware into a single C++ one.""" + + cdef: + dict factories + + def __init__(self, dict factories): + self.factories = factories + + def start_call(self, info, headers): + instances = {} + for key, factory in self.factories.items(): + instance = factory.start_call(info, headers) + if instance: + # TODO: prevent duplicate keys + instances[key] = instance + if instances: + wrapper = _ServerMiddlewareWrapper(instances) + return wrapper + return None + + +cdef class _ServerMiddlewareWrapper(ServerMiddleware): + cdef: + dict middleware + + def __init__(self, dict middleware): + self.middleware = middleware + + def sending_headers(self): + headers = collections.defaultdict(list) + for instance in self.middleware.values(): + more_headers = instance.sending_headers() + if not more_headers: + continue + # Manually merge with existing headers (since headers are + # multi-valued) + for key, values in more_headers.items(): + # ARROW-16606 gRPC aborts given non-lowercase headers + key = key.lower() + if isinstance(values, (bytes, str)): + values = (values,) + headers[key].extend(values) + return headers + + def call_completed(self, exception): + for instance in self.middleware.values(): + instance.call_completed(exception) + + +cdef class _FlightServerFinalizer(_Weakrefable): + """ + A finalizer that shuts down the server on destruction. + + See ARROW-16597. If the server is still active at interpreter + exit, the process may segfault. + """ + + cdef: + shared_ptr[PyFlightServer] server + + def finalize(self): + cdef: + PyFlightServer* server = self.server.get() + CStatus status + if server == NULL: + return + try: + with nogil: + status = server.Shutdown() + if status.ok(): + status = server.Wait() + check_flight_status(status) + finally: + self.server.reset() + + +cdef class FlightServerBase(_Weakrefable): + """A Flight service definition. + + To start the server, create an instance of this class with an + appropriate location. The server will be running as soon as the + instance is created; it is not required to call :meth:`serve`. + + Override methods to define your Flight service. + + Parameters + ---------- + location : str, tuple or Location optional, default None + Location to serve on. Either a gRPC URI like `grpc://localhost:port`, + a tuple of (host, port) pair, or a Location instance. + If None is passed then the server will be started on localhost with a + system provided random port. + auth_handler : ServerAuthHandler optional, default None + An authentication mechanism to use. May be None. + tls_certificates : list optional, default None + A list of (certificate, key) pairs. + verify_client : boolean optional, default False + If True, then enable mutual TLS: require the client to present + a client certificate, and validate the certificate. + root_certificates : bytes optional, default None + If enabling mutual TLS, this specifies the PEM-encoded root + certificate used to validate client certificates. + middleware : dict optional, default None + A dictionary of :class:`ServerMiddlewareFactory` instances. The + string keys can be used to retrieve the middleware instance within + RPC handlers (see :meth:`ServerCallContext.get_middleware`). + + """ + + cdef: + shared_ptr[PyFlightServer] server + object finalizer + + def __init__(self, location=None, auth_handler=None, + tls_certificates=None, verify_client=None, + root_certificates=None, middleware=None): + self.finalizer = None + if isinstance(location, (bytes, str)): + location = Location(location) + elif isinstance(location, (tuple, type(None))): + if location is None: + location = ('localhost', 0) + host, port = location + if tls_certificates: + location = Location.for_grpc_tls(host, port) + else: + location = Location.for_grpc_tcp(host, port) + elif not isinstance(location, Location): + raise TypeError('`location` argument must be a string, tuple or a ' + 'Location instance') + self.init(location, auth_handler, tls_certificates, verify_client, + tobytes(root_certificates or b""), middleware) + + cdef init(self, Location location, ServerAuthHandler auth_handler, + list tls_certificates, c_bool verify_client, + bytes root_certificates, dict middleware): + cdef: + PyFlightServerVtable vtable = PyFlightServerVtable() + PyFlightServer* c_server + unique_ptr[CFlightServerOptions] c_options + CCertKeyPair c_cert + function[cb_server_middleware_start_call] start_call = \ + &_server_middleware_start_call + pair[c_string, shared_ptr[CServerMiddlewareFactory]] c_middleware + + c_options.reset(new CFlightServerOptions(Location.unwrap(location))) + # mTLS configuration + c_options.get().verify_client = verify_client + c_options.get().root_certificates = root_certificates + + if auth_handler: + if not isinstance(auth_handler, ServerAuthHandler): + raise TypeError("auth_handler must be a ServerAuthHandler, " + "not a '{}'".format(type(auth_handler))) + c_options.get().auth_handler.reset( + ( auth_handler).to_handler()) + + if tls_certificates: + for cert, key in tls_certificates: + c_cert.pem_cert = tobytes(cert) + c_cert.pem_key = tobytes(key) + c_options.get().tls_certificates.push_back(c_cert) + + if middleware: + non_tracing_middleware = {} + enable_tracing = None + for key, factory in middleware.items(): + if isinstance(factory, TracingServerMiddlewareFactory): + if enable_tracing is not None: + raise ValueError( + "Can only provide " + "TracingServerMiddlewareFactory once") + if tobytes(key) == CPyServerMiddlewareName: + raise ValueError(f"Middleware key cannot be {key}") + enable_tracing = key + else: + non_tracing_middleware[key] = factory + + if enable_tracing: + c_middleware.first = tobytes(enable_tracing) + c_middleware.second = MakeTracingServerMiddlewareFactory() + c_options.get().middleware.push_back(c_middleware) + + py_middleware = _ServerMiddlewareFactoryWrapper( + non_tracing_middleware) + c_middleware.first = CPyServerMiddlewareName + c_middleware.second.reset(new CPyServerMiddlewareFactory( + py_middleware, + start_call)) + c_options.get().middleware.push_back(c_middleware) + + vtable.list_flights = &_list_flights + vtable.get_flight_info = &_get_flight_info + vtable.get_schema = &_get_schema + vtable.do_put = &_do_put + vtable.do_get = &_do_get + vtable.do_exchange = &_do_exchange + vtable.list_actions = &_list_actions + vtable.do_action = &_do_action + + c_server = new PyFlightServer(self, vtable) + self.server.reset(c_server) + with nogil: + check_flight_status(c_server.Init(deref(c_options))) + cdef _FlightServerFinalizer finalizer = _FlightServerFinalizer() + finalizer.server = self.server + self.finalizer = weakref.finalize(self, finalizer.finalize) + + @property + def port(self): + """ + Get the port that this server is listening on. + + Returns a non-positive value if the operation is invalid + (e.g. init() was not called or server is listening on a domain + socket). + """ + return self.server.get().port() + + def list_flights(self, context, criteria): + """List flights available on this service. + + Applications should override this method to implement their + own behavior. The default method raises a NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + criteria : bytes + Filter criteria provided by the client. + + Returns + ------- + iterator of FlightInfo + + """ + raise NotImplementedError + + def get_flight_info(self, context, descriptor): + """Get information about a flight. + + Applications should override this method to implement their + own behavior. The default method raises a NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + descriptor : FlightDescriptor + The descriptor for the flight provided by the client. + + Returns + ------- + FlightInfo + + """ + raise NotImplementedError + + def get_schema(self, context, descriptor): + """Get the schema of a flight. + + Applications should override this method to implement their + own behavior. The default method raises a NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + descriptor : FlightDescriptor + The descriptor for the flight provided by the client. + + Returns + ------- + Schema + + """ + raise NotImplementedError + + def do_put(self, context, descriptor, reader: MetadataRecordBatchReader, + writer: FlightMetadataWriter): + """Write data to a flight. + + Applications should override this method to implement their + own behavior. The default method raises a NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + descriptor : FlightDescriptor + The descriptor for the flight provided by the client. + reader : MetadataRecordBatchReader + A reader for data uploaded by the client. + writer : FlightMetadataWriter + A writer to send responses to the client. + + """ + raise NotImplementedError + + def do_get(self, context, ticket): + """Write data to a flight. + + Applications should override this method to implement their + own behavior. The default method raises a NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + ticket : Ticket + The ticket for the flight. + + Returns + ------- + FlightDataStream + A stream of data to send back to the client. + + """ + raise NotImplementedError + + def do_exchange(self, context, descriptor, reader, writer): + """Write data to a flight. + + Applications should override this method to implement their + own behavior. The default method raises a NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + descriptor : FlightDescriptor + The descriptor for the flight provided by the client. + reader : MetadataRecordBatchReader + A reader for data uploaded by the client. + writer : MetadataRecordBatchWriter + A writer to send responses to the client. + + """ + raise NotImplementedError + + def list_actions(self, context): + """List custom actions available on this server. + + Applications should override this method to implement their + own behavior. The default method raises a NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + + Returns + ------- + iterator of ActionType or tuple + + """ + raise NotImplementedError + + def do_action(self, context, action): + """Execute a custom action. + + This method should return an iterator, or it should be a + generator. Applications should override this method to + implement their own behavior. The default method raises a + NotImplementedError. + + Parameters + ---------- + context : ServerCallContext + Common contextual information. + action : Action + The action to execute. + + Returns + ------- + iterator of bytes + + """ + raise NotImplementedError + + def serve(self): + """Block until the server shuts down. + + This method only returns if shutdown() is called or a signal is + received. + """ + if self.server.get() == nullptr: + raise ValueError("run() on uninitialized FlightServerBase") + with nogil: + check_flight_status(self.server.get().ServeWithSignals()) + + def run(self): + """Block until the server shuts down. + + .. deprecated:: 0.15.0 + Use the ``FlightServer.serve`` method instead + """ + warnings.warn("The 'FlightServer.run' method is deprecated, use " + "FlightServer.serve method instead") + self.serve() + + def shutdown(self): + """Shut down the server, blocking until current requests finish. + + Do not call this directly from the implementation of a Flight + method, as then the server will block forever waiting for that + request to finish. Instead, call this method from a background + thread. + + This method should only be called once. + """ + # Must not hold the GIL: shutdown waits for pending RPCs to + # complete. Holding the GIL means Python-implemented Flight + # methods will never get to run, so this will hang + # indefinitely. + if self.server.get() == nullptr: + raise ValueError("shutdown() on uninitialized FlightServerBase") + with nogil: + check_flight_status(self.server.get().Shutdown()) + + def wait(self): + """Block until server is terminated with shutdown.""" + with nogil: + self.server.get().Wait() + + def __enter__(self): + return self + + def __exit__(self, exc_type, exc_value, traceback): + if self.finalizer: + self.finalizer() + + +def connect(location, **kwargs): + """ + Connect to a Flight server. + + Parameters + ---------- + location : str, tuple, or Location + Location to connect to. Either a URI like "grpc://localhost:port", + a tuple of (host, port), or a Location instance. + tls_root_certs : bytes or None + PEM-encoded. + cert_chain: str or None + If provided, enables TLS mutual authentication. + private_key: str or None + If provided, enables TLS mutual authentication. + override_hostname : str or None + Override the hostname checked by TLS. Insecure, use with caution. + middleware : list or None + A list of ClientMiddlewareFactory instances to apply. + write_size_limit_bytes : int or None + A soft limit on the size of a data payload sent to the + server. Enabled if positive. If enabled, writing a record + batch that (when serialized) exceeds this limit will raise an + exception; the client can retry the write with a smaller + batch. + disable_server_verification : boolean or None + Disable verifying the server when using TLS. + Insecure, use with caution. + generic_options : list or None + A list of generic (string, int or string) options to pass to + the underlying transport. + + Returns + ------- + client : FlightClient + """ + return FlightClient(location, **kwargs) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_fs.cpython-310-x86_64-linux-gnu.so b/vllm/lib/python3.10/site-packages/pyarrow/_fs.cpython-310-x86_64-linux-gnu.so new file mode 100644 index 0000000000000000000000000000000000000000..7a30cfca34e1a8a626e82d67e6cd853667e6e125 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_fs.cpython-310-x86_64-linux-gnu.so @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73c0243b4f340126fe086337dcfece841c60fd9382fca302f07f52eaa7c829fa +size 501392 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_fs.pxd b/vllm/lib/python3.10/site-packages/pyarrow/_fs.pxd new file mode 100644 index 0000000000000000000000000000000000000000..0df75530bbd6ec3552131e11acc5b0406627fe65 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_fs.pxd @@ -0,0 +1,91 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: language_level = 3 + +from pyarrow.includes.common cimport * +from pyarrow.includes.libarrow_fs cimport * +from pyarrow.lib import _detect_compression, frombytes, tobytes +from pyarrow.lib cimport * + + +cpdef enum FileType: + NotFound = CFileType_NotFound + Unknown = CFileType_Unknown + File = CFileType_File + Directory = CFileType_Directory + + +cdef class FileInfo(_Weakrefable): + cdef: + CFileInfo info + + @staticmethod + cdef wrap(CFileInfo info) + + cdef inline CFileInfo unwrap(self) nogil + + @staticmethod + cdef CFileInfo unwrap_safe(obj) + + +cdef class FileSelector(_Weakrefable): + cdef: + CFileSelector selector + + @staticmethod + cdef FileSelector wrap(CFileSelector selector) + + cdef inline CFileSelector unwrap(self) nogil + + +cdef class FileSystem(_Weakrefable): + cdef: + shared_ptr[CFileSystem] wrapped + CFileSystem* fs + + cdef init(self, const shared_ptr[CFileSystem]& wrapped) + + @staticmethod + cdef wrap(const shared_ptr[CFileSystem]& sp) + + cdef inline shared_ptr[CFileSystem] unwrap(self) nogil + + +cdef class LocalFileSystem(FileSystem): + cdef init(self, const shared_ptr[CFileSystem]& wrapped) + + +cdef class SubTreeFileSystem(FileSystem): + cdef: + CSubTreeFileSystem* subtreefs + + cdef init(self, const shared_ptr[CFileSystem]& wrapped) + + +cdef class _MockFileSystem(FileSystem): + cdef: + CMockFileSystem* mockfs + + cdef init(self, const shared_ptr[CFileSystem]& wrapped) + + +cdef class PyFileSystem(FileSystem): + cdef: + CPyFileSystem* pyfs + + cdef init(self, const shared_ptr[CFileSystem]& wrapped) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_fs.pyx b/vllm/lib/python3.10/site-packages/pyarrow/_fs.pyx new file mode 100644 index 0000000000000000000000000000000000000000..dbfb6ed114553bc8435d0f61090c554e4ede5632 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_fs.pyx @@ -0,0 +1,1628 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: language_level = 3 + +from cpython.datetime cimport datetime, PyDateTime_DateTime +from cython cimport binding + +from pyarrow.includes.common cimport * +from pyarrow.includes.libarrow_python cimport PyDateTime_to_TimePoint +from pyarrow.lib import _detect_compression, frombytes, tobytes +from pyarrow.lib cimport * +from pyarrow.util import _stringify_path + +from abc import ABC, abstractmethod +from datetime import datetime, timezone +import os +import pathlib +import sys + + +cdef _init_ca_paths(): + cdef CFileSystemGlobalOptions options + + import ssl + paths = ssl.get_default_verify_paths() + if paths.cafile: + options.tls_ca_file_path = os.fsencode(paths.cafile) + if paths.capath: + options.tls_ca_dir_path = os.fsencode(paths.capath) + check_status(CFileSystemsInitialize(options)) + + +if sys.platform == 'linux': + # ARROW-9261: On Linux, we may need to fixup the paths to TLS CA certs + # (especially in manylinux packages) since the values hardcoded at + # compile-time in libcurl may be wrong. + _init_ca_paths() + + +cdef inline c_string _path_as_bytes(path) except *: + # handle only abstract paths, not bound to any filesystem like pathlib is, + # so we only accept plain strings + if not isinstance(path, (bytes, str)): + raise TypeError('Path must be a string') + # tobytes always uses utf-8, which is more or less ok, at least on Windows + # since the C++ side then decodes from utf-8. On Unix, os.fsencode may be + # better. + return tobytes(path) + + +cdef object _wrap_file_type(CFileType ty): + return FileType( ty) + + +cdef CFileType _unwrap_file_type(FileType ty) except *: + if ty == FileType.Unknown: + return CFileType_Unknown + elif ty == FileType.NotFound: + return CFileType_NotFound + elif ty == FileType.File: + return CFileType_File + elif ty == FileType.Directory: + return CFileType_Directory + assert 0 + + +def _file_type_to_string(ty): + # Python 3.11 changed str(IntEnum) to return the string representation + # of the integer value: https://github.com/python/cpython/issues/94763 + return f"{ty.__class__.__name__}.{ty._name_}" + + +cdef class FileInfo(_Weakrefable): + """ + FileSystem entry info. + + Parameters + ---------- + path : str + The full path to the filesystem entry. + type : FileType + The type of the filesystem entry. + mtime : datetime or float, default None + If given, the modification time of the filesystem entry. + If a float is given, it is the number of seconds since the + Unix epoch. + mtime_ns : int, default None + If given, the modification time of the filesystem entry, + in nanoseconds since the Unix epoch. + `mtime` and `mtime_ns` are mutually exclusive. + size : int, default None + If given, the filesystem entry size in bytes. This should only + be given if `type` is `FileType.File`. + + Examples + -------- + Generate a file: + + >>> from pyarrow import fs + >>> local = fs.LocalFileSystem() + >>> path_fs = local_path + '/pyarrow-fs-example.dat' + >>> with local.open_output_stream(path_fs) as stream: + ... stream.write(b'data') + 4 + + Get FileInfo object using ``get_file_info()``: + + >>> file_info = local.get_file_info(path_fs) + >>> file_info + + + Inspect FileInfo attributes: + + >>> file_info.type + + + >>> file_info.is_file + True + + >>> file_info.path + '/.../pyarrow-fs-example.dat' + + >>> file_info.base_name + 'pyarrow-fs-example.dat' + + >>> file_info.size + 4 + + >>> file_info.extension + 'dat' + + >>> file_info.mtime # doctest: +SKIP + datetime.datetime(2022, 6, 29, 7, 56, 10, 873922, tzinfo=datetime.timezone.utc) + + >>> file_info.mtime_ns # doctest: +SKIP + 1656489370873922073 + """ + + def __init__(self, path, FileType type=FileType.Unknown, *, + mtime=None, mtime_ns=None, size=None): + self.info.set_path(tobytes(path)) + self.info.set_type(_unwrap_file_type(type)) + if mtime is not None: + if mtime_ns is not None: + raise TypeError("Only one of mtime and mtime_ns " + "can be given") + if isinstance(mtime, datetime): + self.info.set_mtime(PyDateTime_to_TimePoint( + mtime)) + else: + self.info.set_mtime(TimePoint_from_s(mtime)) + elif mtime_ns is not None: + self.info.set_mtime(TimePoint_from_ns(mtime_ns)) + if size is not None: + self.info.set_size(size) + + @staticmethod + cdef wrap(CFileInfo info): + cdef FileInfo self = FileInfo.__new__(FileInfo) + self.info = move(info) + return self + + cdef inline CFileInfo unwrap(self) nogil: + return self.info + + @staticmethod + cdef CFileInfo unwrap_safe(obj): + if not isinstance(obj, FileInfo): + raise TypeError("Expected FileInfo instance, got {0}" + .format(type(obj))) + return ( obj).unwrap() + + def __repr__(self): + def getvalue(attr): + try: + return getattr(self, attr) + except ValueError: + return '' + + s = (f'>> file_info = local.get_file_info(path) + >>> file_info.path + '/.../pyarrow-fs-example.dat' + """ + return frombytes(self.info.path()) + + @property + def base_name(self): + """ + The file base name. + + Component after the last directory separator. + + Examples + -------- + >>> file_info = local.get_file_info(path) + >>> file_info.base_name + 'pyarrow-fs-example.dat' + """ + return frombytes(self.info.base_name()) + + @property + def size(self): + """ + The size in bytes, if available. + + Only regular files are guaranteed to have a size. + + Returns + ------- + size : int or None + """ + cdef int64_t size + size = self.info.size() + return (size if size != -1 else None) + + @property + def extension(self): + """ + The file extension. + + Examples + -------- + >>> file_info = local.get_file_info(path) + >>> file_info.extension + 'dat' + """ + return frombytes(self.info.extension()) + + @property + def mtime(self): + """ + The time of last modification, if available. + + Returns + ------- + mtime : datetime.datetime or None + + Examples + -------- + >>> file_info = local.get_file_info(path) + >>> file_info.mtime # doctest: +SKIP + datetime.datetime(2022, 6, 29, 7, 56, 10, 873922, tzinfo=datetime.timezone.utc) + """ + cdef int64_t nanoseconds + nanoseconds = TimePoint_to_ns(self.info.mtime()) + return (datetime.fromtimestamp(nanoseconds / 1.0e9, timezone.utc) + if nanoseconds != -1 else None) + + @property + def mtime_ns(self): + """ + The time of last modification, if available, expressed in nanoseconds + since the Unix epoch. + + Returns + ------- + mtime_ns : int or None + + Examples + -------- + >>> file_info = local.get_file_info(path) + >>> file_info.mtime_ns # doctest: +SKIP + 1656489370873922073 + """ + cdef int64_t nanoseconds + nanoseconds = TimePoint_to_ns(self.info.mtime()) + return (nanoseconds if nanoseconds != -1 else None) + + +cdef class FileSelector(_Weakrefable): + """ + File and directory selector. + + It contains a set of options that describes how to search for files and + directories. + + Parameters + ---------- + base_dir : str + The directory in which to select files. Relative paths also work, use + '.' for the current directory and '..' for the parent. + allow_not_found : bool, default False + The behavior if `base_dir` doesn't exist in the filesystem. + If false, an error is returned. + If true, an empty selection is returned. + recursive : bool, default False + Whether to recurse into subdirectories. + + Examples + -------- + List the contents of a directory and subdirectories: + + >>> selector_1 = fs.FileSelector(local_path, recursive=True) + >>> local.get_file_info(selector_1) # doctest: +SKIP + [, + , + ] + + List only the contents of the base directory: + + >>> selector_2 = fs.FileSelector(local_path) + >>> local.get_file_info(selector_2) # doctest: +SKIP + [, + ] + + Return empty selection if the directory doesn't exist: + + >>> selector_not_found = fs.FileSelector(local_path + '/missing', + ... recursive=True, + ... allow_not_found=True) + >>> local.get_file_info(selector_not_found) + [] + """ + + def __init__(self, base_dir, bint allow_not_found=False, + bint recursive=False): + self.base_dir = base_dir + self.recursive = recursive + self.allow_not_found = allow_not_found + + @staticmethod + cdef FileSelector wrap(CFileSelector wrapped): + cdef FileSelector self = FileSelector.__new__(FileSelector) + self.selector = move(wrapped) + return self + + cdef inline CFileSelector unwrap(self) nogil: + return self.selector + + @property + def base_dir(self): + return frombytes(self.selector.base_dir) + + @base_dir.setter + def base_dir(self, base_dir): + self.selector.base_dir = _path_as_bytes(base_dir) + + @property + def allow_not_found(self): + return self.selector.allow_not_found + + @allow_not_found.setter + def allow_not_found(self, bint allow_not_found): + self.selector.allow_not_found = allow_not_found + + @property + def recursive(self): + return self.selector.recursive + + @recursive.setter + def recursive(self, bint recursive): + self.selector.recursive = recursive + + def __repr__(self): + return ("".format(self)) + + +cdef class FileSystem(_Weakrefable): + """ + Abstract file system API. + """ + + def __init__(self): + raise TypeError("FileSystem is an abstract class, instantiate one of " + "the subclasses instead: LocalFileSystem or " + "SubTreeFileSystem") + + @staticmethod + @binding(True) # Required for cython < 3 + def _from_uri(uri): + fs, _path = FileSystem.from_uri(uri) + return fs + + @staticmethod + def from_uri(uri): + """ + Create a new FileSystem from URI or Path. + + Recognized URI schemes are "file", "mock", "s3fs", "gs", "gcs", "hdfs" and "viewfs". + In addition, the argument can be a pathlib.Path object, or a string + describing an absolute local path. + + Parameters + ---------- + uri : string + URI-based path, for example: file:///some/local/path. + + Returns + ------- + tuple of (FileSystem, str path) + With (filesystem, path) tuple where path is the abstract path + inside the FileSystem instance. + + Examples + -------- + Create a new FileSystem subclass from a URI: + + >>> uri = 'file:///{}/pyarrow-fs-example.dat'.format(local_path) + >>> local_new, path_new = fs.FileSystem.from_uri(uri) + >>> local_new + >> path_new + '/.../pyarrow-fs-example.dat' + + Or from a s3 bucket: + + >>> fs.FileSystem.from_uri("s3://usgs-landsat/collection02/") + (, 'usgs-landsat/collection02') + """ + cdef: + c_string c_path + c_string c_uri + CResult[shared_ptr[CFileSystem]] result + + if isinstance(uri, pathlib.Path): + # Make absolute + uri = uri.resolve().absolute() + c_uri = tobytes(_stringify_path(uri)) + with nogil: + result = CFileSystemFromUriOrPath(c_uri, &c_path) + return FileSystem.wrap(GetResultValue(result)), frombytes(c_path) + + cdef init(self, const shared_ptr[CFileSystem]& wrapped): + self.wrapped = wrapped + self.fs = wrapped.get() + + @staticmethod + cdef wrap(const shared_ptr[CFileSystem]& sp): + cdef FileSystem self + + typ = frombytes(sp.get().type_name()) + if typ == 'local': + self = LocalFileSystem.__new__(LocalFileSystem) + elif typ == 'mock': + self = _MockFileSystem.__new__(_MockFileSystem) + elif typ == 'subtree': + self = SubTreeFileSystem.__new__(SubTreeFileSystem) + elif typ == 's3': + from pyarrow._s3fs import S3FileSystem + self = S3FileSystem.__new__(S3FileSystem) + elif typ == 'gcs': + from pyarrow._gcsfs import GcsFileSystem + self = GcsFileSystem.__new__(GcsFileSystem) + elif typ == 'abfs': + from pyarrow._azurefs import AzureFileSystem + self = AzureFileSystem.__new__(AzureFileSystem) + elif typ == 'hdfs': + from pyarrow._hdfs import HadoopFileSystem + self = HadoopFileSystem.__new__(HadoopFileSystem) + elif typ.startswith('py::'): + self = PyFileSystem.__new__(PyFileSystem) + else: + raise TypeError('Cannot wrap FileSystem pointer') + + self.init(sp) + return self + + cdef inline shared_ptr[CFileSystem] unwrap(self) nogil: + return self.wrapped + + def equals(self, FileSystem other not None): + """ + Parameters + ---------- + other : pyarrow.fs.FileSystem + + Returns + ------- + bool + """ + return self.fs.Equals(other.unwrap()) + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return NotImplemented + + @property + def type_name(self): + """ + The filesystem's type name. + """ + return frombytes(self.fs.type_name()) + + def get_file_info(self, paths_or_selector): + """ + Get info for the given files. + + Any symlink is automatically dereferenced, recursively. A non-existing + or unreachable file returns a FileStat object and has a FileType of + value NotFound. An exception indicates a truly exceptional condition + (low-level I/O error, etc.). + + Parameters + ---------- + paths_or_selector : FileSelector, path-like or list of path-likes + Either a selector object, a path-like object or a list of + path-like objects. The selector's base directory will not be + part of the results, even if it exists. If it doesn't exist, + use `allow_not_found`. + + Returns + ------- + FileInfo or list of FileInfo + Single FileInfo object is returned for a single path, otherwise + a list of FileInfo objects is returned. + + Examples + -------- + >>> local + + >>> local.get_file_info("/{}/pyarrow-fs-example.dat".format(local_path)) + + """ + cdef: + CFileInfo info + c_string path + vector[CFileInfo] infos + vector[c_string] paths + CFileSelector selector + + if isinstance(paths_or_selector, FileSelector): + with nogil: + selector = (paths_or_selector).selector + infos = GetResultValue(self.fs.GetFileInfo(selector)) + elif isinstance(paths_or_selector, (list, tuple)): + paths = [_path_as_bytes(s) for s in paths_or_selector] + with nogil: + infos = GetResultValue(self.fs.GetFileInfo(paths)) + elif isinstance(paths_or_selector, (bytes, str)): + path =_path_as_bytes(paths_or_selector) + with nogil: + info = GetResultValue(self.fs.GetFileInfo(path)) + return FileInfo.wrap(info) + else: + raise TypeError('Must pass either path(s) or a FileSelector') + + return [FileInfo.wrap(info) for info in infos] + + def create_dir(self, path, *, bint recursive=True): + """ + Create a directory and subdirectories. + + This function succeeds if the directory already exists. + + Parameters + ---------- + path : str + The path of the new directory. + recursive : bool, default True + Create nested directories as well. + """ + cdef c_string directory = _path_as_bytes(path) + with nogil: + check_status(self.fs.CreateDir(directory, recursive=recursive)) + + def delete_dir(self, path): + """ + Delete a directory and its contents, recursively. + + Parameters + ---------- + path : str + The path of the directory to be deleted. + """ + cdef c_string directory = _path_as_bytes(path) + with nogil: + check_status(self.fs.DeleteDir(directory)) + + def delete_dir_contents(self, path, *, + bint accept_root_dir=False, + bint missing_dir_ok=False): + """ + Delete a directory's contents, recursively. + + Like delete_dir, but doesn't delete the directory itself. + + Parameters + ---------- + path : str + The path of the directory to be deleted. + accept_root_dir : boolean, default False + Allow deleting the root directory's contents + (if path is empty or "/") + missing_dir_ok : boolean, default False + If False then an error is raised if path does + not exist + """ + cdef c_string directory = _path_as_bytes(path) + if accept_root_dir and directory.strip(b"/") == b"": + with nogil: + check_status(self.fs.DeleteRootDirContents()) + else: + with nogil: + check_status(self.fs.DeleteDirContents(directory, + missing_dir_ok)) + + def move(self, src, dest): + """ + Move / rename a file or directory. + + If the destination exists: + - if it is a non-empty directory, an error is returned + - otherwise, if it has the same type as the source, it is replaced + - otherwise, behavior is unspecified (implementation-dependent). + + Parameters + ---------- + src : str + The path of the file or the directory to be moved. + dest : str + The destination path where the file or directory is moved to. + + Examples + -------- + Create a new folder with a file: + + >>> local.create_dir('/tmp/other_dir') + >>> local.copy_file(path,'/tmp/move_example.dat') + + Move the file: + + >>> local.move('/tmp/move_example.dat', + ... '/tmp/other_dir/move_example_2.dat') + + Inspect the file info: + + >>> local.get_file_info('/tmp/other_dir/move_example_2.dat') + + >>> local.get_file_info('/tmp/move_example.dat') + + + Delete the folder: + >>> local.delete_dir('/tmp/other_dir') + """ + cdef: + c_string source = _path_as_bytes(src) + c_string destination = _path_as_bytes(dest) + with nogil: + check_status(self.fs.Move(source, destination)) + + def copy_file(self, src, dest): + """ + Copy a file. + + If the destination exists and is a directory, an error is returned. + Otherwise, it is replaced. + + Parameters + ---------- + src : str + The path of the file to be copied from. + dest : str + The destination path where the file is copied to. + + Examples + -------- + >>> local.copy_file(path, + ... local_path + '/pyarrow-fs-example_copy.dat') + + Inspect the file info: + + >>> local.get_file_info(local_path + '/pyarrow-fs-example_copy.dat') + + >>> local.get_file_info(path) + + """ + cdef: + c_string source = _path_as_bytes(src) + c_string destination = _path_as_bytes(dest) + with nogil: + check_status(self.fs.CopyFile(source, destination)) + + def delete_file(self, path): + """ + Delete a file. + + Parameters + ---------- + path : str + The path of the file to be deleted. + """ + cdef c_string file = _path_as_bytes(path) + with nogil: + check_status(self.fs.DeleteFile(file)) + + def _wrap_input_stream(self, stream, path, compression, buffer_size): + if buffer_size is not None and buffer_size != 0: + stream = BufferedInputStream(stream, buffer_size) + if compression == 'detect': + compression = _detect_compression(path) + if compression is not None: + stream = CompressedInputStream(stream, compression) + return stream + + def _wrap_output_stream(self, stream, path, compression, buffer_size): + if buffer_size is not None and buffer_size != 0: + stream = BufferedOutputStream(stream, buffer_size) + if compression == 'detect': + compression = _detect_compression(path) + if compression is not None: + stream = CompressedOutputStream(stream, compression) + return stream + + def open_input_file(self, path): + """ + Open an input file for random access reading. + + Parameters + ---------- + path : str + The source to open for reading. + + Returns + ------- + stream : NativeFile + + Examples + -------- + Print the data from the file with `open_input_file()`: + + >>> with local.open_input_file(path) as f: + ... print(f.readall()) + b'data' + """ + cdef: + c_string pathstr = _path_as_bytes(path) + NativeFile stream = NativeFile() + shared_ptr[CRandomAccessFile] in_handle + + with nogil: + in_handle = GetResultValue(self.fs.OpenInputFile(pathstr)) + + stream.set_random_access_file(in_handle) + stream.is_readable = True + return stream + + def open_input_stream(self, path, compression='detect', buffer_size=None): + """ + Open an input stream for sequential reading. + + Parameters + ---------- + path : str + The source to open for reading. + compression : str optional, default 'detect' + The compression algorithm to use for on-the-fly decompression. + If "detect" and source is a file path, then compression will be + chosen based on the file extension. + If None, no compression will be applied. Otherwise, a well-known + algorithm name must be supplied (e.g. "gzip"). + buffer_size : int optional, default None + If None or 0, no buffering will happen. Otherwise the size of the + temporary read buffer. + + Returns + ------- + stream : NativeFile + + Examples + -------- + Print the data from the file with `open_input_stream()`: + + >>> with local.open_input_stream(path) as f: + ... print(f.readall()) + b'data' + """ + cdef: + c_string pathstr = _path_as_bytes(path) + NativeFile stream = NativeFile() + shared_ptr[CInputStream] in_handle + + with nogil: + in_handle = GetResultValue(self.fs.OpenInputStream(pathstr)) + + stream.set_input_stream(in_handle) + stream.is_readable = True + + return self._wrap_input_stream( + stream, path=path, compression=compression, buffer_size=buffer_size + ) + + def open_output_stream(self, path, compression='detect', + buffer_size=None, metadata=None): + """ + Open an output stream for sequential writing. + + If the target already exists, existing data is truncated. + + Parameters + ---------- + path : str + The source to open for writing. + compression : str optional, default 'detect' + The compression algorithm to use for on-the-fly compression. + If "detect" and source is a file path, then compression will be + chosen based on the file extension. + If None, no compression will be applied. Otherwise, a well-known + algorithm name must be supplied (e.g. "gzip"). + buffer_size : int optional, default None + If None or 0, no buffering will happen. Otherwise the size of the + temporary write buffer. + metadata : dict optional, default None + If not None, a mapping of string keys to string values. + Some filesystems support storing metadata along the file + (such as "Content-Type"). + Unsupported metadata keys will be ignored. + + Returns + ------- + stream : NativeFile + + Examples + -------- + >>> local = fs.LocalFileSystem() + >>> with local.open_output_stream(path) as stream: + ... stream.write(b'data') + 4 + """ + cdef: + c_string pathstr = _path_as_bytes(path) + NativeFile stream = NativeFile() + shared_ptr[COutputStream] out_handle + shared_ptr[const CKeyValueMetadata] c_metadata + + if metadata is not None: + c_metadata = pyarrow_unwrap_metadata(KeyValueMetadata(metadata)) + + with nogil: + out_handle = GetResultValue( + self.fs.OpenOutputStream(pathstr, c_metadata)) + + stream.set_output_stream(out_handle) + stream.is_writable = True + + return self._wrap_output_stream( + stream, path=path, compression=compression, buffer_size=buffer_size + ) + + def open_append_stream(self, path, compression='detect', + buffer_size=None, metadata=None): + """ + Open an output stream for appending. + + If the target doesn't exist, a new empty file is created. + + .. note:: + Some filesystem implementations do not support efficient + appending to an existing file, in which case this method will + raise NotImplementedError. + Consider writing to multiple files (using e.g. the dataset layer) + instead. + + Parameters + ---------- + path : str + The source to open for writing. + compression : str optional, default 'detect' + The compression algorithm to use for on-the-fly compression. + If "detect" and source is a file path, then compression will be + chosen based on the file extension. + If None, no compression will be applied. Otherwise, a well-known + algorithm name must be supplied (e.g. "gzip"). + buffer_size : int optional, default None + If None or 0, no buffering will happen. Otherwise the size of the + temporary write buffer. + metadata : dict optional, default None + If not None, a mapping of string keys to string values. + Some filesystems support storing metadata along the file + (such as "Content-Type"). + Unsupported metadata keys will be ignored. + + Returns + ------- + stream : NativeFile + + Examples + -------- + Append new data to a FileSystem subclass with nonempty file: + + >>> with local.open_append_stream(path) as f: + ... f.write(b'+newly added') + 12 + + Print out the content fo the file: + + >>> with local.open_input_file(path) as f: + ... print(f.readall()) + b'data+newly added' + """ + cdef: + c_string pathstr = _path_as_bytes(path) + NativeFile stream = NativeFile() + shared_ptr[COutputStream] out_handle + shared_ptr[const CKeyValueMetadata] c_metadata + + if metadata is not None: + c_metadata = pyarrow_unwrap_metadata(KeyValueMetadata(metadata)) + + with nogil: + out_handle = GetResultValue( + self.fs.OpenAppendStream(pathstr, c_metadata)) + + stream.set_output_stream(out_handle) + stream.is_writable = True + + return self._wrap_output_stream( + stream, path=path, compression=compression, buffer_size=buffer_size + ) + + def normalize_path(self, path): + """ + Normalize filesystem path. + + Parameters + ---------- + path : str + The path to normalize + + Returns + ------- + normalized_path : str + The normalized path + """ + cdef: + c_string c_path = _path_as_bytes(path) + c_string c_path_normalized + + c_path_normalized = GetResultValue(self.fs.NormalizePath(c_path)) + return frombytes(c_path_normalized) + + +cdef class LocalFileSystem(FileSystem): + """ + A FileSystem implementation accessing files on the local machine. + + Details such as symlinks are abstracted away (symlinks are always followed, + except when deleting an entry). + + Parameters + ---------- + use_mmap : bool, default False + Whether open_input_stream and open_input_file should return + a mmap'ed file or a regular file. + + Examples + -------- + Create a FileSystem object with LocalFileSystem constructor: + + >>> from pyarrow import fs + >>> local = fs.LocalFileSystem() + >>> local + + + and write data on to the file: + + >>> with local.open_output_stream('/tmp/local_fs.dat') as stream: + ... stream.write(b'data') + 4 + >>> with local.open_input_stream('/tmp/local_fs.dat') as stream: + ... print(stream.readall()) + b'data' + + Create a FileSystem object inferred from a URI of the saved file: + + >>> local_new, path = fs.LocalFileSystem().from_uri('/tmp/local_fs.dat') + >>> local_new + >> path + '/tmp/local_fs.dat' + + Check if FileSystems `local` and `local_new` are equal: + + >>> local.equals(local_new) + True + + Compare two different FileSystems: + + >>> local2 = fs.LocalFileSystem(use_mmap=True) + >>> local.equals(local2) + False + + Copy a file and print out the data: + + >>> local.copy_file('/tmp/local_fs.dat', '/tmp/local_fs-copy.dat') + >>> with local.open_input_stream('/tmp/local_fs-copy.dat') as stream: + ... print(stream.readall()) + ... + b'data' + + Open an output stream for appending, add text and print the new data: + + >>> with local.open_append_stream('/tmp/local_fs-copy.dat') as f: + ... f.write(b'+newly added') + 12 + + >>> with local.open_input_stream('/tmp/local_fs-copy.dat') as f: + ... print(f.readall()) + b'data+newly added' + + Create a directory, copy a file into it and then delete the whole directory: + + >>> local.create_dir('/tmp/new_folder') + >>> local.copy_file('/tmp/local_fs.dat', '/tmp/new_folder/local_fs.dat') + >>> local.get_file_info('/tmp/new_folder') + + >>> local.delete_dir('/tmp/new_folder') + >>> local.get_file_info('/tmp/new_folder') + + + Create a directory, copy a file into it and then delete + the content of the directory: + + >>> local.create_dir('/tmp/new_folder') + >>> local.copy_file('/tmp/local_fs.dat', '/tmp/new_folder/local_fs.dat') + >>> local.get_file_info('/tmp/new_folder/local_fs.dat') + + >>> local.delete_dir_contents('/tmp/new_folder') + >>> local.get_file_info('/tmp/new_folder') + + >>> local.get_file_info('/tmp/new_folder/local_fs.dat') + + + Create a directory, copy a file into it and then delete + the file from the directory: + + >>> local.create_dir('/tmp/new_folder') + >>> local.copy_file('/tmp/local_fs.dat', '/tmp/new_folder/local_fs.dat') + >>> local.delete_file('/tmp/new_folder/local_fs.dat') + >>> local.get_file_info('/tmp/new_folder/local_fs.dat') + + >>> local.get_file_info('/tmp/new_folder') + + + Move the file: + + >>> local.move('/tmp/local_fs-copy.dat', '/tmp/new_folder/local_fs-copy.dat') + >>> local.get_file_info('/tmp/new_folder/local_fs-copy.dat') + + >>> local.get_file_info('/tmp/local_fs-copy.dat') + + + To finish delete the file left: + >>> local.delete_file('/tmp/local_fs.dat') + """ + + def __init__(self, *, use_mmap=False): + cdef: + shared_ptr[CFileSystem] fs + c_string c_uri + + # from_uri needs a non-empty path, so just use a placeholder of /_ + c_uri = tobytes(f"file:///_?use_mmap={int(use_mmap)}") + with nogil: + fs = GetResultValue(CFileSystemFromUri(c_uri)) + self.init( fs) + + def __reduce__(self): + uri = frombytes(GetResultValue(self.fs.MakeUri(b"/_"))) + return FileSystem._from_uri, (uri,) + + +cdef class SubTreeFileSystem(FileSystem): + """ + Delegates to another implementation after prepending a fixed base path. + + This is useful to expose a logical view of a subtree of a filesystem, + for example a directory in a LocalFileSystem. + + Note, that this makes no security guarantee. For example, symlinks may + allow to "escape" the subtree and access other parts of the underlying + filesystem. + + Parameters + ---------- + base_path : str + The root of the subtree. + base_fs : FileSystem + FileSystem object the operations delegated to. + + Examples + -------- + Create a LocalFileSystem instance: + + >>> from pyarrow import fs + >>> local = fs.LocalFileSystem() + >>> with local.open_output_stream('/tmp/local_fs.dat') as stream: + ... stream.write(b'data') + 4 + + Create a directory and a SubTreeFileSystem instance: + + >>> local.create_dir('/tmp/sub_tree') + >>> subtree = fs.SubTreeFileSystem('/tmp/sub_tree', local) + + Write data into the existing file: + + >>> with subtree.open_append_stream('sub_tree_fs.dat') as f: + ... f.write(b'+newly added') + 12 + + Print out the attributes: + + >>> subtree.base_fs + + >>> subtree.base_path + '/tmp/sub_tree/' + + Get info for the given directory or given file: + + >>> subtree.get_file_info('') + + >>> subtree.get_file_info('sub_tree_fs.dat') + + + Delete the file and directory: + + >>> subtree.delete_file('sub_tree_fs.dat') + >>> local.delete_dir('/tmp/sub_tree') + >>> local.delete_file('/tmp/local_fs.dat') + + For usage of the methods see examples for :func:`~pyarrow.fs.LocalFileSystem`. + """ + + def __init__(self, base_path, FileSystem base_fs): + cdef: + c_string pathstr + shared_ptr[CSubTreeFileSystem] wrapped + + pathstr = _path_as_bytes(base_path) + wrapped = make_shared[CSubTreeFileSystem](pathstr, base_fs.wrapped) + + self.init( wrapped) + + cdef init(self, const shared_ptr[CFileSystem]& wrapped): + FileSystem.init(self, wrapped) + self.subtreefs = wrapped.get() + + def __repr__(self): + return ("SubTreeFileSystem(base_path={}, base_fs={}" + .format(self.base_path, self.base_fs)) + + def __reduce__(self): + return SubTreeFileSystem, ( + frombytes(self.subtreefs.base_path()), + FileSystem.wrap(self.subtreefs.base_fs()) + ) + + @property + def base_path(self): + return frombytes(self.subtreefs.base_path()) + + @property + def base_fs(self): + return FileSystem.wrap(self.subtreefs.base_fs()) + + +cdef class _MockFileSystem(FileSystem): + + def __init__(self, datetime current_time=None): + cdef shared_ptr[CMockFileSystem] wrapped + + current_time = current_time or datetime.now() + wrapped = make_shared[CMockFileSystem]( + PyDateTime_to_TimePoint( current_time) + ) + + self.init( wrapped) + + cdef init(self, const shared_ptr[CFileSystem]& wrapped): + FileSystem.init(self, wrapped) + self.mockfs = wrapped.get() + + +cdef class PyFileSystem(FileSystem): + """ + A FileSystem with behavior implemented in Python. + + Parameters + ---------- + handler : FileSystemHandler + The handler object implementing custom filesystem behavior. + + Examples + -------- + Create an fsspec-based filesystem object for GitHub: + + >>> from fsspec.implementations import github + >>> gfs = github.GithubFileSystem('apache', 'arrow') # doctest: +SKIP + + Get a PyArrow FileSystem object: + + >>> from pyarrow.fs import PyFileSystem, FSSpecHandler + >>> pa_fs = PyFileSystem(FSSpecHandler(gfs)) # doctest: +SKIP + + Use :func:`~pyarrow.fs.FileSystem` functionality ``get_file_info()``: + + >>> pa_fs.get_file_info('README.md') # doctest: +SKIP + + """ + + def __init__(self, handler): + cdef: + CPyFileSystemVtable vtable + shared_ptr[CPyFileSystem] wrapped + + if not isinstance(handler, FileSystemHandler): + raise TypeError("Expected a FileSystemHandler instance, got {0}" + .format(type(handler))) + + vtable.get_type_name = _cb_get_type_name + vtable.equals = _cb_equals + vtable.get_file_info = _cb_get_file_info + vtable.get_file_info_vector = _cb_get_file_info_vector + vtable.get_file_info_selector = _cb_get_file_info_selector + vtable.create_dir = _cb_create_dir + vtable.delete_dir = _cb_delete_dir + vtable.delete_dir_contents = _cb_delete_dir_contents + vtable.delete_root_dir_contents = _cb_delete_root_dir_contents + vtable.delete_file = _cb_delete_file + vtable.move = _cb_move + vtable.copy_file = _cb_copy_file + vtable.open_input_stream = _cb_open_input_stream + vtable.open_input_file = _cb_open_input_file + vtable.open_output_stream = _cb_open_output_stream + vtable.open_append_stream = _cb_open_append_stream + vtable.normalize_path = _cb_normalize_path + + wrapped = CPyFileSystem.Make(handler, move(vtable)) + self.init( wrapped) + + cdef init(self, const shared_ptr[CFileSystem]& wrapped): + FileSystem.init(self, wrapped) + self.pyfs = wrapped.get() + + @property + def handler(self): + """ + The filesystem's underlying handler. + + Returns + ------- + handler : FileSystemHandler + """ + return self.pyfs.handler() + + def __reduce__(self): + return PyFileSystem, (self.handler,) + + +class FileSystemHandler(ABC): + """ + An abstract class exposing methods to implement PyFileSystem's behavior. + """ + + @abstractmethod + def get_type_name(self): + """ + Implement PyFileSystem.type_name. + """ + + @abstractmethod + def get_file_info(self, paths): + """ + Implement PyFileSystem.get_file_info(paths). + + Parameters + ---------- + paths : list of str + paths for which we want to retrieve the info. + """ + + @abstractmethod + def get_file_info_selector(self, selector): + """ + Implement PyFileSystem.get_file_info(selector). + + Parameters + ---------- + selector : FileSelector + selector for which we want to retrieve the info. + """ + + @abstractmethod + def create_dir(self, path, recursive): + """ + Implement PyFileSystem.create_dir(...). + + Parameters + ---------- + path : str + path of the directory. + recursive : bool + if the parent directories should be created too. + """ + + @abstractmethod + def delete_dir(self, path): + """ + Implement PyFileSystem.delete_dir(...). + + Parameters + ---------- + path : str + path of the directory. + """ + + @abstractmethod + def delete_dir_contents(self, path, missing_dir_ok=False): + """ + Implement PyFileSystem.delete_dir_contents(...). + + Parameters + ---------- + path : str + path of the directory. + missing_dir_ok : bool + if False an error should be raised if path does not exist + """ + + @abstractmethod + def delete_root_dir_contents(self): + """ + Implement PyFileSystem.delete_dir_contents("/", accept_root_dir=True). + """ + + @abstractmethod + def delete_file(self, path): + """ + Implement PyFileSystem.delete_file(...). + + Parameters + ---------- + path : str + path of the file. + """ + + @abstractmethod + def move(self, src, dest): + """ + Implement PyFileSystem.move(...). + + Parameters + ---------- + src : str + path of what should be moved. + dest : str + path of where it should be moved to. + """ + + @abstractmethod + def copy_file(self, src, dest): + """ + Implement PyFileSystem.copy_file(...). + + Parameters + ---------- + src : str + path of what should be copied. + dest : str + path of where it should be copied to. + """ + + @abstractmethod + def open_input_stream(self, path): + """ + Implement PyFileSystem.open_input_stream(...). + + Parameters + ---------- + path : str + path of what should be opened. + """ + + @abstractmethod + def open_input_file(self, path): + """ + Implement PyFileSystem.open_input_file(...). + + Parameters + ---------- + path : str + path of what should be opened. + """ + + @abstractmethod + def open_output_stream(self, path, metadata): + """ + Implement PyFileSystem.open_output_stream(...). + + Parameters + ---------- + path : str + path of what should be opened. + metadata : mapping + Mapping of string keys to string values. + Some filesystems support storing metadata along the file + (such as "Content-Type"). + """ + + @abstractmethod + def open_append_stream(self, path, metadata): + """ + Implement PyFileSystem.open_append_stream(...). + + Parameters + ---------- + path : str + path of what should be opened. + metadata : mapping + Mapping of string keys to string values. + Some filesystems support storing metadata along the file + (such as "Content-Type"). + """ + + @abstractmethod + def normalize_path(self, path): + """ + Implement PyFileSystem.normalize_path(...). + + Parameters + ---------- + path : str + path of what should be normalized. + """ + +# Callback definitions for CPyFileSystemVtable + + +cdef void _cb_get_type_name(handler, c_string* out) except *: + out[0] = tobytes("py::" + handler.get_type_name()) + +cdef c_bool _cb_equals(handler, const CFileSystem& c_other) except False: + if c_other.type_name().startswith(b"py::"): + return ( c_other).handler() == handler + + return False + +cdef void _cb_get_file_info(handler, const c_string& path, + CFileInfo* out) except *: + infos = handler.get_file_info([frombytes(path)]) + if not isinstance(infos, list) or len(infos) != 1: + raise TypeError("get_file_info should have returned a 1-element list") + out[0] = FileInfo.unwrap_safe(infos[0]) + +cdef void _cb_get_file_info_vector(handler, const vector[c_string]& paths, + vector[CFileInfo]* out) except *: + py_paths = [frombytes(paths[i]) for i in range(len(paths))] + infos = handler.get_file_info(py_paths) + if not isinstance(infos, list): + raise TypeError("get_file_info should have returned a list") + out[0].clear() + out[0].reserve(len(infos)) + for info in infos: + out[0].push_back(FileInfo.unwrap_safe(info)) + +cdef void _cb_get_file_info_selector(handler, const CFileSelector& selector, + vector[CFileInfo]* out) except *: + infos = handler.get_file_info_selector(FileSelector.wrap(selector)) + if not isinstance(infos, list): + raise TypeError("get_file_info_selector should have returned a list") + out[0].clear() + out[0].reserve(len(infos)) + for info in infos: + out[0].push_back(FileInfo.unwrap_safe(info)) + +cdef void _cb_create_dir(handler, const c_string& path, + c_bool recursive) except *: + handler.create_dir(frombytes(path), recursive) + +cdef void _cb_delete_dir(handler, const c_string& path) except *: + handler.delete_dir(frombytes(path)) + +cdef void _cb_delete_dir_contents(handler, const c_string& path, + c_bool missing_dir_ok) except *: + handler.delete_dir_contents(frombytes(path), missing_dir_ok) + +cdef void _cb_delete_root_dir_contents(handler) except *: + handler.delete_root_dir_contents() + +cdef void _cb_delete_file(handler, const c_string& path) except *: + handler.delete_file(frombytes(path)) + +cdef void _cb_move(handler, const c_string& src, + const c_string& dest) except *: + handler.move(frombytes(src), frombytes(dest)) + +cdef void _cb_copy_file(handler, const c_string& src, + const c_string& dest) except *: + handler.copy_file(frombytes(src), frombytes(dest)) + +cdef void _cb_open_input_stream(handler, const c_string& path, + shared_ptr[CInputStream]* out) except *: + stream = handler.open_input_stream(frombytes(path)) + if not isinstance(stream, NativeFile): + raise TypeError("open_input_stream should have returned " + "a PyArrow file") + out[0] = ( stream).get_input_stream() + +cdef void _cb_open_input_file(handler, const c_string& path, + shared_ptr[CRandomAccessFile]* out) except *: + stream = handler.open_input_file(frombytes(path)) + if not isinstance(stream, NativeFile): + raise TypeError("open_input_file should have returned " + "a PyArrow file") + out[0] = ( stream).get_random_access_file() + +cdef void _cb_open_output_stream( + handler, const c_string& path, + const shared_ptr[const CKeyValueMetadata]& metadata, + shared_ptr[COutputStream]* out) except *: + stream = handler.open_output_stream( + frombytes(path), pyarrow_wrap_metadata(metadata)) + if not isinstance(stream, NativeFile): + raise TypeError("open_output_stream should have returned " + "a PyArrow file") + out[0] = ( stream).get_output_stream() + +cdef void _cb_open_append_stream( + handler, const c_string& path, + const shared_ptr[const CKeyValueMetadata]& metadata, + shared_ptr[COutputStream]* out) except *: + stream = handler.open_append_stream( + frombytes(path), pyarrow_wrap_metadata(metadata)) + if not isinstance(stream, NativeFile): + raise TypeError("open_append_stream should have returned " + "a PyArrow file") + out[0] = ( stream).get_output_stream() + +cdef void _cb_normalize_path(handler, const c_string& path, + c_string* out) except *: + out[0] = tobytes(handler.normalize_path(frombytes(path))) + + +def _copy_files(FileSystem source_fs, str source_path, + FileSystem destination_fs, str destination_path, + int64_t chunk_size, c_bool use_threads): + # low-level helper exposed through pyarrow/fs.py::copy_files + cdef: + CFileLocator c_source + vector[CFileLocator] c_sources + CFileLocator c_destination + vector[CFileLocator] c_destinations + + c_source.filesystem = source_fs.unwrap() + c_source.path = tobytes(source_path) + c_sources.push_back(c_source) + + c_destination.filesystem = destination_fs.unwrap() + c_destination.path = tobytes(destination_path) + c_destinations.push_back(c_destination) + + with nogil: + check_status(CCopyFiles( + c_sources, c_destinations, + c_default_io_context(), chunk_size, use_threads, + )) + + +def _copy_files_selector(FileSystem source_fs, FileSelector source_sel, + FileSystem destination_fs, str destination_base_dir, + int64_t chunk_size, c_bool use_threads): + # low-level helper exposed through pyarrow/fs.py::copy_files + cdef c_string c_destination_base_dir = tobytes(destination_base_dir) + + with nogil: + check_status(CCopyFilesWithSelector( + source_fs.unwrap(), source_sel.unwrap(), + destination_fs.unwrap(), c_destination_base_dir, + c_default_io_context(), chunk_size, use_threads, + )) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_hdfs.cpython-310-x86_64-linux-gnu.so b/vllm/lib/python3.10/site-packages/pyarrow/_hdfs.cpython-310-x86_64-linux-gnu.so new file mode 100644 index 0000000000000000000000000000000000000000..8011203548bd7e5be89cb31b6b1846831b32b663 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_hdfs.cpython-310-x86_64-linux-gnu.so @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe1f5de542566e82c277c8d849336c3c386fdbca748d2cdf1051218ef8562b4a +size 131816 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_parquet.pyx b/vllm/lib/python3.10/site-packages/pyarrow/_parquet.pyx new file mode 100644 index 0000000000000000000000000000000000000000..254bfe3b09a9cd51d5c8d3207d5b59a9b5bdcd3d --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_parquet.pyx @@ -0,0 +1,2266 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: profile=False +# distutils: language = c++ + +from collections.abc import Sequence +from textwrap import indent +import warnings + +from cython.operator cimport dereference as deref +from pyarrow.includes.common cimport * +from pyarrow.includes.libarrow cimport * +from pyarrow.includes.libarrow_python cimport * +from pyarrow.lib cimport (_Weakrefable, Buffer, Schema, + check_status, + MemoryPool, maybe_unbox_memory_pool, + Table, KeyValueMetadata, + pyarrow_wrap_chunked_array, + pyarrow_wrap_schema, + pyarrow_unwrap_metadata, + pyarrow_unwrap_schema, + pyarrow_wrap_table, + pyarrow_wrap_batch, + pyarrow_wrap_scalar, + NativeFile, get_reader, get_writer, + string_to_timeunit) + +from pyarrow.lib import (ArrowException, NativeFile, BufferOutputStream, + _stringify_path, + tobytes, frombytes, is_threading_enabled) + +cimport cpython as cp + +_DEFAULT_ROW_GROUP_SIZE = 1024*1024 +_MAX_ROW_GROUP_SIZE = 64*1024*1024 + +cdef class Statistics(_Weakrefable): + """Statistics for a single column in a single row group.""" + + def __cinit__(self): + pass + + def __repr__(self): + return """{} + has_min_max: {} + min: {} + max: {} + null_count: {} + distinct_count: {} + num_values: {} + physical_type: {} + logical_type: {} + converted_type (legacy): {}""".format(object.__repr__(self), + self.has_min_max, + self.min, + self.max, + self.null_count, + self.distinct_count, + self.num_values, + self.physical_type, + str(self.logical_type), + self.converted_type) + + def to_dict(self): + """ + Get dictionary representation of statistics. + + Returns + ------- + dict + Dictionary with a key for each attribute of this class. + """ + d = dict( + has_min_max=self.has_min_max, + min=self.min, + max=self.max, + null_count=self.null_count, + distinct_count=self.distinct_count, + num_values=self.num_values, + physical_type=self.physical_type + ) + return d + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return NotImplemented + + def equals(self, Statistics other): + """ + Return whether the two column statistics objects are equal. + + Parameters + ---------- + other : Statistics + Statistics to compare against. + + Returns + ------- + are_equal : bool + """ + return self.statistics.get().Equals(deref(other.statistics.get())) + + @property + def has_min_max(self): + """Whether min and max are present (bool).""" + return self.statistics.get().HasMinMax() + + @property + def has_null_count(self): + """Whether null count is present (bool).""" + return self.statistics.get().HasNullCount() + + @property + def has_distinct_count(self): + """Whether distinct count is preset (bool).""" + return self.statistics.get().HasDistinctCount() + + @property + def min_raw(self): + """Min value as physical type (bool, int, float, or bytes).""" + if self.has_min_max: + return _cast_statistic_raw_min(self.statistics.get()) + else: + return None + + @property + def max_raw(self): + """Max value as physical type (bool, int, float, or bytes).""" + if self.has_min_max: + return _cast_statistic_raw_max(self.statistics.get()) + else: + return None + + @property + def min(self): + """ + Min value as logical type. + + Returned as the Python equivalent of logical type, such as datetime.date + for dates and decimal.Decimal for decimals. + """ + if self.has_min_max: + min_scalar, _ = _cast_statistics(self.statistics.get()) + return min_scalar.as_py() + else: + return None + + @property + def max(self): + """ + Max value as logical type. + + Returned as the Python equivalent of logical type, such as datetime.date + for dates and decimal.Decimal for decimals. + """ + if self.has_min_max: + _, max_scalar = _cast_statistics(self.statistics.get()) + return max_scalar.as_py() + else: + return None + + @property + def null_count(self): + """Number of null values in chunk (int).""" + if self.has_null_count: + return self.statistics.get().null_count() + else: + return None + + @property + def distinct_count(self): + """Distinct number of values in chunk (int).""" + if self.has_distinct_count: + return self.statistics.get().distinct_count() + else: + return None + + @property + def num_values(self): + """Number of non-null values (int).""" + return self.statistics.get().num_values() + + @property + def physical_type(self): + """Physical type of column (str).""" + raw_physical_type = self.statistics.get().physical_type() + return physical_type_name_from_enum(raw_physical_type) + + @property + def logical_type(self): + """Logical type of column (:class:`ParquetLogicalType`).""" + return wrap_logical_type(self.statistics.get().descr().logical_type()) + + @property + def converted_type(self): + """Legacy converted type (str or None).""" + raw_converted_type = self.statistics.get().descr().converted_type() + return converted_type_name_from_enum(raw_converted_type) + + +cdef class ParquetLogicalType(_Weakrefable): + """Logical type of parquet type.""" + cdef: + shared_ptr[const CParquetLogicalType] type + + def __cinit__(self): + pass + + cdef init(self, const shared_ptr[const CParquetLogicalType]& type): + self.type = type + + def __repr__(self): + return "{}\n {}".format(object.__repr__(self), str(self)) + + def __str__(self): + return frombytes(self.type.get().ToString(), safe=True) + + def to_json(self): + """ + Get a JSON string containing type and type parameters. + + Returns + ------- + json : str + JSON representation of type, with at least a field called 'Type' + which contains the type name. If the type is parameterized, such + as a decimal with scale and precision, will contain those as fields + as well. + """ + return frombytes(self.type.get().ToJSON()) + + @property + def type(self): + """Name of the logical type (str).""" + return logical_type_name_from_enum(self.type.get().type()) + + +cdef wrap_logical_type(const shared_ptr[const CParquetLogicalType]& type): + cdef ParquetLogicalType out = ParquetLogicalType() + out.init(type) + return out + + +cdef _cast_statistic_raw_min(CStatistics* statistics): + cdef ParquetType physical_type = statistics.physical_type() + cdef uint32_t type_length = statistics.descr().type_length() + if physical_type == ParquetType_BOOLEAN: + return ( statistics).min() + elif physical_type == ParquetType_INT32: + return ( statistics).min() + elif physical_type == ParquetType_INT64: + return ( statistics).min() + elif physical_type == ParquetType_FLOAT: + return ( statistics).min() + elif physical_type == ParquetType_DOUBLE: + return ( statistics).min() + elif physical_type == ParquetType_BYTE_ARRAY: + return _box_byte_array(( statistics).min()) + elif physical_type == ParquetType_FIXED_LEN_BYTE_ARRAY: + return _box_flba(( statistics).min(), type_length) + + +cdef _cast_statistic_raw_max(CStatistics* statistics): + cdef ParquetType physical_type = statistics.physical_type() + cdef uint32_t type_length = statistics.descr().type_length() + if physical_type == ParquetType_BOOLEAN: + return ( statistics).max() + elif physical_type == ParquetType_INT32: + return ( statistics).max() + elif physical_type == ParquetType_INT64: + return ( statistics).max() + elif physical_type == ParquetType_FLOAT: + return ( statistics).max() + elif physical_type == ParquetType_DOUBLE: + return ( statistics).max() + elif physical_type == ParquetType_BYTE_ARRAY: + return _box_byte_array(( statistics).max()) + elif physical_type == ParquetType_FIXED_LEN_BYTE_ARRAY: + return _box_flba(( statistics).max(), type_length) + + +cdef _cast_statistics(CStatistics* statistics): + cdef: + shared_ptr[CScalar] c_min + shared_ptr[CScalar] c_max + check_status(StatisticsAsScalars(statistics[0], &c_min, &c_max)) + return (pyarrow_wrap_scalar(c_min), pyarrow_wrap_scalar(c_max)) + + +cdef _box_byte_array(ParquetByteArray val): + return cp.PyBytes_FromStringAndSize( val.ptr, val.len) + + +cdef _box_flba(ParquetFLBA val, uint32_t len): + return cp.PyBytes_FromStringAndSize( val.ptr, len) + + +cdef class ColumnChunkMetaData(_Weakrefable): + """Column metadata for a single row group.""" + + def __cinit__(self): + pass + + def __repr__(self): + statistics = indent(repr(self.statistics), 4 * ' ') + return """{0} + file_offset: {1} + file_path: {2} + physical_type: {3} + num_values: {4} + path_in_schema: {5} + is_stats_set: {6} + statistics: +{7} + compression: {8} + encodings: {9} + has_dictionary_page: {10} + dictionary_page_offset: {11} + data_page_offset: {12} + total_compressed_size: {13} + total_uncompressed_size: {14}""".format(object.__repr__(self), + self.file_offset, + self.file_path, + self.physical_type, + self.num_values, + self.path_in_schema, + self.is_stats_set, + statistics, + self.compression, + self.encodings, + self.has_dictionary_page, + self.dictionary_page_offset, + self.data_page_offset, + self.total_compressed_size, + self.total_uncompressed_size) + + def to_dict(self): + """ + Get dictionary representation of the column chunk metadata. + + Returns + ------- + dict + Dictionary with a key for each attribute of this class. + """ + statistics = self.statistics.to_dict() if self.is_stats_set else None + d = dict( + file_offset=self.file_offset, + file_path=self.file_path, + physical_type=self.physical_type, + num_values=self.num_values, + path_in_schema=self.path_in_schema, + is_stats_set=self.is_stats_set, + statistics=statistics, + compression=self.compression, + encodings=self.encodings, + has_dictionary_page=self.has_dictionary_page, + dictionary_page_offset=self.dictionary_page_offset, + data_page_offset=self.data_page_offset, + total_compressed_size=self.total_compressed_size, + total_uncompressed_size=self.total_uncompressed_size + ) + return d + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return NotImplemented + + def equals(self, ColumnChunkMetaData other): + """ + Return whether the two column chunk metadata objects are equal. + + Parameters + ---------- + other : ColumnChunkMetaData + Metadata to compare against. + + Returns + ------- + are_equal : bool + """ + return self.metadata.Equals(deref(other.metadata)) + + @property + def file_offset(self): + """Offset into file where column chunk is located (int).""" + return self.metadata.file_offset() + + @property + def file_path(self): + """Optional file path if set (str or None).""" + return frombytes(self.metadata.file_path()) + + @property + def physical_type(self): + """Physical type of column (str).""" + return physical_type_name_from_enum(self.metadata.type()) + + @property + def num_values(self): + """Total number of values (int).""" + return self.metadata.num_values() + + @property + def path_in_schema(self): + """Nested path to field, separated by periods (str).""" + path = self.metadata.path_in_schema().get().ToDotString() + return frombytes(path) + + @property + def is_stats_set(self): + """Whether or not statistics are present in metadata (bool).""" + return self.metadata.is_stats_set() + + @property + def statistics(self): + """Statistics for column chunk (:class:`Statistics`).""" + if not self.metadata.is_stats_set(): + return None + statistics = Statistics() + statistics.init(self.metadata.statistics(), self) + return statistics + + @property + def compression(self): + """ + Type of compression used for column (str). + + One of 'UNCOMPRESSED', 'SNAPPY', 'GZIP', 'LZO', 'BROTLI', 'LZ4', 'ZSTD', + or 'UNKNOWN'. + """ + return compression_name_from_enum(self.metadata.compression()) + + @property + def encodings(self): + """ + Encodings used for column (tuple of str). + + One of 'PLAIN', 'BIT_PACKED', 'RLE', 'BYTE_STREAM_SPLIT', 'DELTA_BINARY_PACKED', + 'DELTA_LENGTH_BYTE_ARRAY', 'DELTA_BYTE_ARRAY'. + """ + return tuple(map(encoding_name_from_enum, self.metadata.encodings())) + + @property + def has_dictionary_page(self): + """Whether there is dictionary data present in the column chunk (bool).""" + return bool(self.metadata.has_dictionary_page()) + + @property + def dictionary_page_offset(self): + """Offset of dictionary page relative to column chunk offset (int).""" + if self.has_dictionary_page: + return self.metadata.dictionary_page_offset() + else: + return None + + @property + def data_page_offset(self): + """Offset of data page relative to column chunk offset (int).""" + return self.metadata.data_page_offset() + + @property + def has_index_page(self): + """Not yet supported.""" + raise NotImplementedError('not supported in parquet-cpp') + + @property + def index_page_offset(self): + """Not yet supported.""" + raise NotImplementedError("parquet-cpp doesn't return valid values") + + @property + def total_compressed_size(self): + """Compressed size in bytes (int).""" + return self.metadata.total_compressed_size() + + @property + def total_uncompressed_size(self): + """Uncompressed size in bytes (int).""" + return self.metadata.total_uncompressed_size() + + @property + def has_offset_index(self): + """Whether the column chunk has an offset index""" + return self.metadata.GetOffsetIndexLocation().has_value() + + @property + def has_column_index(self): + """Whether the column chunk has a column index""" + return self.metadata.GetColumnIndexLocation().has_value() + + @property + def metadata(self): + """Additional metadata as key value pairs (dict[bytes, bytes]).""" + cdef: + unordered_map[c_string, c_string] metadata + const CKeyValueMetadata* underlying_metadata + underlying_metadata = self.metadata.key_value_metadata().get() + if underlying_metadata != NULL: + underlying_metadata.ToUnorderedMap(&metadata) + return metadata + else: + return None + + +cdef class SortingColumn: + """ + Sorting specification for a single column. + + Returned by :meth:`RowGroupMetaData.sorting_columns` and used in + :class:`ParquetWriter` to specify the sort order of the data. + + Parameters + ---------- + column_index : int + Index of column that data is sorted by. + descending : bool, default False + Whether column is sorted in descending order. + nulls_first : bool, default False + Whether null values appear before valid values. + + Notes + ----- + + Column indices are zero-based, refer only to leaf fields, and are in + depth-first order. This may make the column indices for nested schemas + different from what you expect. In most cases, it will be easier to + specify the sort order using column names instead of column indices + and converting using the ``from_ordering`` method. + + Examples + -------- + + In other APIs, sort order is specified by names, such as: + + >>> sort_order = [('id', 'ascending'), ('timestamp', 'descending')] + + For Parquet, the column index must be used instead: + + >>> import pyarrow.parquet as pq + >>> [pq.SortingColumn(0), pq.SortingColumn(1, descending=True)] + [SortingColumn(column_index=0, descending=False, nulls_first=False), SortingColumn(column_index=1, descending=True, nulls_first=False)] + + Convert the sort_order into the list of sorting columns with + ``from_ordering`` (note that the schema must be provided as well): + + >>> import pyarrow as pa + >>> schema = pa.schema([('id', pa.int64()), ('timestamp', pa.timestamp('ms'))]) + >>> sorting_columns = pq.SortingColumn.from_ordering(schema, sort_order) + >>> sorting_columns + (SortingColumn(column_index=0, descending=False, nulls_first=False), SortingColumn(column_index=1, descending=True, nulls_first=False)) + + Convert back to the sort order with ``to_ordering``: + + >>> pq.SortingColumn.to_ordering(schema, sorting_columns) + ((('id', 'ascending'), ('timestamp', 'descending')), 'at_end') + + See Also + -------- + RowGroupMetaData.sorting_columns + """ + cdef int column_index + cdef c_bool descending + cdef c_bool nulls_first + + def __init__(self, int column_index, c_bool descending=False, c_bool nulls_first=False): + self.column_index = column_index + self.descending = descending + self.nulls_first = nulls_first + + @classmethod + def from_ordering(cls, Schema schema, sort_keys, null_placement='at_end'): + """ + Create a tuple of SortingColumn objects from the same arguments as + :class:`pyarrow.compute.SortOptions`. + + Parameters + ---------- + schema : Schema + Schema of the input data. + sort_keys : Sequence of (name, order) tuples + Names of field/column keys (str) to sort the input on, + along with the order each field/column is sorted in. + Accepted values for `order` are "ascending", "descending". + null_placement : {'at_start', 'at_end'}, default 'at_end' + Where null values should appear in the sort order. + + Returns + ------- + sorting_columns : tuple of SortingColumn + """ + if null_placement == 'at_start': + nulls_first = True + elif null_placement == 'at_end': + nulls_first = False + else: + raise ValueError('null_placement must be "at_start" or "at_end"') + + col_map = _name_to_index_map(schema) + + sorting_columns = [] + + for sort_key in sort_keys: + if isinstance(sort_key, str): + name = sort_key + descending = False + elif (isinstance(sort_key, tuple) and len(sort_key) == 2 and + isinstance(sort_key[0], str) and + isinstance(sort_key[1], str)): + name, descending = sort_key + if descending == "descending": + descending = True + elif descending == "ascending": + descending = False + else: + raise ValueError("Invalid sort key direction: {0}" + .format(descending)) + else: + raise ValueError("Invalid sort key: {0}".format(sort_key)) + + try: + column_index = col_map[name] + except KeyError: + raise ValueError("Sort key name '{0}' not found in schema:\n{1}" + .format(name, schema)) + + sorting_columns.append( + cls(column_index, descending=descending, nulls_first=nulls_first) + ) + + return tuple(sorting_columns) + + @staticmethod + def to_ordering(Schema schema, sorting_columns): + """ + Convert a tuple of SortingColumn objects to the same format as + :class:`pyarrow.compute.SortOptions`. + + Parameters + ---------- + schema : Schema + Schema of the input data. + sorting_columns : tuple of SortingColumn + Columns to sort the input on. + + Returns + ------- + sort_keys : tuple of (name, order) tuples + null_placement : {'at_start', 'at_end'} + """ + col_map = {i: name for name, i in _name_to_index_map(schema).items()} + + sort_keys = [] + nulls_first = None + + for sorting_column in sorting_columns: + name = col_map[sorting_column.column_index] + if sorting_column.descending: + order = "descending" + else: + order = "ascending" + sort_keys.append((name, order)) + if nulls_first is None: + nulls_first = sorting_column.nulls_first + elif nulls_first != sorting_column.nulls_first: + raise ValueError("Sorting columns have inconsistent null placement") + + if nulls_first: + null_placement = "at_start" + else: + null_placement = "at_end" + + return tuple(sort_keys), null_placement + + def __repr__(self): + return """{}(column_index={}, descending={}, nulls_first={})""".format( + self.__class__.__name__, + self.column_index, self.descending, self.nulls_first) + + def __eq__(self, SortingColumn other): + return (self.column_index == other.column_index and + self.descending == other.descending and + self.nulls_first == other.nulls_first) + + def __hash__(self): + return hash((self.column_index, self.descending, self.nulls_first)) + + @property + def column_index(self): + """"Index of column data is sorted by (int).""" + return self.column_index + + @property + def descending(self): + """Whether column is sorted in descending order (bool).""" + return self.descending + + @property + def nulls_first(self): + """Whether null values appear before valid values (bool).""" + return self.nulls_first + + def to_dict(self): + """ + Get dictionary representation of the SortingColumn. + + Returns + ------- + dict + Dictionary with a key for each attribute of this class. + """ + d = dict( + column_index=self.column_index, + descending=self.descending, + nulls_first=self.nulls_first + ) + return d + + +cdef class RowGroupMetaData(_Weakrefable): + """Metadata for a single row group.""" + + def __cinit__(self, FileMetaData parent, int index): + if index < 0 or index >= parent.num_row_groups: + raise IndexError('{0} out of bounds'.format(index)) + self.up_metadata = parent._metadata.RowGroup(index) + self.metadata = self.up_metadata.get() + self.parent = parent + self.index = index + + def __reduce__(self): + return RowGroupMetaData, (self.parent, self.index) + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return NotImplemented + + def equals(self, RowGroupMetaData other): + """ + Return whether the two row group metadata objects are equal. + + Parameters + ---------- + other : RowGroupMetaData + Metadata to compare against. + + Returns + ------- + are_equal : bool + """ + return self.metadata.Equals(deref(other.metadata)) + + def column(self, int i): + """ + Get column metadata at given index. + + Parameters + ---------- + i : int + Index of column to get metadata for. + + Returns + ------- + ColumnChunkMetaData + Metadata for column within this chunk. + """ + if i < 0 or i >= self.num_columns: + raise IndexError('{0} out of bounds'.format(i)) + chunk = ColumnChunkMetaData() + chunk.init(self, i) + return chunk + + def __repr__(self): + return """{0} + num_columns: {1} + num_rows: {2} + total_byte_size: {3} + sorting_columns: {4}""".format(object.__repr__(self), + self.num_columns, + self.num_rows, + self.total_byte_size, + self.sorting_columns) + + def to_dict(self): + """ + Get dictionary representation of the row group metadata. + + Returns + ------- + dict + Dictionary with a key for each attribute of this class. + """ + columns = [] + d = dict( + num_columns=self.num_columns, + num_rows=self.num_rows, + total_byte_size=self.total_byte_size, + columns=columns, + sorting_columns=[col.to_dict() for col in self.sorting_columns] + ) + for i in range(self.num_columns): + columns.append(self.column(i).to_dict()) + return d + + @property + def num_columns(self): + """Number of columns in this row group (int).""" + return self.metadata.num_columns() + + @property + def num_rows(self): + """Number of rows in this row group (int).""" + return self.metadata.num_rows() + + @property + def total_byte_size(self): + """Total byte size of all the uncompressed column data in this row group (int).""" + return self.metadata.total_byte_size() + + @property + def sorting_columns(self): + """Columns the row group is sorted by (tuple of :class:`SortingColumn`)).""" + out = [] + cdef vector[CSortingColumn] sorting_columns = self.metadata.sorting_columns() + for sorting_col in sorting_columns: + out.append(SortingColumn( + sorting_col.column_idx, + sorting_col.descending, + sorting_col.nulls_first + )) + return tuple(out) + + +def _reconstruct_filemetadata(Buffer serialized): + cdef: + FileMetaData metadata = FileMetaData.__new__(FileMetaData) + CBuffer *buffer = serialized.buffer.get() + uint32_t metadata_len = buffer.size() + + metadata.init(CFileMetaData_Make(buffer.data(), &metadata_len)) + + return metadata + + +cdef class FileMetaData(_Weakrefable): + """Parquet metadata for a single file.""" + + def __cinit__(self): + pass + + def __reduce__(self): + cdef: + NativeFile sink = BufferOutputStream() + COutputStream* c_sink = sink.get_output_stream().get() + with nogil: + self._metadata.WriteTo(c_sink) + + cdef Buffer buffer = sink.getvalue() + return _reconstruct_filemetadata, (buffer,) + + def __hash__(self): + return hash((self.schema, + self.num_rows, + self.num_row_groups, + self.format_version, + self.serialized_size)) + + def __repr__(self): + return """{0} + created_by: {1} + num_columns: {2} + num_rows: {3} + num_row_groups: {4} + format_version: {5} + serialized_size: {6}""".format(object.__repr__(self), + self.created_by, self.num_columns, + self.num_rows, self.num_row_groups, + self.format_version, + self.serialized_size) + + def to_dict(self): + """ + Get dictionary representation of the file metadata. + + Returns + ------- + dict + Dictionary with a key for each attribute of this class. + """ + row_groups = [] + d = dict( + created_by=self.created_by, + num_columns=self.num_columns, + num_rows=self.num_rows, + num_row_groups=self.num_row_groups, + row_groups=row_groups, + format_version=self.format_version, + serialized_size=self.serialized_size + ) + for i in range(self.num_row_groups): + row_groups.append(self.row_group(i).to_dict()) + return d + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return NotImplemented + + def equals(self, FileMetaData other not None): + """ + Return whether the two file metadata objects are equal. + + Parameters + ---------- + other : FileMetaData + Metadata to compare against. + + Returns + ------- + are_equal : bool + """ + return self._metadata.Equals(deref(other._metadata)) + + @property + def schema(self): + """Schema of the file (:class:`ParquetSchema`).""" + if self._schema is None: + self._schema = ParquetSchema(self) + return self._schema + + @property + def serialized_size(self): + """Size of the original thrift encoded metadata footer (int).""" + return self._metadata.size() + + @property + def num_columns(self): + """Number of columns in file (int).""" + return self._metadata.num_columns() + + @property + def num_rows(self): + """Total number of rows in file (int).""" + return self._metadata.num_rows() + + @property + def num_row_groups(self): + """Number of row groups in file (int).""" + return self._metadata.num_row_groups() + + @property + def format_version(self): + """ + Parquet format version used in file (str, such as '1.0', '2.4'). + + If version is missing or unparsable, will default to assuming '2.6'. + """ + cdef ParquetVersion version = self._metadata.version() + if version == ParquetVersion_V1: + return '1.0' + elif version == ParquetVersion_V2_0: + return 'pseudo-2.0' + elif version == ParquetVersion_V2_4: + return '2.4' + elif version == ParquetVersion_V2_6: + return '2.6' + else: + warnings.warn('Unrecognized file version, assuming 2.6: {}' + .format(version)) + return '2.6' + + @property + def created_by(self): + """ + String describing source of the parquet file (str). + + This typically includes library name and version number. For example, Arrow 7.0's + writer returns 'parquet-cpp-arrow version 7.0.0'. + """ + return frombytes(self._metadata.created_by()) + + @property + def metadata(self): + """Additional metadata as key value pairs (dict[bytes, bytes]).""" + cdef: + unordered_map[c_string, c_string] metadata + const CKeyValueMetadata* underlying_metadata + underlying_metadata = self._metadata.key_value_metadata().get() + if underlying_metadata != NULL: + underlying_metadata.ToUnorderedMap(&metadata) + return metadata + else: + return None + + def row_group(self, int i): + """ + Get metadata for row group at index i. + + Parameters + ---------- + i : int + Row group index to get. + + Returns + ------- + row_group_metadata : RowGroupMetaData + """ + return RowGroupMetaData(self, i) + + def set_file_path(self, path): + """ + Set ColumnChunk file paths to the given value. + + This method modifies the ``file_path`` field of each ColumnChunk + in the FileMetaData to be a particular value. + + Parameters + ---------- + path : str + The file path to set on all ColumnChunks. + """ + cdef: + c_string c_path = tobytes(path) + self._metadata.set_file_path(c_path) + + def append_row_groups(self, FileMetaData other): + """ + Append row groups from other FileMetaData object. + + Parameters + ---------- + other : FileMetaData + Other metadata to append row groups from. + """ + cdef shared_ptr[CFileMetaData] c_metadata + + c_metadata = other.sp_metadata + self._metadata.AppendRowGroups(deref(c_metadata)) + + def write_metadata_file(self, where): + """ + Write the metadata to a metadata-only Parquet file. + + Parameters + ---------- + where : path or file-like object + Where to write the metadata. Should be a writable path on + the local filesystem, or a writable file-like object. + """ + cdef: + shared_ptr[COutputStream] sink + c_string c_where + + try: + where = _stringify_path(where) + except TypeError: + get_writer(where, &sink) + else: + c_where = tobytes(where) + with nogil: + sink = GetResultValue(FileOutputStream.Open(c_where)) + + with nogil: + check_status( + WriteMetaDataFile(deref(self._metadata), sink.get())) + + +cdef class ParquetSchema(_Weakrefable): + """A Parquet schema.""" + + def __cinit__(self, FileMetaData container): + self.parent = container + self.schema = container._metadata.schema() + + def __repr__(self): + return "{0}\n{1}".format( + object.__repr__(self), + frombytes(self.schema.ToString(), safe=True)) + + def __reduce__(self): + return ParquetSchema, (self.parent,) + + def __len__(self): + return self.schema.num_columns() + + def __getitem__(self, i): + return self.column(i) + + def __hash__(self): + return hash(self.schema.ToString()) + + @property + def names(self): + """Name of each field (list of str).""" + return [self[i].name for i in range(len(self))] + + def to_arrow_schema(self): + """ + Convert Parquet schema to effective Arrow schema. + + Returns + ------- + schema : Schema + """ + cdef shared_ptr[CSchema] sp_arrow_schema + + with nogil: + check_status(FromParquetSchema( + self.schema, default_arrow_reader_properties(), + self.parent._metadata.key_value_metadata(), + &sp_arrow_schema)) + + return pyarrow_wrap_schema(sp_arrow_schema) + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return NotImplemented + + def equals(self, ParquetSchema other): + """ + Return whether the two schemas are equal. + + Parameters + ---------- + other : ParquetSchema + Schema to compare against. + + Returns + ------- + are_equal : bool + """ + return self.schema.Equals(deref(other.schema)) + + def column(self, i): + """ + Return the schema for a single column. + + Parameters + ---------- + i : int + Index of column in schema. + + Returns + ------- + column_schema : ColumnSchema + """ + if i < 0 or i >= len(self): + raise IndexError('{0} out of bounds'.format(i)) + + return ColumnSchema(self, i) + + +cdef class ColumnSchema(_Weakrefable): + """Schema for a single column.""" + cdef: + int index + ParquetSchema parent + const ColumnDescriptor* descr + + def __cinit__(self, ParquetSchema schema, int index): + self.parent = schema + self.index = index # for pickling support + self.descr = schema.schema.Column(index) + + def __eq__(self, other): + try: + return self.equals(other) + except TypeError: + return NotImplemented + + def __reduce__(self): + return ColumnSchema, (self.parent, self.index) + + def equals(self, ColumnSchema other): + """ + Return whether the two column schemas are equal. + + Parameters + ---------- + other : ColumnSchema + Schema to compare against. + + Returns + ------- + are_equal : bool + """ + return self.descr.Equals(deref(other.descr)) + + def __repr__(self): + physical_type = self.physical_type + converted_type = self.converted_type + if converted_type == 'DECIMAL': + converted_type = 'DECIMAL({0}, {1})'.format(self.precision, + self.scale) + elif physical_type == 'FIXED_LEN_BYTE_ARRAY': + converted_type = ('FIXED_LEN_BYTE_ARRAY(length={0})' + .format(self.length)) + + return """ + name: {0} + path: {1} + max_definition_level: {2} + max_repetition_level: {3} + physical_type: {4} + logical_type: {5} + converted_type (legacy): {6}""".format(self.name, self.path, + self.max_definition_level, + self.max_repetition_level, + physical_type, + str(self.logical_type), + converted_type) + + @property + def name(self): + """Name of field (str).""" + return frombytes(self.descr.name()) + + @property + def path(self): + """Nested path to field, separated by periods (str).""" + return frombytes(self.descr.path().get().ToDotString()) + + @property + def max_definition_level(self): + """Maximum definition level (int).""" + return self.descr.max_definition_level() + + @property + def max_repetition_level(self): + """Maximum repetition level (int).""" + return self.descr.max_repetition_level() + + @property + def physical_type(self): + """Name of physical type (str).""" + return physical_type_name_from_enum(self.descr.physical_type()) + + @property + def logical_type(self): + """Logical type of column (:class:`ParquetLogicalType`).""" + return wrap_logical_type(self.descr.logical_type()) + + @property + def converted_type(self): + """Legacy converted type (str or None).""" + return converted_type_name_from_enum(self.descr.converted_type()) + + # FIXED_LEN_BYTE_ARRAY attribute + @property + def length(self): + """Array length if fixed length byte array type, None otherwise (int or None).""" + return self.descr.type_length() + + # Decimal attributes + @property + def precision(self): + """Precision if decimal type, None otherwise (int or None).""" + return self.descr.type_precision() + + @property + def scale(self): + """Scale if decimal type, None otherwise (int or None).""" + return self.descr.type_scale() + + +cdef physical_type_name_from_enum(ParquetType type_): + return { + ParquetType_BOOLEAN: 'BOOLEAN', + ParquetType_INT32: 'INT32', + ParquetType_INT64: 'INT64', + ParquetType_INT96: 'INT96', + ParquetType_FLOAT: 'FLOAT', + ParquetType_DOUBLE: 'DOUBLE', + ParquetType_BYTE_ARRAY: 'BYTE_ARRAY', + ParquetType_FIXED_LEN_BYTE_ARRAY: 'FIXED_LEN_BYTE_ARRAY', + }.get(type_, 'UNKNOWN') + + +cdef logical_type_name_from_enum(ParquetLogicalTypeId type_): + return { + ParquetLogicalType_UNDEFINED: 'UNDEFINED', + ParquetLogicalType_STRING: 'STRING', + ParquetLogicalType_MAP: 'MAP', + ParquetLogicalType_LIST: 'LIST', + ParquetLogicalType_ENUM: 'ENUM', + ParquetLogicalType_DECIMAL: 'DECIMAL', + ParquetLogicalType_DATE: 'DATE', + ParquetLogicalType_TIME: 'TIME', + ParquetLogicalType_TIMESTAMP: 'TIMESTAMP', + ParquetLogicalType_INT: 'INT', + ParquetLogicalType_FLOAT16: 'FLOAT16', + ParquetLogicalType_JSON: 'JSON', + ParquetLogicalType_BSON: 'BSON', + ParquetLogicalType_UUID: 'UUID', + ParquetLogicalType_NONE: 'NONE', + }.get(type_, 'UNKNOWN') + + +cdef converted_type_name_from_enum(ParquetConvertedType type_): + return { + ParquetConvertedType_NONE: 'NONE', + ParquetConvertedType_UTF8: 'UTF8', + ParquetConvertedType_MAP: 'MAP', + ParquetConvertedType_MAP_KEY_VALUE: 'MAP_KEY_VALUE', + ParquetConvertedType_LIST: 'LIST', + ParquetConvertedType_ENUM: 'ENUM', + ParquetConvertedType_DECIMAL: 'DECIMAL', + ParquetConvertedType_DATE: 'DATE', + ParquetConvertedType_TIME_MILLIS: 'TIME_MILLIS', + ParquetConvertedType_TIME_MICROS: 'TIME_MICROS', + ParquetConvertedType_TIMESTAMP_MILLIS: 'TIMESTAMP_MILLIS', + ParquetConvertedType_TIMESTAMP_MICROS: 'TIMESTAMP_MICROS', + ParquetConvertedType_UINT_8: 'UINT_8', + ParquetConvertedType_UINT_16: 'UINT_16', + ParquetConvertedType_UINT_32: 'UINT_32', + ParquetConvertedType_UINT_64: 'UINT_64', + ParquetConvertedType_INT_8: 'INT_8', + ParquetConvertedType_INT_16: 'INT_16', + ParquetConvertedType_INT_32: 'INT_32', + ParquetConvertedType_INT_64: 'INT_64', + ParquetConvertedType_JSON: 'JSON', + ParquetConvertedType_BSON: 'BSON', + ParquetConvertedType_INTERVAL: 'INTERVAL', + }.get(type_, 'UNKNOWN') + + +cdef encoding_name_from_enum(ParquetEncoding encoding_): + return { + ParquetEncoding_PLAIN: 'PLAIN', + ParquetEncoding_PLAIN_DICTIONARY: 'PLAIN_DICTIONARY', + ParquetEncoding_RLE: 'RLE', + ParquetEncoding_BIT_PACKED: 'BIT_PACKED', + ParquetEncoding_DELTA_BINARY_PACKED: 'DELTA_BINARY_PACKED', + ParquetEncoding_DELTA_LENGTH_BYTE_ARRAY: 'DELTA_LENGTH_BYTE_ARRAY', + ParquetEncoding_DELTA_BYTE_ARRAY: 'DELTA_BYTE_ARRAY', + ParquetEncoding_RLE_DICTIONARY: 'RLE_DICTIONARY', + ParquetEncoding_BYTE_STREAM_SPLIT: 'BYTE_STREAM_SPLIT', + }.get(encoding_, 'UNKNOWN') + + +cdef encoding_enum_from_name(str encoding_name): + enc = { + 'PLAIN': ParquetEncoding_PLAIN, + 'BIT_PACKED': ParquetEncoding_BIT_PACKED, + 'RLE': ParquetEncoding_RLE, + 'BYTE_STREAM_SPLIT': ParquetEncoding_BYTE_STREAM_SPLIT, + 'DELTA_BINARY_PACKED': ParquetEncoding_DELTA_BINARY_PACKED, + 'DELTA_LENGTH_BYTE_ARRAY': ParquetEncoding_DELTA_LENGTH_BYTE_ARRAY, + 'DELTA_BYTE_ARRAY': ParquetEncoding_DELTA_BYTE_ARRAY, + 'RLE_DICTIONARY': 'dict', + 'PLAIN_DICTIONARY': 'dict', + }.get(encoding_name, None) + if enc is None: + raise ValueError(f"Unsupported column encoding: {encoding_name!r}") + elif enc == 'dict': + raise ValueError(f"{encoding_name!r} is already used by default.") + else: + return enc + + +cdef compression_name_from_enum(ParquetCompression compression_): + return { + ParquetCompression_UNCOMPRESSED: 'UNCOMPRESSED', + ParquetCompression_SNAPPY: 'SNAPPY', + ParquetCompression_GZIP: 'GZIP', + ParquetCompression_LZO: 'LZO', + ParquetCompression_BROTLI: 'BROTLI', + ParquetCompression_LZ4: 'LZ4', + ParquetCompression_ZSTD: 'ZSTD', + }.get(compression_, 'UNKNOWN') + + +cdef int check_compression_name(name) except -1: + if name.upper() not in {'NONE', 'SNAPPY', 'GZIP', 'LZO', 'BROTLI', 'LZ4', + 'ZSTD'}: + raise ArrowException("Unsupported compression: " + name) + return 0 + + +cdef ParquetCompression compression_from_name(name): + name = name.upper() + if name == 'SNAPPY': + return ParquetCompression_SNAPPY + elif name == 'GZIP': + return ParquetCompression_GZIP + elif name == 'LZO': + return ParquetCompression_LZO + elif name == 'BROTLI': + return ParquetCompression_BROTLI + elif name == 'LZ4': + return ParquetCompression_LZ4 + elif name == 'ZSTD': + return ParquetCompression_ZSTD + else: + return ParquetCompression_UNCOMPRESSED + + +cdef class ParquetReader(_Weakrefable): + cdef: + object source + CMemoryPool* pool + UniquePtrNoGIL[FileReader] reader + FileMetaData _metadata + shared_ptr[CRandomAccessFile] rd_handle + + cdef public: + _column_idx_map + + def __cinit__(self, MemoryPool memory_pool=None): + self.pool = maybe_unbox_memory_pool(memory_pool) + self._metadata = None + + def open(self, object source not None, *, bint use_memory_map=False, + read_dictionary=None, FileMetaData metadata=None, + int buffer_size=0, bint pre_buffer=False, + coerce_int96_timestamp_unit=None, + FileDecryptionProperties decryption_properties=None, + thrift_string_size_limit=None, + thrift_container_size_limit=None, + page_checksum_verification=False): + """ + Open a parquet file for reading. + + Parameters + ---------- + source : str, pathlib.Path, pyarrow.NativeFile, or file-like object + use_memory_map : bool, default False + read_dictionary : iterable[int or str], optional + metadata : FileMetaData, optional + buffer_size : int, default 0 + pre_buffer : bool, default False + coerce_int96_timestamp_unit : str, optional + decryption_properties : FileDecryptionProperties, optional + thrift_string_size_limit : int, optional + thrift_container_size_limit : int, optional + page_checksum_verification : bool, default False + """ + cdef: + shared_ptr[CFileMetaData] c_metadata + CReaderProperties properties = default_reader_properties() + ArrowReaderProperties arrow_props = ( + default_arrow_reader_properties()) + FileReaderBuilder builder + + if pre_buffer and not is_threading_enabled(): + pre_buffer = False + + if metadata is not None: + c_metadata = metadata.sp_metadata + + if buffer_size > 0: + properties.enable_buffered_stream() + properties.set_buffer_size(buffer_size) + elif buffer_size == 0: + properties.disable_buffered_stream() + else: + raise ValueError('Buffer size must be larger than zero') + + if thrift_string_size_limit is not None: + if thrift_string_size_limit <= 0: + raise ValueError("thrift_string_size_limit " + "must be larger than zero") + properties.set_thrift_string_size_limit(thrift_string_size_limit) + if thrift_container_size_limit is not None: + if thrift_container_size_limit <= 0: + raise ValueError("thrift_container_size_limit " + "must be larger than zero") + properties.set_thrift_container_size_limit( + thrift_container_size_limit) + + if decryption_properties is not None: + properties.file_decryption_properties( + decryption_properties.unwrap()) + + arrow_props.set_pre_buffer(pre_buffer) + + properties.set_page_checksum_verification(page_checksum_verification) + + if coerce_int96_timestamp_unit is None: + # use the default defined in default_arrow_reader_properties() + pass + else: + arrow_props.set_coerce_int96_timestamp_unit( + string_to_timeunit(coerce_int96_timestamp_unit)) + + self.source = source + get_reader(source, use_memory_map, &self.rd_handle) + + with nogil: + check_status(builder.Open(self.rd_handle, properties, c_metadata)) + + # Set up metadata + with nogil: + c_metadata = builder.raw_reader().metadata() + self._metadata = result = FileMetaData() + result.init(c_metadata) + + if read_dictionary is not None: + self._set_read_dictionary(read_dictionary, &arrow_props) + + with nogil: + check_status(builder.memory_pool(self.pool) + .properties(arrow_props) + .Build(&self.reader)) + + cdef _set_read_dictionary(self, read_dictionary, + ArrowReaderProperties* props): + for column in read_dictionary: + if not isinstance(column, int): + column = self.column_name_idx(column) + props.set_read_dictionary(column, True) + + @property + def column_paths(self): + cdef: + FileMetaData container = self.metadata + const CFileMetaData* metadata = container._metadata + vector[c_string] path + int i = 0 + + paths = [] + for i in range(0, metadata.num_columns()): + path = (metadata.schema().Column(i) + .path().get().ToDotVector()) + paths.append([frombytes(x) for x in path]) + + return paths + + @property + def metadata(self): + return self._metadata + + @property + def schema_arrow(self): + cdef shared_ptr[CSchema] out + with nogil: + check_status(self.reader.get().GetSchema(&out)) + return pyarrow_wrap_schema(out) + + @property + def num_row_groups(self): + return self.reader.get().num_row_groups() + + def set_use_threads(self, bint use_threads): + """ + Parameters + ---------- + use_threads : bool + """ + if is_threading_enabled(): + self.reader.get().set_use_threads(use_threads) + else: + self.reader.get().set_use_threads(False) + + def set_batch_size(self, int64_t batch_size): + """ + Parameters + ---------- + batch_size : int64 + """ + self.reader.get().set_batch_size(batch_size) + + def iter_batches(self, int64_t batch_size, row_groups, column_indices=None, + bint use_threads=True): + """ + Parameters + ---------- + batch_size : int64 + row_groups : list[int] + column_indices : list[int], optional + use_threads : bool, default True + + Yields + ------ + next : RecordBatch + """ + cdef: + vector[int] c_row_groups + vector[int] c_column_indices + shared_ptr[CRecordBatch] record_batch + UniquePtrNoGIL[CRecordBatchReader] recordbatchreader + + self.set_batch_size(batch_size) + + if use_threads: + self.set_use_threads(use_threads) + + for row_group in row_groups: + c_row_groups.push_back(row_group) + + if column_indices is not None: + for index in column_indices: + c_column_indices.push_back(index) + with nogil: + check_status( + self.reader.get().GetRecordBatchReader( + c_row_groups, c_column_indices, &recordbatchreader + ) + ) + else: + with nogil: + check_status( + self.reader.get().GetRecordBatchReader( + c_row_groups, &recordbatchreader + ) + ) + + while True: + with nogil: + check_status( + recordbatchreader.get().ReadNext(&record_batch) + ) + if record_batch.get() == NULL: + break + + yield pyarrow_wrap_batch(record_batch) + + def read_row_group(self, int i, column_indices=None, + bint use_threads=True): + """ + Parameters + ---------- + i : int + column_indices : list[int], optional + use_threads : bool, default True + + Returns + ------- + table : pyarrow.Table + """ + return self.read_row_groups([i], column_indices, use_threads) + + def read_row_groups(self, row_groups not None, column_indices=None, + bint use_threads=True): + """ + Parameters + ---------- + row_groups : list[int] + column_indices : list[int], optional + use_threads : bool, default True + + Returns + ------- + table : pyarrow.Table + """ + cdef: + shared_ptr[CTable] ctable + vector[int] c_row_groups + vector[int] c_column_indices + + self.set_use_threads(use_threads) + + for row_group in row_groups: + c_row_groups.push_back(row_group) + + if column_indices is not None: + for index in column_indices: + c_column_indices.push_back(index) + + with nogil: + check_status(self.reader.get() + .ReadRowGroups(c_row_groups, c_column_indices, + &ctable)) + else: + # Read all columns + with nogil: + check_status(self.reader.get() + .ReadRowGroups(c_row_groups, &ctable)) + return pyarrow_wrap_table(ctable) + + def read_all(self, column_indices=None, bint use_threads=True): + """ + Parameters + ---------- + column_indices : list[int], optional + use_threads : bool, default True + + Returns + ------- + table : pyarrow.Table + """ + cdef: + shared_ptr[CTable] ctable + vector[int] c_column_indices + + self.set_use_threads(use_threads) + + if column_indices is not None: + for index in column_indices: + c_column_indices.push_back(index) + + with nogil: + check_status(self.reader.get() + .ReadTable(c_column_indices, &ctable)) + else: + # Read all columns + with nogil: + check_status(self.reader.get() + .ReadTable(&ctable)) + return pyarrow_wrap_table(ctable) + + def scan_contents(self, column_indices=None, batch_size=65536): + """ + Parameters + ---------- + column_indices : list[int], optional + batch_size : int32, default 65536 + + Returns + ------- + num_rows : int64 + """ + cdef: + vector[int] c_column_indices + int32_t c_batch_size + int64_t c_num_rows + + if column_indices is not None: + for index in column_indices: + c_column_indices.push_back(index) + + c_batch_size = batch_size + + with nogil: + check_status(self.reader.get() + .ScanContents(c_column_indices, c_batch_size, + &c_num_rows)) + + return c_num_rows + + def column_name_idx(self, column_name): + """ + Find the index of a column by its name. + + Parameters + ---------- + column_name : str + Name of the column; separation of nesting levels is done via ".". + + Returns + ------- + column_idx : int + Integer index of the column in the schema. + """ + cdef: + FileMetaData container = self.metadata + const CFileMetaData* metadata = container._metadata + int i = 0 + + if self._column_idx_map is None: + self._column_idx_map = {} + for i in range(0, metadata.num_columns()): + col_bytes = tobytes(metadata.schema().Column(i) + .path().get().ToDotString()) + self._column_idx_map[col_bytes] = i + + return self._column_idx_map[tobytes(column_name)] + + def read_column(self, int column_index): + """ + Read the column at the specified index. + + Parameters + ---------- + column_index : int + Index of the column. + + Returns + ------- + column : pyarrow.ChunkedArray + """ + cdef shared_ptr[CChunkedArray] out + with nogil: + check_status(self.reader.get() + .ReadColumn(column_index, &out)) + return pyarrow_wrap_chunked_array(out) + + def close(self): + if not self.closed: + with nogil: + check_status(self.rd_handle.get().Close()) + + @property + def closed(self): + if self.rd_handle == NULL: + return True + with nogil: + closed = self.rd_handle.get().closed() + return closed + + +cdef CSortingColumn _convert_sorting_column(SortingColumn sorting_column): + cdef CSortingColumn c_sorting_column + + c_sorting_column.column_idx = sorting_column.column_index + c_sorting_column.descending = sorting_column.descending + c_sorting_column.nulls_first = sorting_column.nulls_first + + return c_sorting_column + + +cdef vector[CSortingColumn] _convert_sorting_columns(sorting_columns) except *: + if not (isinstance(sorting_columns, Sequence) + and all(isinstance(col, SortingColumn) for col in sorting_columns)): + raise ValueError( + "'sorting_columns' must be a list of `SortingColumn`") + + cdef vector[CSortingColumn] c_sorting_columns = [_convert_sorting_column(col) + for col in sorting_columns] + + return c_sorting_columns + + +cdef shared_ptr[WriterProperties] _create_writer_properties( + use_dictionary=None, + compression=None, + version=None, + write_statistics=None, + data_page_size=None, + compression_level=None, + use_byte_stream_split=False, + column_encoding=None, + data_page_version=None, + FileEncryptionProperties encryption_properties=None, + write_batch_size=None, + dictionary_pagesize_limit=None, + write_page_index=False, + write_page_checksum=False, + sorting_columns=None, + store_decimal_as_integer=False) except *: + + """General writer properties""" + cdef: + shared_ptr[WriterProperties] properties + WriterProperties.Builder props + + # data_page_version + + if data_page_version is not None: + if data_page_version == "1.0": + props.data_page_version(ParquetDataPageVersion_V1) + elif data_page_version == "2.0": + props.data_page_version(ParquetDataPageVersion_V2) + else: + raise ValueError("Unsupported Parquet data page version: {0}" + .format(data_page_version)) + + # version + + if version is not None: + if version == "1.0": + props.version(ParquetVersion_V1) + elif version in ("2.0", "pseudo-2.0"): + warnings.warn( + "Parquet format '2.0' pseudo version is deprecated, use " + "'2.4' or '2.6' for fine-grained feature selection", + FutureWarning, stacklevel=2) + props.version(ParquetVersion_V2_0) + elif version == "2.4": + props.version(ParquetVersion_V2_4) + elif version == "2.6": + props.version(ParquetVersion_V2_6) + else: + raise ValueError("Unsupported Parquet format version: {0}" + .format(version)) + + # compression + + if isinstance(compression, basestring): + check_compression_name(compression) + props.compression(compression_from_name(compression)) + elif compression is not None: + for column, codec in compression.iteritems(): + check_compression_name(codec) + props.compression(tobytes(column), compression_from_name(codec)) + + if isinstance(compression_level, int): + props.compression_level(compression_level) + elif compression_level is not None: + for column, level in compression_level.iteritems(): + props.compression_level(tobytes(column), level) + + # use_dictionary + + if isinstance(use_dictionary, bool): + if use_dictionary: + props.enable_dictionary() + if column_encoding is not None: + raise ValueError( + "To use 'column_encoding' set 'use_dictionary' to False") + else: + props.disable_dictionary() + elif use_dictionary is not None: + # Deactivate dictionary encoding by default + props.disable_dictionary() + for column in use_dictionary: + props.enable_dictionary(tobytes(column)) + if (column_encoding is not None and + column_encoding.get(column) is not None): + raise ValueError( + "To use 'column_encoding' set 'use_dictionary' to False") + + # write_statistics + + if isinstance(write_statistics, bool): + if write_statistics: + props.enable_statistics() + else: + props.disable_statistics() + elif write_statistics is not None: + # Deactivate statistics by default and enable for specified columns + props.disable_statistics() + for column in write_statistics: + props.enable_statistics(tobytes(column)) + + # sorting_columns + + if sorting_columns is not None: + props.set_sorting_columns(_convert_sorting_columns(sorting_columns)) + + # use_byte_stream_split + + if isinstance(use_byte_stream_split, bool): + if use_byte_stream_split: + if column_encoding is not None: + raise ValueError( + "'use_byte_stream_split' cannot be passed" + "together with 'column_encoding'") + else: + props.encoding(ParquetEncoding_BYTE_STREAM_SPLIT) + elif use_byte_stream_split is not None: + for column in use_byte_stream_split: + if column_encoding is None: + column_encoding = {column: 'BYTE_STREAM_SPLIT'} + elif column_encoding.get(column, None) is None: + column_encoding[column] = 'BYTE_STREAM_SPLIT' + else: + raise ValueError( + "'use_byte_stream_split' cannot be passed" + "together with 'column_encoding'") + + # store_decimal_as_integer + + if isinstance(store_decimal_as_integer, bool): + if store_decimal_as_integer: + props.enable_store_decimal_as_integer() + else: + props.disable_store_decimal_as_integer() + else: + raise TypeError("'store_decimal_as_integer' must be a boolean") + + # column_encoding + # encoding map - encode individual columns + + if column_encoding is not None: + if isinstance(column_encoding, dict): + for column, _encoding in column_encoding.items(): + props.encoding(tobytes(column), + encoding_enum_from_name(_encoding)) + elif isinstance(column_encoding, str): + props.encoding(encoding_enum_from_name(column_encoding)) + else: + raise TypeError( + "'column_encoding' should be a dictionary or a string") + + if data_page_size is not None: + props.data_pagesize(data_page_size) + + if write_batch_size is not None: + props.write_batch_size(write_batch_size) + + if dictionary_pagesize_limit is not None: + props.dictionary_pagesize_limit(dictionary_pagesize_limit) + + # encryption + + if encryption_properties is not None: + props.encryption( + (encryption_properties).unwrap()) + + # For backwards compatibility reasons we cap the maximum row group size + # at 64Mi rows. This could be changed in the future, though it would be + # a breaking change. + # + # The user can always specify a smaller row group size (and the default + # is smaller) when calling write_table. If the call to write_table uses + # a size larger than this then it will be latched to this value. + props.max_row_group_length(_MAX_ROW_GROUP_SIZE) + + # checksum + + if write_page_checksum: + props.enable_page_checksum() + else: + props.disable_page_checksum() + + # page index + + if write_page_index: + props.enable_write_page_index() + else: + props.disable_write_page_index() + + properties = props.build() + + return properties + + +cdef shared_ptr[ArrowWriterProperties] _create_arrow_writer_properties( + use_deprecated_int96_timestamps=False, + coerce_timestamps=None, + allow_truncated_timestamps=False, + writer_engine_version=None, + use_compliant_nested_type=True, + store_schema=True) except *: + """Arrow writer properties""" + cdef: + shared_ptr[ArrowWriterProperties] arrow_properties + ArrowWriterProperties.Builder arrow_props + + # Store the original Arrow schema so things like dictionary types can + # be automatically reconstructed + if store_schema: + arrow_props.store_schema() + + # int96 support + + if use_deprecated_int96_timestamps: + arrow_props.enable_deprecated_int96_timestamps() + else: + arrow_props.disable_deprecated_int96_timestamps() + + # coerce_timestamps + + if coerce_timestamps == 'ms': + arrow_props.coerce_timestamps(TimeUnit_MILLI) + elif coerce_timestamps == 'us': + arrow_props.coerce_timestamps(TimeUnit_MICRO) + elif coerce_timestamps is not None: + raise ValueError('Invalid value for coerce_timestamps: {0}' + .format(coerce_timestamps)) + + # allow_truncated_timestamps + + if allow_truncated_timestamps: + arrow_props.allow_truncated_timestamps() + else: + arrow_props.disallow_truncated_timestamps() + + # use_compliant_nested_type + + if use_compliant_nested_type: + arrow_props.enable_compliant_nested_types() + else: + arrow_props.disable_compliant_nested_types() + + # writer_engine_version + + if writer_engine_version == "V1": + warnings.warn("V1 parquet writer engine is a no-op. Use V2.") + arrow_props.set_engine_version(ArrowWriterEngineVersion.V1) + elif writer_engine_version != "V2": + raise ValueError("Unsupported Writer Engine Version: {0}" + .format(writer_engine_version)) + + arrow_properties = arrow_props.build() + + return arrow_properties + +cdef _name_to_index_map(Schema arrow_schema): + cdef: + shared_ptr[CSchema] sp_arrow_schema + shared_ptr[SchemaDescriptor] sp_parquet_schema + shared_ptr[WriterProperties] props = _create_writer_properties() + shared_ptr[ArrowWriterProperties] arrow_props = _create_arrow_writer_properties( + use_deprecated_int96_timestamps=False, + coerce_timestamps=None, + allow_truncated_timestamps=False, + writer_engine_version="V2" + ) + + sp_arrow_schema = pyarrow_unwrap_schema(arrow_schema) + + with nogil: + check_status(ToParquetSchema( + sp_arrow_schema.get(), deref(props.get()), deref(arrow_props.get()), &sp_parquet_schema)) + + out = dict() + + cdef SchemaDescriptor* parquet_schema = sp_parquet_schema.get() + + for i in range(parquet_schema.num_columns()): + name = frombytes(parquet_schema.Column(i).path().get().ToDotString()) + out[name] = i + + return out + + +cdef class ParquetWriter(_Weakrefable): + cdef: + unique_ptr[FileWriter] writer + shared_ptr[COutputStream] sink + bint own_sink + + cdef readonly: + object use_dictionary + object use_deprecated_int96_timestamps + object use_byte_stream_split + object column_encoding + object coerce_timestamps + object allow_truncated_timestamps + object compression + object compression_level + object data_page_version + object use_compliant_nested_type + object version + object write_statistics + object writer_engine_version + int row_group_size + int64_t data_page_size + FileEncryptionProperties encryption_properties + int64_t write_batch_size + int64_t dictionary_pagesize_limit + object store_schema + object store_decimal_as_integer + + def __cinit__(self, where, Schema schema not None, use_dictionary=None, + compression=None, version=None, + write_statistics=None, + MemoryPool memory_pool=None, + use_deprecated_int96_timestamps=False, + coerce_timestamps=None, + data_page_size=None, + allow_truncated_timestamps=False, + compression_level=None, + use_byte_stream_split=False, + column_encoding=None, + writer_engine_version=None, + data_page_version=None, + use_compliant_nested_type=True, + encryption_properties=None, + write_batch_size=None, + dictionary_pagesize_limit=None, + store_schema=True, + write_page_index=False, + write_page_checksum=False, + sorting_columns=None, + store_decimal_as_integer=False): + cdef: + shared_ptr[WriterProperties] properties + shared_ptr[ArrowWriterProperties] arrow_properties + c_string c_where + CMemoryPool* pool + + try: + where = _stringify_path(where) + except TypeError: + get_writer(where, &self.sink) + self.own_sink = False + else: + c_where = tobytes(where) + with nogil: + self.sink = GetResultValue(FileOutputStream.Open(c_where)) + self.own_sink = True + + properties = _create_writer_properties( + use_dictionary=use_dictionary, + compression=compression, + version=version, + write_statistics=write_statistics, + data_page_size=data_page_size, + compression_level=compression_level, + use_byte_stream_split=use_byte_stream_split, + column_encoding=column_encoding, + data_page_version=data_page_version, + encryption_properties=encryption_properties, + write_batch_size=write_batch_size, + dictionary_pagesize_limit=dictionary_pagesize_limit, + write_page_index=write_page_index, + write_page_checksum=write_page_checksum, + sorting_columns=sorting_columns, + store_decimal_as_integer=store_decimal_as_integer, + ) + arrow_properties = _create_arrow_writer_properties( + use_deprecated_int96_timestamps=use_deprecated_int96_timestamps, + coerce_timestamps=coerce_timestamps, + allow_truncated_timestamps=allow_truncated_timestamps, + writer_engine_version=writer_engine_version, + use_compliant_nested_type=use_compliant_nested_type, + store_schema=store_schema, + ) + + pool = maybe_unbox_memory_pool(memory_pool) + with nogil: + self.writer = move(GetResultValue( + FileWriter.Open(deref(schema.schema), pool, + self.sink, properties, arrow_properties))) + + def close(self): + with nogil: + check_status(self.writer.get().Close()) + if self.own_sink: + check_status(self.sink.get().Close()) + + def write_table(self, Table table, row_group_size=None): + cdef: + CTable* ctable = table.table + int64_t c_row_group_size + + if row_group_size is None or row_group_size == -1: + c_row_group_size = min(ctable.num_rows(), _DEFAULT_ROW_GROUP_SIZE) + elif row_group_size == 0: + raise ValueError('Row group size cannot be 0') + else: + c_row_group_size = row_group_size + + with nogil: + check_status(self.writer.get() + .WriteTable(deref(ctable), c_row_group_size)) + + def add_key_value_metadata(self, key_value_metadata): + cdef: + shared_ptr[const CKeyValueMetadata] c_metadata + + c_metadata = pyarrow_unwrap_metadata(KeyValueMetadata(key_value_metadata)) + with nogil: + check_status(self.writer.get() + .AddKeyValueMetadata(c_metadata)) + + @property + def metadata(self): + cdef: + shared_ptr[CFileMetaData] metadata + FileMetaData result + with nogil: + metadata = self.writer.get().metadata() + if metadata: + result = FileMetaData() + result.init(metadata) + return result + raise RuntimeError( + 'file metadata is only available after writer close') diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_parquet_encryption.cpython-310-x86_64-linux-gnu.so b/vllm/lib/python3.10/site-packages/pyarrow/_parquet_encryption.cpython-310-x86_64-linux-gnu.so new file mode 100644 index 0000000000000000000000000000000000000000..ec52d4989da86bda942d5378b7a6d53e6f2159a1 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_parquet_encryption.cpython-310-x86_64-linux-gnu.so @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c5439a88a794d415e8530186f617d8ac3dd79deba6a29ad981d91710cef58e6 +size 285904 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/_pyarrow_cpp_tests.pxd b/vllm/lib/python3.10/site-packages/pyarrow/_pyarrow_cpp_tests.pxd new file mode 100644 index 0000000000000000000000000000000000000000..91c0220d7310870a7803ecceb2c32b8b32f8c11d --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/_pyarrow_cpp_tests.pxd @@ -0,0 +1,33 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# distutils: language = c++ +# cython: language_level = 3 + +from pyarrow.includes.common cimport * +from pyarrow.includes.libarrow cimport CStatus + + +ctypedef CStatus cb_test_func() + +cdef extern from "arrow/python/python_test.h" namespace "arrow::py::testing" nogil: + + cdef cppclass CTestCase "arrow::py::testing::TestCase": + c_string name + cb_test_func func + + vector[CTestCase] GetCppTestCases() diff --git a/vllm/lib/python3.10/site-packages/pyarrow/compat.pxi b/vllm/lib/python3.10/site-packages/pyarrow/compat.pxi new file mode 100644 index 0000000000000000000000000000000000000000..8cf106d5609b50dd84c082dcfd36aee5b16fbee4 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/compat.pxi @@ -0,0 +1,71 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + + +def encode_file_path(path): + if isinstance(path, str): + # POSIX systems can handle utf-8. UTF8 is converted to utf16-le in + # libarrow + encoded_path = path.encode('utf-8') + else: + encoded_path = path + + # Windows file system requires utf-16le for file names; Arrow C++ libraries + # will convert utf8 to utf16 + return encoded_path + + +# Starting with Python 3.7, dicts are guaranteed to be insertion-ordered. +ordered_dict = dict + + +try: + import cloudpickle as pickle +except ImportError: + import pickle + + +def tobytes(o): + """ + Encode a unicode or bytes string to bytes. + + Parameters + ---------- + o : str or bytes + Input string. + """ + if isinstance(o, str): + return o.encode('utf8') + else: + return o + + +def frombytes(o, *, safe=False): + """ + Decode the given bytestring to unicode. + + Parameters + ---------- + o : bytes-like + Input object. + safe : bool, default False + If true, raise on encoding errors. + """ + if safe: + return o.decode('utf8', errors='replace') + else: + return o.decode('utf8') diff --git a/vllm/lib/python3.10/site-packages/pyarrow/flight.py b/vllm/lib/python3.10/site-packages/pyarrow/flight.py new file mode 100644 index 0000000000000000000000000000000000000000..b1836907c6744161c86f32e873316923c60b4226 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/flight.py @@ -0,0 +1,69 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +try: + from pyarrow._flight import ( # noqa:F401 + connect, + Action, + ActionType, + BasicAuth, + CallInfo, + CertKeyPair, + ClientAuthHandler, + ClientMiddleware, + ClientMiddlewareFactory, + DescriptorType, + FlightCallOptions, + FlightCancelledError, + FlightClient, + FlightDataStream, + FlightDescriptor, + FlightEndpoint, + FlightError, + FlightInfo, + FlightInternalError, + FlightMetadataReader, + FlightMetadataWriter, + FlightMethod, + FlightServerBase, + FlightServerError, + FlightStreamChunk, + FlightStreamReader, + FlightStreamWriter, + FlightTimedOutError, + FlightUnauthenticatedError, + FlightUnauthorizedError, + FlightUnavailableError, + FlightWriteSizeExceededError, + GeneratorStream, + Location, + MetadataRecordBatchReader, + MetadataRecordBatchWriter, + RecordBatchStream, + Result, + SchemaResult, + ServerAuthHandler, + ServerCallContext, + ServerMiddleware, + ServerMiddlewareFactory, + Ticket, + TracingServerMiddlewareFactory, + ) +except ImportError as exc: + raise ImportError( + f"The pyarrow installation is not built with support for 'flight' ({str(exc)})" + ) from None diff --git a/vllm/lib/python3.10/site-packages/pyarrow/ipc.pxi b/vllm/lib/python3.10/site-packages/pyarrow/ipc.pxi new file mode 100644 index 0000000000000000000000000000000000000000..e15b0ea40ed2e7de9d5a7f1776d26ff40909b4c4 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/ipc.pxi @@ -0,0 +1,1403 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from cpython.pycapsule cimport PyCapsule_CheckExact, PyCapsule_GetPointer, PyCapsule_New + +from collections import namedtuple +import warnings +from cython import sizeof + +cpdef enum MetadataVersion: + V1 = CMetadataVersion_V1 + V2 = CMetadataVersion_V2 + V3 = CMetadataVersion_V3 + V4 = CMetadataVersion_V4 + V5 = CMetadataVersion_V5 + + +cdef object _wrap_metadata_version(CMetadataVersion version): + return MetadataVersion( version) + + +cdef CMetadataVersion _unwrap_metadata_version( + MetadataVersion version) except *: + if version == MetadataVersion.V1: + return CMetadataVersion_V1 + elif version == MetadataVersion.V2: + return CMetadataVersion_V2 + elif version == MetadataVersion.V3: + return CMetadataVersion_V3 + elif version == MetadataVersion.V4: + return CMetadataVersion_V4 + elif version == MetadataVersion.V5: + return CMetadataVersion_V5 + raise ValueError("Not a metadata version: " + repr(version)) + + +_WriteStats = namedtuple( + 'WriteStats', + ('num_messages', 'num_record_batches', 'num_dictionary_batches', + 'num_dictionary_deltas', 'num_replaced_dictionaries')) + + +class WriteStats(_WriteStats): + """IPC write statistics + + Parameters + ---------- + num_messages : int + Number of messages. + num_record_batches : int + Number of record batches. + num_dictionary_batches : int + Number of dictionary batches. + num_dictionary_deltas : int + Delta of dictionaries. + num_replaced_dictionaries : int + Number of replaced dictionaries. + """ + __slots__ = () + + +@staticmethod +cdef _wrap_write_stats(CIpcWriteStats c): + return WriteStats(c.num_messages, c.num_record_batches, + c.num_dictionary_batches, c.num_dictionary_deltas, + c.num_replaced_dictionaries) + + +_ReadStats = namedtuple( + 'ReadStats', + ('num_messages', 'num_record_batches', 'num_dictionary_batches', + 'num_dictionary_deltas', 'num_replaced_dictionaries')) + + +class ReadStats(_ReadStats): + """IPC read statistics + + Parameters + ---------- + num_messages : int + Number of messages. + num_record_batches : int + Number of record batches. + num_dictionary_batches : int + Number of dictionary batches. + num_dictionary_deltas : int + Delta of dictionaries. + num_replaced_dictionaries : int + Number of replaced dictionaries. + """ + __slots__ = () + + +@staticmethod +cdef _wrap_read_stats(CIpcReadStats c): + return ReadStats(c.num_messages, c.num_record_batches, + c.num_dictionary_batches, c.num_dictionary_deltas, + c.num_replaced_dictionaries) + + +cdef class IpcReadOptions(_Weakrefable): + """ + Serialization options for reading IPC format. + + Parameters + ---------- + ensure_native_endian : bool, default True + Whether to convert incoming data to platform-native endianness. + use_threads : bool + Whether to use the global CPU thread pool to parallelize any + computational tasks like decompression + included_fields : list + If empty (the default), return all deserialized fields. + If non-empty, the values are the indices of fields to read on + the top-level schema + """ + __slots__ = () + + # cdef block is in lib.pxd + + def __init__(self, *, bint ensure_native_endian=True, + bint use_threads=True, list included_fields=None): + self.c_options = CIpcReadOptions.Defaults() + self.ensure_native_endian = ensure_native_endian + self.use_threads = use_threads + if included_fields is not None: + self.included_fields = included_fields + + @property + def ensure_native_endian(self): + return self.c_options.ensure_native_endian + + @ensure_native_endian.setter + def ensure_native_endian(self, bint value): + self.c_options.ensure_native_endian = value + + @property + def use_threads(self): + return self.c_options.use_threads + + @use_threads.setter + def use_threads(self, bint value): + self.c_options.use_threads = value + + @property + def included_fields(self): + return self.c_options.included_fields + + @included_fields.setter + def included_fields(self, list value not None): + self.c_options.included_fields = value + + +cdef class IpcWriteOptions(_Weakrefable): + """ + Serialization options for the IPC format. + + Parameters + ---------- + metadata_version : MetadataVersion, default MetadataVersion.V5 + The metadata version to write. V5 is the current and latest, + V4 is the pre-1.0 metadata version (with incompatible Union layout). + allow_64bit : bool, default False + If true, allow field lengths that don't fit in a signed 32-bit int. + use_legacy_format : bool, default False + Whether to use the pre-Arrow 0.15 IPC format. + compression : str, Codec, or None + compression codec to use for record batch buffers. + If None then batch buffers will be uncompressed. + Must be "lz4", "zstd" or None. + To specify a compression_level use `pyarrow.Codec` + use_threads : bool + Whether to use the global CPU thread pool to parallelize any + computational tasks like compression. + emit_dictionary_deltas : bool + Whether to emit dictionary deltas. Default is false for maximum + stream compatibility. + unify_dictionaries : bool + If true then calls to write_table will attempt to unify dictionaries + across all batches in the table. This can help avoid the need for + replacement dictionaries (which the file format does not support) + but requires computing the unified dictionary and then remapping + the indices arrays. + + This parameter is ignored when writing to the IPC stream format as + the IPC stream format can support replacement dictionaries. + """ + __slots__ = () + + # cdef block is in lib.pxd + + def __init__(self, *, metadata_version=MetadataVersion.V5, + bint allow_64bit=False, use_legacy_format=False, + compression=None, bint use_threads=True, + bint emit_dictionary_deltas=False, + bint unify_dictionaries=False): + self.c_options = CIpcWriteOptions.Defaults() + self.allow_64bit = allow_64bit + self.use_legacy_format = use_legacy_format + self.metadata_version = metadata_version + if compression is not None: + self.compression = compression + self.use_threads = use_threads + self.emit_dictionary_deltas = emit_dictionary_deltas + self.unify_dictionaries = unify_dictionaries + + @property + def allow_64bit(self): + return self.c_options.allow_64bit + + @allow_64bit.setter + def allow_64bit(self, bint value): + self.c_options.allow_64bit = value + + @property + def use_legacy_format(self): + return self.c_options.write_legacy_ipc_format + + @use_legacy_format.setter + def use_legacy_format(self, bint value): + self.c_options.write_legacy_ipc_format = value + + @property + def metadata_version(self): + return _wrap_metadata_version(self.c_options.metadata_version) + + @metadata_version.setter + def metadata_version(self, value): + self.c_options.metadata_version = _unwrap_metadata_version(value) + + @property + def compression(self): + if self.c_options.codec == nullptr: + return None + else: + return frombytes(self.c_options.codec.get().name()) + + @compression.setter + def compression(self, value): + if value is None: + self.c_options.codec.reset() + elif isinstance(value, str): + codec_type = _ensure_compression(value) + if codec_type != CCompressionType_ZSTD and codec_type != CCompressionType_LZ4_FRAME: + raise ValueError("Compression type must be lz4, zstd or None") + self.c_options.codec = shared_ptr[CCodec](GetResultValue( + CCodec.Create(codec_type)).release()) + elif isinstance(value, Codec): + if value.name != "lz4" and value.name != "zstd": + raise ValueError("Compression type must be lz4, zstd or None") + self.c_options.codec = (value).wrapped + else: + raise TypeError( + "Property `compression` must be None, str, or pyarrow.Codec") + + @property + def use_threads(self): + return self.c_options.use_threads + + @use_threads.setter + def use_threads(self, bint value): + self.c_options.use_threads = value + + @property + def emit_dictionary_deltas(self): + return self.c_options.emit_dictionary_deltas + + @emit_dictionary_deltas.setter + def emit_dictionary_deltas(self, bint value): + self.c_options.emit_dictionary_deltas = value + + @property + def unify_dictionaries(self): + return self.c_options.unify_dictionaries + + @unify_dictionaries.setter + def unify_dictionaries(self, bint value): + self.c_options.unify_dictionaries = value + + +cdef class Message(_Weakrefable): + """ + Container for an Arrow IPC message with metadata and optional body + """ + + def __cinit__(self): + pass + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly, use " + "`pyarrow.ipc.read_message` function instead." + .format(self.__class__.__name__)) + + @property + def type(self): + return frombytes(FormatMessageType(self.message.get().type())) + + @property + def metadata(self): + return pyarrow_wrap_buffer(self.message.get().metadata()) + + @property + def metadata_version(self): + return _wrap_metadata_version(self.message.get().metadata_version()) + + @property + def body(self): + cdef shared_ptr[CBuffer] body = self.message.get().body() + if body.get() == NULL: + return None + else: + return pyarrow_wrap_buffer(body) + + def equals(self, Message other): + """ + Returns True if the message contents (metadata and body) are identical + + Parameters + ---------- + other : Message + + Returns + ------- + are_equal : bool + """ + cdef c_bool result + with nogil: + result = self.message.get().Equals(deref(other.message.get())) + return result + + def serialize_to(self, NativeFile sink, alignment=8, memory_pool=None): + """ + Write message to generic OutputStream + + Parameters + ---------- + sink : NativeFile + alignment : int, default 8 + Byte alignment for metadata and body + memory_pool : MemoryPool, default None + Uses default memory pool if not specified + """ + cdef: + int64_t output_length = 0 + COutputStream* out + CIpcWriteOptions options + + options.alignment = alignment + out = sink.get_output_stream().get() + with nogil: + check_status(self.message.get() + .SerializeTo(out, options, &output_length)) + + def serialize(self, alignment=8, memory_pool=None): + """ + Write message as encapsulated IPC message + + Parameters + ---------- + alignment : int, default 8 + Byte alignment for metadata and body + memory_pool : MemoryPool, default None + Uses default memory pool if not specified + + Returns + ------- + serialized : Buffer + """ + stream = BufferOutputStream(memory_pool) + self.serialize_to(stream, alignment=alignment, memory_pool=memory_pool) + return stream.getvalue() + + def __repr__(self): + if self.message == nullptr: + return """pyarrow.Message(uninitialized)""" + + metadata_len = self.metadata.size + body = self.body + body_len = 0 if body is None else body.size + + return """pyarrow.Message +type: {0} +metadata length: {1} +body length: {2}""".format(self.type, metadata_len, body_len) + + +cdef class MessageReader(_Weakrefable): + """ + Interface for reading Message objects from some source (like an + InputStream) + """ + cdef: + unique_ptr[CMessageReader] reader + + def __cinit__(self): + pass + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly, use " + "`pyarrow.ipc.MessageReader.open_stream` function " + "instead.".format(self.__class__.__name__)) + + @staticmethod + def open_stream(source): + """ + Open stream from source, if you want to use memory map use + MemoryMappedFile as source. + + Parameters + ---------- + source : bytes/buffer-like, pyarrow.NativeFile, or file-like Python object + A readable source, like an InputStream + """ + cdef: + MessageReader result = MessageReader.__new__(MessageReader) + shared_ptr[CInputStream] in_stream + unique_ptr[CMessageReader] reader + + _get_input_stream(source, &in_stream) + with nogil: + reader = CMessageReader.Open(in_stream) + result.reader.reset(reader.release()) + + return result + + def __iter__(self): + return self + + def __next__(self): + return self.read_next_message() + + def read_next_message(self): + """ + Read next Message from the stream. + + Raises + ------ + StopIteration + At end of stream + """ + cdef Message result = Message.__new__(Message) + + with nogil: + result.message = move(GetResultValue(self.reader.get() + .ReadNextMessage())) + + if result.message.get() == NULL: + raise StopIteration + + return result + +# ---------------------------------------------------------------------- +# File and stream readers and writers + +cdef class _CRecordBatchWriter(_Weakrefable): + """The base RecordBatchWriter wrapper. + + Provides common implementations of convenience methods. Should not + be instantiated directly by user code. + """ + + # cdef block is in lib.pxd + + def write(self, table_or_batch): + """ + Write RecordBatch or Table to stream. + + Parameters + ---------- + table_or_batch : {RecordBatch, Table} + """ + if isinstance(table_or_batch, RecordBatch): + self.write_batch(table_or_batch) + elif isinstance(table_or_batch, Table): + self.write_table(table_or_batch) + else: + raise ValueError(type(table_or_batch)) + + def write_batch(self, RecordBatch batch, custom_metadata=None): + """ + Write RecordBatch to stream. + + Parameters + ---------- + batch : RecordBatch + custom_metadata : mapping or KeyValueMetadata + Keys and values must be string-like / coercible to bytes + """ + metadata = ensure_metadata(custom_metadata, allow_none=True) + c_meta = pyarrow_unwrap_metadata(metadata) + + with nogil: + check_status(self.writer.get() + .WriteRecordBatch(deref(batch.batch), c_meta)) + + def write_table(self, Table table, max_chunksize=None): + """ + Write Table to stream in (contiguous) RecordBatch objects. + + Parameters + ---------- + table : Table + max_chunksize : int, default None + Maximum number of rows for RecordBatch chunks. Individual chunks may + be smaller depending on the chunk layout of individual columns. + """ + cdef: + # max_chunksize must be > 0 to have any impact + int64_t c_max_chunksize = -1 + + if max_chunksize is not None: + c_max_chunksize = max_chunksize + + with nogil: + check_status(self.writer.get().WriteTable(table.table[0], + c_max_chunksize)) + + def close(self): + """ + Close stream and write end-of-stream 0 marker. + """ + with nogil: + check_status(self.writer.get().Close()) + + def __enter__(self): + return self + + def __exit__(self, exc_type, exc_val, exc_tb): + self.close() + + @property + def stats(self): + """ + Current IPC write statistics. + """ + if not self.writer: + raise ValueError("Operation on closed writer") + return _wrap_write_stats(self.writer.get().stats()) + + +cdef class _RecordBatchStreamWriter(_CRecordBatchWriter): + cdef: + CIpcWriteOptions options + bint closed + + def __cinit__(self): + pass + + def __dealloc__(self): + pass + + @property + def _use_legacy_format(self): + # For testing (see test_ipc.py) + return self.options.write_legacy_ipc_format + + @property + def _metadata_version(self): + # For testing (see test_ipc.py) + return _wrap_metadata_version(self.options.metadata_version) + + def _open(self, sink, Schema schema not None, + IpcWriteOptions options=IpcWriteOptions()): + cdef: + shared_ptr[COutputStream] c_sink + + self.options = options.c_options + get_writer(sink, &c_sink) + with nogil: + self.writer = GetResultValue( + MakeStreamWriter(c_sink, schema.sp_schema, + self.options)) + + +cdef _get_input_stream(object source, shared_ptr[CInputStream]* out): + try: + source = as_buffer(source) + except TypeError: + # Non-buffer-like + pass + + get_input_stream(source, True, out) + + +class _ReadPandasMixin: + + def read_pandas(self, **options): + """ + Read contents of stream to a pandas.DataFrame. + + Read all record batches as a pyarrow.Table then convert it to a + pandas.DataFrame using Table.to_pandas. + + Parameters + ---------- + **options + Arguments to forward to :meth:`Table.to_pandas`. + + Returns + ------- + df : pandas.DataFrame + """ + table = self.read_all() + return table.to_pandas(**options) + + +cdef class RecordBatchReader(_Weakrefable): + """Base class for reading stream of record batches. + + Record batch readers function as iterators of record batches that also + provide the schema (without the need to get any batches). + + Warnings + -------- + Do not call this class's constructor directly, use one of the + ``RecordBatchReader.from_*`` functions instead. + + Notes + ----- + To import and export using the Arrow C stream interface, use the + ``_import_from_c`` and ``_export_to_c`` methods. However, keep in mind this + interface is intended for expert users. + + Examples + -------- + >>> import pyarrow as pa + >>> schema = pa.schema([('x', pa.int64())]) + >>> def iter_record_batches(): + ... for i in range(2): + ... yield pa.RecordBatch.from_arrays([pa.array([1, 2, 3])], schema=schema) + >>> reader = pa.RecordBatchReader.from_batches(schema, iter_record_batches()) + >>> print(reader.schema) + x: int64 + >>> for batch in reader: + ... print(batch) + pyarrow.RecordBatch + x: int64 + ---- + x: [1,2,3] + pyarrow.RecordBatch + x: int64 + ---- + x: [1,2,3] + """ + + # cdef block is in lib.pxd + + def __init__(self): + raise TypeError("Do not call {}'s constructor directly, " + "use one of the RecordBatchReader.from_* functions instead." + .format(self.__class__.__name__)) + + def __iter__(self): + return self + + def __next__(self): + return self.read_next_batch() + + @property + def schema(self): + """ + Shared schema of the record batches in the stream. + + Returns + ------- + Schema + """ + cdef shared_ptr[CSchema] c_schema + + with nogil: + c_schema = self.reader.get().schema() + + return pyarrow_wrap_schema(c_schema) + + def read_next_batch(self): + """ + Read next RecordBatch from the stream. + + Raises + ------ + StopIteration: + At end of stream. + + Returns + ------- + RecordBatch + """ + cdef shared_ptr[CRecordBatch] batch + + with nogil: + check_status(self.reader.get().ReadNext(&batch)) + + if batch.get() == NULL: + raise StopIteration + + return pyarrow_wrap_batch(batch) + + def read_next_batch_with_custom_metadata(self): + """ + Read next RecordBatch from the stream along with its custom metadata. + + Raises + ------ + StopIteration: + At end of stream. + + Returns + ------- + batch : RecordBatch + custom_metadata : KeyValueMetadata + """ + cdef: + CRecordBatchWithMetadata batch_with_metadata + + with nogil: + batch_with_metadata = GetResultValue(self.reader.get().ReadNext()) + + if batch_with_metadata.batch.get() == NULL: + raise StopIteration + + return _wrap_record_batch_with_metadata(batch_with_metadata) + + def iter_batches_with_custom_metadata(self): + """ + Iterate over record batches from the stream along with their custom + metadata. + + Yields + ------ + RecordBatchWithMetadata + """ + while True: + try: + yield self.read_next_batch_with_custom_metadata() + except StopIteration: + return + + def read_all(self): + """ + Read all record batches as a pyarrow.Table. + + Returns + ------- + Table + """ + cdef shared_ptr[CTable] table + with nogil: + check_status(self.reader.get().ToTable().Value(&table)) + return pyarrow_wrap_table(table) + + read_pandas = _ReadPandasMixin.read_pandas + + def close(self): + """ + Release any resources associated with the reader. + """ + with nogil: + check_status(self.reader.get().Close()) + + def __enter__(self): + return self + + def __exit__(self, exc_type, exc_val, exc_tb): + self.close() + + def cast(self, target_schema): + """ + Wrap this reader with one that casts each batch lazily as it is pulled. + Currently only a safe cast to target_schema is implemented. + + Parameters + ---------- + target_schema : Schema + Schema to cast to, the names and order of fields must match. + + Returns + ------- + RecordBatchReader + """ + cdef: + shared_ptr[CSchema] c_schema + shared_ptr[CRecordBatchReader] c_reader + RecordBatchReader out + + if self.schema.names != target_schema.names: + raise ValueError("Target schema's field names are not matching " + f"the table's field names: {self.schema.names}, " + f"{target_schema.names}") + + c_schema = pyarrow_unwrap_schema(target_schema) + c_reader = GetResultValue(CCastingRecordBatchReader.Make( + self.reader, c_schema)) + + out = RecordBatchReader.__new__(RecordBatchReader) + out.reader = c_reader + return out + + def _export_to_c(self, out_ptr): + """ + Export to a C ArrowArrayStream struct, given its pointer. + + Parameters + ---------- + out_ptr: int + The raw pointer to a C ArrowArrayStream struct. + + Be careful: if you don't pass the ArrowArrayStream struct to a + consumer, array memory will leak. This is a low-level function + intended for expert users. + """ + cdef: + void* c_ptr = _as_c_pointer(out_ptr) + with nogil: + check_status(ExportRecordBatchReader( + self.reader, c_ptr)) + + @staticmethod + def _import_from_c(in_ptr): + """ + Import RecordBatchReader from a C ArrowArrayStream struct, + given its pointer. + + Parameters + ---------- + in_ptr: int + The raw pointer to a C ArrowArrayStream struct. + + This is a low-level function intended for expert users. + """ + cdef: + void* c_ptr = _as_c_pointer(in_ptr) + shared_ptr[CRecordBatchReader] c_reader + RecordBatchReader self + + with nogil: + c_reader = GetResultValue(ImportRecordBatchReader( + c_ptr)) + + self = RecordBatchReader.__new__(RecordBatchReader) + self.reader = c_reader + return self + + def __arrow_c_stream__(self, requested_schema=None): + """ + Export to a C ArrowArrayStream PyCapsule. + + Parameters + ---------- + requested_schema : PyCapsule, default None + The schema to which the stream should be casted, passed as a + PyCapsule containing a C ArrowSchema representation of the + requested schema. + + Returns + ------- + PyCapsule + A capsule containing a C ArrowArrayStream struct. + """ + cdef: + ArrowArrayStream* c_stream + + if requested_schema is not None: + out_schema = Schema._import_from_c_capsule(requested_schema) + if self.schema != out_schema: + return self.cast(out_schema).__arrow_c_stream__() + + stream_capsule = alloc_c_stream(&c_stream) + + with nogil: + check_status(ExportRecordBatchReader(self.reader, c_stream)) + + return stream_capsule + + @staticmethod + def _import_from_c_capsule(stream): + """ + Import RecordBatchReader from a C ArrowArrayStream PyCapsule. + + Parameters + ---------- + stream: PyCapsule + A capsule containing a C ArrowArrayStream PyCapsule. + + Returns + ------- + RecordBatchReader + """ + cdef: + ArrowArrayStream* c_stream + shared_ptr[CRecordBatchReader] c_reader + RecordBatchReader self + + c_stream = PyCapsule_GetPointer( + stream, 'arrow_array_stream' + ) + + with nogil: + c_reader = GetResultValue(ImportRecordBatchReader(c_stream)) + + self = RecordBatchReader.__new__(RecordBatchReader) + self.reader = c_reader + return self + + @staticmethod + def from_stream(data, schema=None): + """ + Create RecordBatchReader from a Arrow-compatible stream object. + + This accepts objects implementing the Arrow PyCapsule Protocol for + streams, i.e. objects that have a ``__arrow_c_stream__`` method. + + Parameters + ---------- + data : Arrow-compatible stream object + Any object that implements the Arrow PyCapsule Protocol for + streams. + schema : Schema, default None + The schema to which the stream should be casted, if supported + by the stream object. + + Returns + ------- + RecordBatchReader + """ + + if not hasattr(data, "__arrow_c_stream__"): + raise TypeError( + "Expected an object implementing the Arrow PyCapsule Protocol for " + "streams (i.e. having a `__arrow_c_stream__` method), " + f"got {type(data)!r}." + ) + + if schema is not None: + if not hasattr(schema, "__arrow_c_schema__"): + raise TypeError( + "Expected an object implementing the Arrow PyCapsule Protocol for " + "schema (i.e. having a `__arrow_c_schema__` method), " + f"got {type(schema)!r}." + ) + requested = schema.__arrow_c_schema__() + else: + requested = None + + capsule = data.__arrow_c_stream__(requested) + return RecordBatchReader._import_from_c_capsule(capsule) + + @staticmethod + def from_batches(Schema schema not None, batches): + """ + Create RecordBatchReader from an iterable of batches. + + Parameters + ---------- + schema : Schema + The shared schema of the record batches + batches : Iterable[RecordBatch] + The batches that this reader will return. + + Returns + ------- + reader : RecordBatchReader + """ + cdef: + shared_ptr[CSchema] c_schema + shared_ptr[CRecordBatchReader] c_reader + RecordBatchReader self + + c_schema = pyarrow_unwrap_schema(schema) + c_reader = GetResultValue(CPyRecordBatchReader.Make( + c_schema, batches)) + + self = RecordBatchReader.__new__(RecordBatchReader) + self.reader = c_reader + return self + + +cdef class _RecordBatchStreamReader(RecordBatchReader): + cdef: + shared_ptr[CInputStream] in_stream + CIpcReadOptions options + CRecordBatchStreamReader* stream_reader + + def __cinit__(self): + pass + + def _open(self, source, IpcReadOptions options=IpcReadOptions(), + MemoryPool memory_pool=None): + self.options = options.c_options + self.options.memory_pool = maybe_unbox_memory_pool(memory_pool) + _get_input_stream(source, &self.in_stream) + with nogil: + self.reader = GetResultValue(CRecordBatchStreamReader.Open( + self.in_stream, self.options)) + self.stream_reader = self.reader.get() + + @property + def stats(self): + """ + Current IPC read statistics. + """ + if not self.reader: + raise ValueError("Operation on closed reader") + return _wrap_read_stats(self.stream_reader.stats()) + + +cdef class _RecordBatchFileWriter(_RecordBatchStreamWriter): + + def _open(self, sink, Schema schema not None, + IpcWriteOptions options=IpcWriteOptions()): + cdef: + shared_ptr[COutputStream] c_sink + + self.options = options.c_options + get_writer(sink, &c_sink) + with nogil: + self.writer = GetResultValue( + MakeFileWriter(c_sink, schema.sp_schema, self.options)) + +_RecordBatchWithMetadata = namedtuple( + 'RecordBatchWithMetadata', + ('batch', 'custom_metadata')) + + +class RecordBatchWithMetadata(_RecordBatchWithMetadata): + """RecordBatch with its custom metadata + + Parameters + ---------- + batch : RecordBatch + custom_metadata : KeyValueMetadata + """ + __slots__ = () + + +@staticmethod +cdef _wrap_record_batch_with_metadata(CRecordBatchWithMetadata c): + return RecordBatchWithMetadata(pyarrow_wrap_batch(c.batch), + pyarrow_wrap_metadata(c.custom_metadata)) + + +cdef class _RecordBatchFileReader(_Weakrefable): + cdef: + SharedPtrNoGIL[CRecordBatchFileReader] reader + shared_ptr[CRandomAccessFile] file + CIpcReadOptions options + + cdef readonly: + Schema schema + + def __cinit__(self): + pass + + def _open(self, source, footer_offset=None, + IpcReadOptions options=IpcReadOptions(), + MemoryPool memory_pool=None): + self.options = options.c_options + self.options.memory_pool = maybe_unbox_memory_pool(memory_pool) + try: + source = as_buffer(source) + except TypeError: + pass + + get_reader(source, False, &self.file) + + cdef int64_t offset = 0 + if footer_offset is not None: + offset = footer_offset + + with nogil: + if offset != 0: + self.reader = GetResultValue( + CRecordBatchFileReader.Open2(self.file.get(), offset, + self.options)) + + else: + self.reader = GetResultValue( + CRecordBatchFileReader.Open(self.file.get(), + self.options)) + + self.schema = pyarrow_wrap_schema(self.reader.get().schema()) + + @property + def num_record_batches(self): + """ + The number of record batches in the IPC file. + """ + return self.reader.get().num_record_batches() + + def get_batch(self, int i): + """ + Read the record batch with the given index. + + Parameters + ---------- + i : int + The index of the record batch in the IPC file. + + Returns + ------- + batch : RecordBatch + """ + cdef shared_ptr[CRecordBatch] batch + + if i < 0 or i >= self.num_record_batches: + raise ValueError('Batch number {0} out of range'.format(i)) + + with nogil: + batch = GetResultValue(self.reader.get().ReadRecordBatch(i)) + + return pyarrow_wrap_batch(batch) + + # TODO(wesm): ARROW-503: Function was renamed. Remove after a period of + # time has passed + get_record_batch = get_batch + + def get_batch_with_custom_metadata(self, int i): + """ + Read the record batch with the given index along with + its custom metadata + + Parameters + ---------- + i : int + The index of the record batch in the IPC file. + + Returns + ------- + batch : RecordBatch + custom_metadata : KeyValueMetadata + """ + cdef: + CRecordBatchWithMetadata batch_with_metadata + + if i < 0 or i >= self.num_record_batches: + raise ValueError('Batch number {0} out of range'.format(i)) + + with nogil: + batch_with_metadata = GetResultValue( + self.reader.get().ReadRecordBatchWithCustomMetadata(i)) + + return _wrap_record_batch_with_metadata(batch_with_metadata) + + def read_all(self): + """ + Read all record batches as a pyarrow.Table + """ + cdef: + vector[shared_ptr[CRecordBatch]] batches + shared_ptr[CTable] table + int i, nbatches + + nbatches = self.num_record_batches + + batches.resize(nbatches) + with nogil: + for i in range(nbatches): + batches[i] = GetResultValue(self.reader.get() + .ReadRecordBatch(i)) + table = GetResultValue( + CTable.FromRecordBatches(self.schema.sp_schema, move(batches))) + + return pyarrow_wrap_table(table) + + read_pandas = _ReadPandasMixin.read_pandas + + def __enter__(self): + return self + + def __exit__(self, exc_type, exc_value, traceback): + pass + + @property + def stats(self): + """ + Current IPC read statistics. + """ + if not self.reader: + raise ValueError("Operation on closed reader") + return _wrap_read_stats(self.reader.get().stats()) + + +def get_tensor_size(Tensor tensor): + """ + Return total size of serialized Tensor including metadata and padding. + + Parameters + ---------- + tensor : Tensor + The tensor for which we want to known the size. + """ + cdef int64_t size + with nogil: + check_status(GetTensorSize(deref(tensor.tp), &size)) + return size + + +def get_record_batch_size(RecordBatch batch): + """ + Return total size of serialized RecordBatch including metadata and padding. + + Parameters + ---------- + batch : RecordBatch + The recordbatch for which we want to know the size. + """ + cdef int64_t size + with nogil: + check_status(GetRecordBatchSize(deref(batch.batch), &size)) + return size + + +def write_tensor(Tensor tensor, NativeFile dest): + """ + Write pyarrow.Tensor to pyarrow.NativeFile object its current position. + + Parameters + ---------- + tensor : pyarrow.Tensor + dest : pyarrow.NativeFile + + Returns + ------- + bytes_written : int + Total number of bytes written to the file + """ + cdef: + int32_t metadata_length + int64_t body_length + + handle = dest.get_output_stream() + + with nogil: + check_status( + WriteTensor(deref(tensor.tp), handle.get(), + &metadata_length, &body_length)) + + return metadata_length + body_length + + +cdef NativeFile as_native_file(source): + if not isinstance(source, NativeFile): + if hasattr(source, 'read'): + source = PythonFile(source) + else: + source = BufferReader(source) + + if not isinstance(source, NativeFile): + raise ValueError('Unable to read message from object with type: {0}' + .format(type(source))) + return source + + +def read_tensor(source): + """Read pyarrow.Tensor from pyarrow.NativeFile object from current + position. If the file source supports zero copy (e.g. a memory map), then + this operation does not allocate any memory. This function not assume that + the stream is aligned + + Parameters + ---------- + source : pyarrow.NativeFile + + Returns + ------- + tensor : Tensor + + """ + cdef: + shared_ptr[CTensor] sp_tensor + CInputStream* c_stream + NativeFile nf = as_native_file(source) + + c_stream = nf.get_input_stream().get() + with nogil: + sp_tensor = GetResultValue(ReadTensor(c_stream)) + return pyarrow_wrap_tensor(sp_tensor) + + +def read_message(source): + """ + Read length-prefixed message from file or buffer-like object + + Parameters + ---------- + source : pyarrow.NativeFile, file-like object, or buffer-like object + + Returns + ------- + message : Message + """ + cdef: + Message result = Message.__new__(Message) + CInputStream* c_stream + + cdef NativeFile nf = as_native_file(source) + c_stream = nf.get_input_stream().get() + + with nogil: + result.message = move( + GetResultValue(ReadMessage(c_stream, c_default_memory_pool()))) + + if result.message == nullptr: + raise EOFError("End of Arrow stream") + + return result + + +def read_schema(obj, DictionaryMemo dictionary_memo=None): + """ + Read Schema from message or buffer + + Parameters + ---------- + obj : buffer or Message + dictionary_memo : DictionaryMemo, optional + Needed to be able to reconstruct dictionary-encoded fields + with read_record_batch + + Returns + ------- + schema : Schema + """ + cdef: + shared_ptr[CSchema] result + shared_ptr[CRandomAccessFile] cpp_file + Message message + CDictionaryMemo temp_memo + CDictionaryMemo* arg_dict_memo + + if dictionary_memo is not None: + arg_dict_memo = dictionary_memo.memo + else: + arg_dict_memo = &temp_memo + + if isinstance(obj, Message): + message = obj + with nogil: + result = GetResultValue(ReadSchema( + deref(message.message.get()), arg_dict_memo)) + else: + get_reader(obj, False, &cpp_file) + with nogil: + result = GetResultValue(ReadSchema(cpp_file.get(), arg_dict_memo)) + + return pyarrow_wrap_schema(result) + + +def read_record_batch(obj, Schema schema, + DictionaryMemo dictionary_memo=None): + """ + Read RecordBatch from message, given a known schema. If reading data from a + complete IPC stream, use ipc.open_stream instead + + Parameters + ---------- + obj : Message or Buffer-like + schema : Schema + dictionary_memo : DictionaryMemo, optional + If message contains dictionaries, must pass a populated + DictionaryMemo + + Returns + ------- + batch : RecordBatch + """ + cdef: + shared_ptr[CRecordBatch] result + Message message + CDictionaryMemo temp_memo + CDictionaryMemo* arg_dict_memo + + if isinstance(obj, Message): + message = obj + else: + message = read_message(obj) + + if dictionary_memo is not None: + arg_dict_memo = dictionary_memo.memo + else: + arg_dict_memo = &temp_memo + + with nogil: + result = GetResultValue( + ReadRecordBatch(deref(message.message.get()), + schema.sp_schema, + arg_dict_memo, + CIpcReadOptions.Defaults())) + + return pyarrow_wrap_batch(result) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/jvm.py b/vllm/lib/python3.10/site-packages/pyarrow/jvm.py new file mode 100644 index 0000000000000000000000000000000000000000..161c5ff4d6d74512dfcd76ddac5a4c4781ad63c3 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/jvm.py @@ -0,0 +1,335 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +""" +Functions to interact with Arrow memory allocated by Arrow Java. + +These functions convert the objects holding the metadata, the actual +data is not copied at all. + +This will only work with a JVM running in the same process such as provided +through jpype. Modules that talk to a remote JVM like py4j will not work as the +memory addresses reported by them are not reachable in the python process. +""" + +import pyarrow as pa + + +class _JvmBufferNanny: + """ + An object that keeps a org.apache.arrow.memory.ArrowBuf's underlying + memory alive. + """ + ref_manager = None + + def __init__(self, jvm_buf): + ref_manager = jvm_buf.getReferenceManager() + # Will raise a java.lang.IllegalArgumentException if the buffer + # is already freed. It seems that exception cannot easily be + # caught... + ref_manager.retain() + self.ref_manager = ref_manager + + def __del__(self): + if self.ref_manager is not None: + self.ref_manager.release() + + +def jvm_buffer(jvm_buf): + """ + Construct an Arrow buffer from org.apache.arrow.memory.ArrowBuf + + Parameters + ---------- + + jvm_buf: org.apache.arrow.memory.ArrowBuf + Arrow Buffer representation on the JVM. + + Returns + ------- + pyarrow.Buffer + Python Buffer that references the JVM memory. + """ + nanny = _JvmBufferNanny(jvm_buf) + address = jvm_buf.memoryAddress() + size = jvm_buf.capacity() + return pa.foreign_buffer(address, size, base=nanny) + + +def _from_jvm_int_type(jvm_type): + """ + Convert a JVM int type to its Python equivalent. + + Parameters + ---------- + jvm_type : org.apache.arrow.vector.types.pojo.ArrowType$Int + + Returns + ------- + typ : pyarrow.DataType + """ + + bit_width = jvm_type.getBitWidth() + if jvm_type.getIsSigned(): + if bit_width == 8: + return pa.int8() + elif bit_width == 16: + return pa.int16() + elif bit_width == 32: + return pa.int32() + elif bit_width == 64: + return pa.int64() + else: + if bit_width == 8: + return pa.uint8() + elif bit_width == 16: + return pa.uint16() + elif bit_width == 32: + return pa.uint32() + elif bit_width == 64: + return pa.uint64() + + +def _from_jvm_float_type(jvm_type): + """ + Convert a JVM float type to its Python equivalent. + + Parameters + ---------- + jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$FloatingPoint + + Returns + ------- + typ: pyarrow.DataType + """ + precision = jvm_type.getPrecision().toString() + if precision == 'HALF': + return pa.float16() + elif precision == 'SINGLE': + return pa.float32() + elif precision == 'DOUBLE': + return pa.float64() + + +def _from_jvm_time_type(jvm_type): + """ + Convert a JVM time type to its Python equivalent. + + Parameters + ---------- + jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$Time + + Returns + ------- + typ: pyarrow.DataType + """ + time_unit = jvm_type.getUnit().toString() + if time_unit == 'SECOND': + assert jvm_type.getBitWidth() == 32 + return pa.time32('s') + elif time_unit == 'MILLISECOND': + assert jvm_type.getBitWidth() == 32 + return pa.time32('ms') + elif time_unit == 'MICROSECOND': + assert jvm_type.getBitWidth() == 64 + return pa.time64('us') + elif time_unit == 'NANOSECOND': + assert jvm_type.getBitWidth() == 64 + return pa.time64('ns') + + +def _from_jvm_timestamp_type(jvm_type): + """ + Convert a JVM timestamp type to its Python equivalent. + + Parameters + ---------- + jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$Timestamp + + Returns + ------- + typ: pyarrow.DataType + """ + time_unit = jvm_type.getUnit().toString() + timezone = jvm_type.getTimezone() + if timezone is not None: + timezone = str(timezone) + if time_unit == 'SECOND': + return pa.timestamp('s', tz=timezone) + elif time_unit == 'MILLISECOND': + return pa.timestamp('ms', tz=timezone) + elif time_unit == 'MICROSECOND': + return pa.timestamp('us', tz=timezone) + elif time_unit == 'NANOSECOND': + return pa.timestamp('ns', tz=timezone) + + +def _from_jvm_date_type(jvm_type): + """ + Convert a JVM date type to its Python equivalent + + Parameters + ---------- + jvm_type: org.apache.arrow.vector.types.pojo.ArrowType$Date + + Returns + ------- + typ: pyarrow.DataType + """ + day_unit = jvm_type.getUnit().toString() + if day_unit == 'DAY': + return pa.date32() + elif day_unit == 'MILLISECOND': + return pa.date64() + + +def field(jvm_field): + """ + Construct a Field from a org.apache.arrow.vector.types.pojo.Field + instance. + + Parameters + ---------- + jvm_field: org.apache.arrow.vector.types.pojo.Field + + Returns + ------- + pyarrow.Field + """ + name = str(jvm_field.getName()) + jvm_type = jvm_field.getType() + + typ = None + if not jvm_type.isComplex(): + type_str = jvm_type.getTypeID().toString() + if type_str == 'Null': + typ = pa.null() + elif type_str == 'Int': + typ = _from_jvm_int_type(jvm_type) + elif type_str == 'FloatingPoint': + typ = _from_jvm_float_type(jvm_type) + elif type_str == 'Utf8': + typ = pa.string() + elif type_str == 'Binary': + typ = pa.binary() + elif type_str == 'FixedSizeBinary': + typ = pa.binary(jvm_type.getByteWidth()) + elif type_str == 'Bool': + typ = pa.bool_() + elif type_str == 'Time': + typ = _from_jvm_time_type(jvm_type) + elif type_str == 'Timestamp': + typ = _from_jvm_timestamp_type(jvm_type) + elif type_str == 'Date': + typ = _from_jvm_date_type(jvm_type) + elif type_str == 'Decimal': + typ = pa.decimal128(jvm_type.getPrecision(), jvm_type.getScale()) + else: + raise NotImplementedError( + "Unsupported JVM type: {}".format(type_str)) + else: + # TODO: The following JVM types are not implemented: + # Struct, List, FixedSizeList, Union, Dictionary + raise NotImplementedError( + "JVM field conversion only implemented for primitive types.") + + nullable = jvm_field.isNullable() + jvm_metadata = jvm_field.getMetadata() + if jvm_metadata.isEmpty(): + metadata = None + else: + metadata = {str(entry.getKey()): str(entry.getValue()) + for entry in jvm_metadata.entrySet()} + return pa.field(name, typ, nullable, metadata) + + +def schema(jvm_schema): + """ + Construct a Schema from a org.apache.arrow.vector.types.pojo.Schema + instance. + + Parameters + ---------- + jvm_schema: org.apache.arrow.vector.types.pojo.Schema + + Returns + ------- + pyarrow.Schema + """ + fields = jvm_schema.getFields() + fields = [field(f) for f in fields] + jvm_metadata = jvm_schema.getCustomMetadata() + if jvm_metadata.isEmpty(): + metadata = None + else: + metadata = {str(entry.getKey()): str(entry.getValue()) + for entry in jvm_metadata.entrySet()} + return pa.schema(fields, metadata) + + +def array(jvm_array): + """ + Construct an (Python) Array from its JVM equivalent. + + Parameters + ---------- + jvm_array : org.apache.arrow.vector.ValueVector + + Returns + ------- + array : Array + """ + if jvm_array.getField().getType().isComplex(): + minor_type_str = jvm_array.getMinorType().toString() + raise NotImplementedError( + "Cannot convert JVM Arrow array of type {}," + " complex types not yet implemented.".format(minor_type_str)) + dtype = field(jvm_array.getField()).type + buffers = [jvm_buffer(buf) + for buf in list(jvm_array.getBuffers(False))] + + # If JVM has an empty Vector, buffer list will be empty so create manually + if len(buffers) == 0: + return pa.array([], type=dtype) + + length = jvm_array.getValueCount() + null_count = jvm_array.getNullCount() + return pa.Array.from_buffers(dtype, length, buffers, null_count) + + +def record_batch(jvm_vector_schema_root): + """ + Construct a (Python) RecordBatch from a JVM VectorSchemaRoot + + Parameters + ---------- + jvm_vector_schema_root : org.apache.arrow.vector.VectorSchemaRoot + + Returns + ------- + record_batch: pyarrow.RecordBatch + """ + pa_schema = schema(jvm_vector_schema_root.getSchema()) + + arrays = [] + for name in pa_schema.names: + arrays.append(array(jvm_vector_schema_root.getVector(name))) + + return pa.RecordBatch.from_arrays( + arrays, + pa_schema.names, + metadata=pa_schema.metadata + ) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/lib.pxd b/vllm/lib/python3.10/site-packages/pyarrow/lib.pxd new file mode 100644 index 0000000000000000000000000000000000000000..25a7945dc3ddcb65285ccc2ff07b024d78ff6b96 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/lib.pxd @@ -0,0 +1,747 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# cython: language_level = 3 + +from cpython cimport PyObject +from libcpp cimport nullptr, bool as c_bool +from libcpp.cast cimport dynamic_cast +from libcpp.memory cimport dynamic_pointer_cast +from pyarrow.includes.common cimport * +from pyarrow.includes.libarrow cimport * +from pyarrow.includes.libarrow_python cimport * + +# Will be available in Cython 3, not backported +# ref: https://github.com/cython/cython/issues/3293#issuecomment-1223058101 +cdef extern from "" namespace "std" nogil: + cdef cppclass nullopt_t: + nullopt_t() + + cdef nullopt_t nullopt + + cdef cppclass optional[T]: + ctypedef T value_type + optional() + optional(nullopt_t) + optional(optional&) except + + optional(T&) except + + c_bool has_value() + T& value() + T& value_or[U](U& default_value) + void swap(optional&) + void reset() + T& emplace(...) + T& operator*() + # T* operator->() # Not Supported + optional& operator=(optional&) + optional& operator=[U](U&) + c_bool operator bool() + c_bool operator!() + c_bool operator==[U](optional&, U&) + c_bool operator!=[U](optional&, U&) + c_bool operator<[U](optional&, U&) + c_bool operator>[U](optional&, U&) + c_bool operator<=[U](optional&, U&) + c_bool operator>=[U](optional&, U&) + + optional[T] make_optional[T](...) except + + +cdef extern from "Python.h": + int PySlice_Check(object) + + +cdef int check_status(const CStatus& status) except -1 nogil +cdef object convert_status(const CStatus& status) + + +cdef class _Weakrefable: + cdef object __weakref__ + + +cdef class IpcWriteOptions(_Weakrefable): + cdef: + CIpcWriteOptions c_options + + +cdef class IpcReadOptions(_Weakrefable): + cdef: + CIpcReadOptions c_options + + +cdef class Message(_Weakrefable): + cdef: + unique_ptr[CMessage] message + + +cdef class MemoryPool(_Weakrefable): + cdef: + CMemoryPool* pool + + cdef void init(self, CMemoryPool* pool) + + +cdef CMemoryPool* maybe_unbox_memory_pool(MemoryPool memory_pool) + + +cdef object box_memory_pool(CMemoryPool* pool) + + +cdef class DataType(_Weakrefable): + cdef: + shared_ptr[CDataType] sp_type + CDataType* type + bytes pep3118_format + + cdef void init(self, const shared_ptr[CDataType]& type) except * + cpdef Field field(self, i) + + +cdef class ListType(DataType): + cdef: + const CListType* list_type + + +cdef class LargeListType(DataType): + cdef: + const CLargeListType* list_type + + +cdef class ListViewType(DataType): + cdef: + const CListViewType* list_view_type + + +cdef class LargeListViewType(DataType): + cdef: + const CLargeListViewType* list_view_type + + +cdef class MapType(DataType): + cdef: + const CMapType* map_type + + +cdef class FixedSizeListType(DataType): + cdef: + const CFixedSizeListType* list_type + + +cdef class StructType(DataType): + cdef: + const CStructType* struct_type + + cdef Field field_by_name(self, name) + + +cdef class DictionaryMemo(_Weakrefable): + cdef: + # Even though the CDictionaryMemo instance is private, we allocate + # it on the heap so as to avoid C++ ABI issues with Python wheels. + shared_ptr[CDictionaryMemo] sp_memo + CDictionaryMemo* memo + + +cdef class DictionaryType(DataType): + cdef: + const CDictionaryType* dict_type + + +cdef class TimestampType(DataType): + cdef: + const CTimestampType* ts_type + + +cdef class Time32Type(DataType): + cdef: + const CTime32Type* time_type + + +cdef class Time64Type(DataType): + cdef: + const CTime64Type* time_type + + +cdef class DurationType(DataType): + cdef: + const CDurationType* duration_type + + +cdef class FixedSizeBinaryType(DataType): + cdef: + const CFixedSizeBinaryType* fixed_size_binary_type + + +cdef class Decimal128Type(FixedSizeBinaryType): + cdef: + const CDecimal128Type* decimal128_type + + +cdef class Decimal256Type(FixedSizeBinaryType): + cdef: + const CDecimal256Type* decimal256_type + + +cdef class RunEndEncodedType(DataType): + cdef: + const CRunEndEncodedType* run_end_encoded_type + + +cdef class BaseExtensionType(DataType): + cdef: + const CExtensionType* ext_type + + +cdef class ExtensionType(BaseExtensionType): + cdef: + const CPyExtensionType* cpy_ext_type + + +cdef class FixedShapeTensorType(BaseExtensionType): + cdef: + const CFixedShapeTensorType* tensor_ext_type + +cdef class Bool8Type(BaseExtensionType): + cdef: + const CBool8Type* bool8_ext_type + +cdef class OpaqueType(BaseExtensionType): + cdef: + const COpaqueType* opaque_ext_type + +cdef class UuidType(BaseExtensionType): + cdef: + const CUuidType* uuid_ext_type + +cdef class PyExtensionType(ExtensionType): + pass + + +cdef class _Metadata(_Weakrefable): + # required because KeyValueMetadata also extends collections.abc.Mapping + # and the first parent class must be an extension type + pass + + +cdef class KeyValueMetadata(_Metadata): + cdef: + shared_ptr[const CKeyValueMetadata] wrapped + const CKeyValueMetadata* metadata + + cdef void init(self, const shared_ptr[const CKeyValueMetadata]& wrapped) + + @staticmethod + cdef wrap(const shared_ptr[const CKeyValueMetadata]& sp) + cdef inline shared_ptr[const CKeyValueMetadata] unwrap(self) nogil + + +cdef class Field(_Weakrefable): + cdef: + shared_ptr[CField] sp_field + CField* field + + cdef readonly: + DataType type + + cdef void init(self, const shared_ptr[CField]& field) + + +cdef class Schema(_Weakrefable): + cdef: + shared_ptr[CSchema] sp_schema + CSchema* schema + + cdef void init(self, const vector[shared_ptr[CField]]& fields) + cdef void init_schema(self, const shared_ptr[CSchema]& schema) + + +cdef class Scalar(_Weakrefable): + cdef: + shared_ptr[CScalar] wrapped + + cdef void init(self, const shared_ptr[CScalar]& wrapped) + + @staticmethod + cdef wrap(const shared_ptr[CScalar]& wrapped) + + cdef inline shared_ptr[CScalar] unwrap(self) nogil + + +cdef class _PandasConvertible(_Weakrefable): + pass + + +cdef class Array(_PandasConvertible): + cdef: + shared_ptr[CArray] sp_array + CArray* ap + + cdef readonly: + DataType type + # To allow Table to propagate metadata to pandas.Series + object _name + + cdef void init(self, const shared_ptr[CArray]& sp_array) except * + cdef getitem(self, int64_t i) + cdef int64_t length(self) + cdef void _assert_cpu(self) except * + + +cdef class Tensor(_Weakrefable): + cdef: + shared_ptr[CTensor] sp_tensor + CTensor* tp + + cdef readonly: + DataType type + bytes _ssize_t_shape + bytes _ssize_t_strides + + cdef void init(self, const shared_ptr[CTensor]& sp_tensor) + + +cdef class SparseCSRMatrix(_Weakrefable): + cdef: + shared_ptr[CSparseCSRMatrix] sp_sparse_tensor + CSparseCSRMatrix* stp + + cdef readonly: + DataType type + + cdef void init(self, const shared_ptr[CSparseCSRMatrix]& sp_sparse_tensor) + + +cdef class SparseCSCMatrix(_Weakrefable): + cdef: + shared_ptr[CSparseCSCMatrix] sp_sparse_tensor + CSparseCSCMatrix* stp + + cdef readonly: + DataType type + + cdef void init(self, const shared_ptr[CSparseCSCMatrix]& sp_sparse_tensor) + + +cdef class SparseCOOTensor(_Weakrefable): + cdef: + shared_ptr[CSparseCOOTensor] sp_sparse_tensor + CSparseCOOTensor* stp + + cdef readonly: + DataType type + + cdef void init(self, const shared_ptr[CSparseCOOTensor]& sp_sparse_tensor) + + +cdef class SparseCSFTensor(_Weakrefable): + cdef: + shared_ptr[CSparseCSFTensor] sp_sparse_tensor + CSparseCSFTensor* stp + + cdef readonly: + DataType type + + cdef void init(self, const shared_ptr[CSparseCSFTensor]& sp_sparse_tensor) + + +cdef class NullArray(Array): + pass + + +cdef class BooleanArray(Array): + pass + + +cdef class NumericArray(Array): + pass + + +cdef class IntegerArray(NumericArray): + pass + + +cdef class FloatingPointArray(NumericArray): + pass + + +cdef class Int8Array(IntegerArray): + pass + + +cdef class UInt8Array(IntegerArray): + pass + + +cdef class Int16Array(IntegerArray): + pass + + +cdef class UInt16Array(IntegerArray): + pass + + +cdef class Int32Array(IntegerArray): + pass + + +cdef class UInt32Array(IntegerArray): + pass + + +cdef class Int64Array(IntegerArray): + pass + + +cdef class UInt64Array(IntegerArray): + pass + + +cdef class HalfFloatArray(FloatingPointArray): + pass + + +cdef class FloatArray(FloatingPointArray): + pass + + +cdef class DoubleArray(FloatingPointArray): + pass + + +cdef class FixedSizeBinaryArray(Array): + pass + + +cdef class Decimal128Array(FixedSizeBinaryArray): + pass + + +cdef class Decimal256Array(FixedSizeBinaryArray): + pass + + +cdef class StructArray(Array): + pass + + +cdef class BaseListArray(Array): + pass + + +cdef class ListArray(BaseListArray): + pass + + +cdef class LargeListArray(BaseListArray): + pass + + +cdef class ListViewArray(BaseListArray): + pass + + +cdef class LargeListViewArray(BaseListArray): + pass + + +cdef class MapArray(ListArray): + pass + + +cdef class FixedSizeListArray(BaseListArray): + pass + + +cdef class UnionArray(Array): + pass + + +cdef class StringArray(Array): + pass + + +cdef class BinaryArray(Array): + pass + + +cdef class StringViewArray(Array): + pass + + +cdef class BinaryViewArray(Array): + pass + + +cdef class DictionaryArray(Array): + cdef: + object _indices, _dictionary + + +cdef class ExtensionArray(Array): + pass + + +cdef class MonthDayNanoIntervalArray(Array): + pass + + +cdef wrap_array_output(PyObject* output) +cdef wrap_datum(const CDatum& datum) + + +cdef class ChunkedArray(_PandasConvertible): + cdef: + shared_ptr[CChunkedArray] sp_chunked_array + CChunkedArray* chunked_array + c_bool _is_cpu + c_bool _init_is_cpu + + cdef readonly: + # To allow Table to propagate metadata to pandas.Series + object _name + + cdef void init(self, const shared_ptr[CChunkedArray]& chunked_array) + cdef getitem(self, int64_t i) + + +cdef class _Tabular(_PandasConvertible): + cdef void _assert_cpu(self) except * + + +cdef class Table(_Tabular): + cdef: + shared_ptr[CTable] sp_table + CTable* table + c_bool _is_cpu + c_bool _init_is_cpu + + cdef void init(self, const shared_ptr[CTable]& table) + + +cdef class RecordBatch(_Tabular): + cdef: + shared_ptr[CRecordBatch] sp_batch + CRecordBatch* batch + Schema _schema + + cdef void init(self, const shared_ptr[CRecordBatch]& table) + + +cdef class Device(_Weakrefable): + cdef: + shared_ptr[CDevice] device + + cdef void init(self, const shared_ptr[CDevice]& device) + + @staticmethod + cdef wrap(const shared_ptr[CDevice]& device) + + cdef inline shared_ptr[CDevice] unwrap(self) nogil + + +cdef class MemoryManager(_Weakrefable): + cdef: + shared_ptr[CMemoryManager] memory_manager + + cdef void init(self, const shared_ptr[CMemoryManager]& memory_manager) + + @staticmethod + cdef wrap(const shared_ptr[CMemoryManager]& mm) + + cdef inline shared_ptr[CMemoryManager] unwrap(self) nogil + + +cdef class Buffer(_Weakrefable): + cdef: + shared_ptr[CBuffer] buffer + Py_ssize_t shape[1] + Py_ssize_t strides[1] + + cdef void init(self, const shared_ptr[CBuffer]& buffer) + cdef getitem(self, int64_t i) + + +cdef class ResizableBuffer(Buffer): + + cdef void init_rz(self, const shared_ptr[CResizableBuffer]& buffer) + + +cdef class NativeFile(_Weakrefable): + cdef: + shared_ptr[CInputStream] input_stream + shared_ptr[CRandomAccessFile] random_access + shared_ptr[COutputStream] output_stream + bint is_readable + bint is_writable + bint is_seekable + bint _is_appending + bint own_file + + # By implementing these "virtual" functions (all functions in Cython + # extension classes are technically virtual in the C++ sense) we can expose + # the arrow::io abstract file interfaces to other components throughout the + # suite of Arrow C++ libraries + cdef set_random_access_file(self, shared_ptr[CRandomAccessFile] handle) + cdef set_input_stream(self, shared_ptr[CInputStream] handle) + cdef set_output_stream(self, shared_ptr[COutputStream] handle) + + cdef shared_ptr[CRandomAccessFile] get_random_access_file(self) except * + cdef shared_ptr[CInputStream] get_input_stream(self) except * + cdef shared_ptr[COutputStream] get_output_stream(self) except * + + +cdef class BufferedInputStream(NativeFile): + pass + + +cdef class BufferedOutputStream(NativeFile): + pass + + +cdef class CompressedInputStream(NativeFile): + pass + + +cdef class CompressedOutputStream(NativeFile): + pass + + +cdef class _CRecordBatchWriter(_Weakrefable): + cdef: + SharedPtrNoGIL[CRecordBatchWriter] writer + + +cdef class RecordBatchReader(_Weakrefable): + cdef: + SharedPtrNoGIL[CRecordBatchReader] reader + + +cdef class CacheOptions(_Weakrefable): + cdef: + CCacheOptions wrapped + + cdef void init(self, CCacheOptions options) + + cdef inline CCacheOptions unwrap(self) + + @staticmethod + cdef wrap(const CCacheOptions options) + + +cdef class Codec(_Weakrefable): + cdef: + shared_ptr[CCodec] wrapped + + cdef inline CCodec* unwrap(self) nogil + + +# This class is only used internally for now +cdef class StopToken: + cdef: + CStopToken stop_token + + cdef void init(self, CStopToken stop_token) + + +cdef get_input_stream(object source, c_bool use_memory_map, + shared_ptr[CInputStream]* reader) +cdef get_reader(object source, c_bool use_memory_map, + shared_ptr[CRandomAccessFile]* reader) +cdef get_writer(object source, shared_ptr[COutputStream]* writer) +cdef NativeFile get_native_file(object source, c_bool use_memory_map) + +cdef shared_ptr[CInputStream] native_transcoding_input_stream( + shared_ptr[CInputStream] stream, src_encoding, + dest_encoding) except * + +cdef shared_ptr[function[StreamWrapFunc]] make_streamwrap_func( + src_encoding, dest_encoding) except * + +# Default is allow_none=False +cpdef DataType ensure_type(object type, bint allow_none=*) + +cdef timeunit_to_string(TimeUnit unit) +cdef TimeUnit string_to_timeunit(unit) except * + +# Exceptions may be raised when converting dict values, so need to +# check exception state on return +cdef shared_ptr[const CKeyValueMetadata] pyarrow_unwrap_metadata( + object meta) except * +cdef object pyarrow_wrap_metadata( + const shared_ptr[const CKeyValueMetadata]& meta) + +# +# Public Cython API for 3rd party code +# +# If you add functions to this list, please also update +# `cpp/src/arrow/python/pyarrow.{h, cc}` +# + +# Wrapping C++ -> Python + +cdef public object pyarrow_wrap_buffer(const shared_ptr[CBuffer]& buf) +cdef public object pyarrow_wrap_resizable_buffer( + const shared_ptr[CResizableBuffer]& buf) + +cdef public object pyarrow_wrap_data_type(const shared_ptr[CDataType]& type) +cdef public object pyarrow_wrap_field(const shared_ptr[CField]& field) +cdef public object pyarrow_wrap_schema(const shared_ptr[CSchema]& type) + +cdef public object pyarrow_wrap_scalar(const shared_ptr[CScalar]& sp_scalar) + +cdef public object pyarrow_wrap_array(const shared_ptr[CArray]& sp_array) +cdef public object pyarrow_wrap_chunked_array( + const shared_ptr[CChunkedArray]& sp_array) + +cdef public object pyarrow_wrap_sparse_coo_tensor( + const shared_ptr[CSparseCOOTensor]& sp_sparse_tensor) +cdef public object pyarrow_wrap_sparse_csc_matrix( + const shared_ptr[CSparseCSCMatrix]& sp_sparse_tensor) +cdef public object pyarrow_wrap_sparse_csf_tensor( + const shared_ptr[CSparseCSFTensor]& sp_sparse_tensor) +cdef public object pyarrow_wrap_sparse_csr_matrix( + const shared_ptr[CSparseCSRMatrix]& sp_sparse_tensor) +cdef public object pyarrow_wrap_tensor(const shared_ptr[CTensor]& sp_tensor) + +cdef public object pyarrow_wrap_batch(const shared_ptr[CRecordBatch]& cbatch) +cdef public object pyarrow_wrap_table(const shared_ptr[CTable]& ctable) + +# Unwrapping Python -> C++ + +cdef public shared_ptr[CBuffer] pyarrow_unwrap_buffer(object buffer) + +cdef public shared_ptr[CDataType] pyarrow_unwrap_data_type(object data_type) +cdef public shared_ptr[CField] pyarrow_unwrap_field(object field) +cdef public shared_ptr[CSchema] pyarrow_unwrap_schema(object schema) + +cdef public shared_ptr[CScalar] pyarrow_unwrap_scalar(object scalar) + +cdef public shared_ptr[CArray] pyarrow_unwrap_array(object array) +cdef public shared_ptr[CChunkedArray] pyarrow_unwrap_chunked_array( + object array) + +cdef public shared_ptr[CSparseCOOTensor] pyarrow_unwrap_sparse_coo_tensor( + object sparse_tensor) +cdef public shared_ptr[CSparseCSCMatrix] pyarrow_unwrap_sparse_csc_matrix( + object sparse_tensor) +cdef public shared_ptr[CSparseCSFTensor] pyarrow_unwrap_sparse_csf_tensor( + object sparse_tensor) +cdef public shared_ptr[CSparseCSRMatrix] pyarrow_unwrap_sparse_csr_matrix( + object sparse_tensor) +cdef public shared_ptr[CTensor] pyarrow_unwrap_tensor(object tensor) + +cdef public shared_ptr[CRecordBatch] pyarrow_unwrap_batch(object batch) +cdef public shared_ptr[CTable] pyarrow_unwrap_table(object table) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/libarrow_acero.so.1801 b/vllm/lib/python3.10/site-packages/pyarrow/libarrow_acero.so.1801 new file mode 100644 index 0000000000000000000000000000000000000000..6974c9a49ef7195c66de5a00cee7821f806d615a --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/libarrow_acero.so.1801 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9c5a5ae814e75fba27dabd0fcd032f665e0de855b38b956fde5dad1a2279c49 +size 2086504 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/libarrow_dataset.so.1801 b/vllm/lib/python3.10/site-packages/pyarrow/libarrow_dataset.so.1801 new file mode 100644 index 0000000000000000000000000000000000000000..a6ef6d812cef821967db82c4d09e1c180a090213 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/libarrow_dataset.so.1801 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71c2a4fa551b629f0f0dcdac19454163f878ef90e771ce77332e473a560186c9 +size 2763968 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/libarrow_python_parquet_encryption.so b/vllm/lib/python3.10/site-packages/pyarrow/libarrow_python_parquet_encryption.so new file mode 100644 index 0000000000000000000000000000000000000000..892a9116bf135db422dc5125396d3580cc849cb8 Binary files /dev/null and b/vllm/lib/python3.10/site-packages/pyarrow/libarrow_python_parquet_encryption.so differ diff --git a/vllm/lib/python3.10/site-packages/pyarrow/libparquet.so.1801 b/vllm/lib/python3.10/site-packages/pyarrow/libparquet.so.1801 new file mode 100644 index 0000000000000000000000000000000000000000..f0b9d2b2f0e31e62725b523f9e3ae4c963cfd3e9 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/libparquet.so.1801 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12a65c1e854352c314e7d10eaa02dcfad07b9e0552855907d310497519f6e586 +size 11042272 diff --git a/vllm/lib/python3.10/site-packages/pyarrow/orc.py b/vllm/lib/python3.10/site-packages/pyarrow/orc.py new file mode 100644 index 0000000000000000000000000000000000000000..6f5feafb3363eb35e642e5c98db07ba1908053d3 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/orc.py @@ -0,0 +1,384 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + + +from numbers import Integral +import warnings + +from pyarrow.lib import Table +import pyarrow._orc as _orc +from pyarrow.fs import _resolve_filesystem_and_path + + +class ORCFile: + """ + Reader interface for a single ORC file + + Parameters + ---------- + source : str or pyarrow.NativeFile + Readable source. For passing Python file objects or byte buffers, + see pyarrow.io.PythonFileInterface or pyarrow.io.BufferReader. + """ + + def __init__(self, source): + self.reader = _orc.ORCReader() + self.reader.open(source) + + @property + def metadata(self): + """The file metadata, as an arrow KeyValueMetadata""" + return self.reader.metadata() + + @property + def schema(self): + """The file schema, as an arrow schema""" + return self.reader.schema() + + @property + def nrows(self): + """The number of rows in the file""" + return self.reader.nrows() + + @property + def nstripes(self): + """The number of stripes in the file""" + return self.reader.nstripes() + + @property + def file_version(self): + """Format version of the ORC file, must be 0.11 or 0.12""" + return self.reader.file_version() + + @property + def software_version(self): + """Software instance and version that wrote this file""" + return self.reader.software_version() + + @property + def compression(self): + """Compression codec of the file""" + return self.reader.compression() + + @property + def compression_size(self): + """Number of bytes to buffer for the compression codec in the file""" + return self.reader.compression_size() + + @property + def writer(self): + """Name of the writer that wrote this file. + If the writer is unknown then its Writer ID + (a number) is returned""" + return self.reader.writer() + + @property + def writer_version(self): + """Version of the writer""" + return self.reader.writer_version() + + @property + def row_index_stride(self): + """Number of rows per an entry in the row index or 0 + if there is no row index""" + return self.reader.row_index_stride() + + @property + def nstripe_statistics(self): + """Number of stripe statistics""" + return self.reader.nstripe_statistics() + + @property + def content_length(self): + """Length of the data stripes in the file in bytes""" + return self.reader.content_length() + + @property + def stripe_statistics_length(self): + """The number of compressed bytes in the file stripe statistics""" + return self.reader.stripe_statistics_length() + + @property + def file_footer_length(self): + """The number of compressed bytes in the file footer""" + return self.reader.file_footer_length() + + @property + def file_postscript_length(self): + """The number of bytes in the file postscript""" + return self.reader.file_postscript_length() + + @property + def file_length(self): + """The number of bytes in the file""" + return self.reader.file_length() + + def _select_names(self, columns=None): + if columns is None: + return None + + schema = self.schema + names = [] + for col in columns: + if isinstance(col, Integral): + col = int(col) + if 0 <= col < len(schema): + col = schema[col].name + names.append(col) + else: + raise ValueError("Column indices must be in 0 <= ind < %d," + " got %d" % (len(schema), col)) + else: + return columns + + return names + + def read_stripe(self, n, columns=None): + """Read a single stripe from the file. + + Parameters + ---------- + n : int + The stripe index + columns : list + If not None, only these columns will be read from the stripe. A + column name may be a prefix of a nested field, e.g. 'a' will select + 'a.b', 'a.c', and 'a.d.e' + + Returns + ------- + pyarrow.RecordBatch + Content of the stripe as a RecordBatch. + """ + columns = self._select_names(columns) + return self.reader.read_stripe(n, columns=columns) + + def read(self, columns=None): + """Read the whole file. + + Parameters + ---------- + columns : list + If not None, only these columns will be read from the file. A + column name may be a prefix of a nested field, e.g. 'a' will select + 'a.b', 'a.c', and 'a.d.e'. Output always follows the + ordering of the file and not the `columns` list. + + Returns + ------- + pyarrow.Table + Content of the file as a Table. + """ + columns = self._select_names(columns) + return self.reader.read(columns=columns) + + +_orc_writer_args_docs = """file_version : {"0.11", "0.12"}, default "0.12" + Determine which ORC file version to use. + `Hive 0.11 / ORC v0 `_ + is the older version + while `Hive 0.12 / ORC v1 `_ + is the newer one. +batch_size : int, default 1024 + Number of rows the ORC writer writes at a time. +stripe_size : int, default 64 * 1024 * 1024 + Size of each ORC stripe in bytes. +compression : string, default 'uncompressed' + The compression codec. + Valid values: {'UNCOMPRESSED', 'SNAPPY', 'ZLIB', 'LZ4', 'ZSTD'} + Note that LZ0 is currently not supported. +compression_block_size : int, default 64 * 1024 + Size of each compression block in bytes. +compression_strategy : string, default 'speed' + The compression strategy i.e. speed vs size reduction. + Valid values: {'SPEED', 'COMPRESSION'} +row_index_stride : int, default 10000 + The row index stride i.e. the number of rows per + an entry in the row index. +padding_tolerance : double, default 0.0 + The padding tolerance. +dictionary_key_size_threshold : double, default 0.0 + The dictionary key size threshold. 0 to disable dictionary encoding. + 1 to always enable dictionary encoding. +bloom_filter_columns : None, set-like or list-like, default None + Columns that use the bloom filter. +bloom_filter_fpp : double, default 0.05 + Upper limit of the false-positive rate of the bloom filter. +""" + + +class ORCWriter: + __doc__ = """ +Writer interface for a single ORC file + +Parameters +---------- +where : str or pyarrow.io.NativeFile + Writable target. For passing Python file objects or byte buffers, + see pyarrow.io.PythonFileInterface, pyarrow.io.BufferOutputStream + or pyarrow.io.FixedSizeBufferWriter. +{} +""".format(_orc_writer_args_docs) + + is_open = False + + def __init__(self, where, *, + file_version='0.12', + batch_size=1024, + stripe_size=64 * 1024 * 1024, + compression='uncompressed', + compression_block_size=65536, + compression_strategy='speed', + row_index_stride=10000, + padding_tolerance=0.0, + dictionary_key_size_threshold=0.0, + bloom_filter_columns=None, + bloom_filter_fpp=0.05, + ): + self.writer = _orc.ORCWriter() + self.writer.open( + where, + file_version=file_version, + batch_size=batch_size, + stripe_size=stripe_size, + compression=compression, + compression_block_size=compression_block_size, + compression_strategy=compression_strategy, + row_index_stride=row_index_stride, + padding_tolerance=padding_tolerance, + dictionary_key_size_threshold=dictionary_key_size_threshold, + bloom_filter_columns=bloom_filter_columns, + bloom_filter_fpp=bloom_filter_fpp + ) + self.is_open = True + + def __del__(self): + self.close() + + def __enter__(self): + return self + + def __exit__(self, *args, **kwargs): + self.close() + + def write(self, table): + """ + Write the table into an ORC file. The schema of the table must + be equal to the schema used when opening the ORC file. + + Parameters + ---------- + table : pyarrow.Table + The table to be written into the ORC file + """ + assert self.is_open + self.writer.write(table) + + def close(self): + """ + Close the ORC file + """ + if self.is_open: + self.writer.close() + self.is_open = False + + +def read_table(source, columns=None, filesystem=None): + filesystem, path = _resolve_filesystem_and_path(source, filesystem) + if filesystem is not None: + source = filesystem.open_input_file(path) + + if columns is not None and len(columns) == 0: + result = ORCFile(source).read().select(columns) + else: + result = ORCFile(source).read(columns=columns) + + return result + + +read_table.__doc__ = """ +Read a Table from an ORC file. + +Parameters +---------- +source : str, pyarrow.NativeFile, or file-like object + If a string passed, can be a single file name. For file-like objects, + only read a single file. Use pyarrow.BufferReader to read a file + contained in a bytes or buffer-like object. +columns : list + If not None, only these columns will be read from the file. A column + name may be a prefix of a nested field, e.g. 'a' will select 'a.b', + 'a.c', and 'a.d.e'. Output always follows the ordering of the file and + not the `columns` list. If empty, no columns will be read. Note + that the table will still have the correct num_rows set despite having + no columns. +filesystem : FileSystem, default None + If nothing passed, will be inferred based on path. + Path will try to be found in the local on-disk filesystem otherwise + it will be parsed as an URI to determine the filesystem. +""" + + +def write_table(table, where, *, + file_version='0.12', + batch_size=1024, + stripe_size=64 * 1024 * 1024, + compression='uncompressed', + compression_block_size=65536, + compression_strategy='speed', + row_index_stride=10000, + padding_tolerance=0.0, + dictionary_key_size_threshold=0.0, + bloom_filter_columns=None, + bloom_filter_fpp=0.05): + if isinstance(where, Table): + warnings.warn( + "The order of the arguments has changed. Pass as " + "'write_table(table, where)' instead. The old order will raise " + "an error in the future.", FutureWarning, stacklevel=2 + ) + table, where = where, table + with ORCWriter( + where, + file_version=file_version, + batch_size=batch_size, + stripe_size=stripe_size, + compression=compression, + compression_block_size=compression_block_size, + compression_strategy=compression_strategy, + row_index_stride=row_index_stride, + padding_tolerance=padding_tolerance, + dictionary_key_size_threshold=dictionary_key_size_threshold, + bloom_filter_columns=bloom_filter_columns, + bloom_filter_fpp=bloom_filter_fpp + ) as writer: + writer.write(table) + + +write_table.__doc__ = """ +Write a table into an ORC file. + +Parameters +---------- +table : pyarrow.lib.Table + The table to be written into the ORC file +where : str or pyarrow.io.NativeFile + Writable target. For passing Python file objects or byte buffers, + see pyarrow.io.PythonFileInterface, pyarrow.io.BufferOutputStream + or pyarrow.io.FixedSizeBufferWriter. +{} +""".format(_orc_writer_args_docs) diff --git a/vllm/lib/python3.10/site-packages/pyarrow/pandas-shim.pxi b/vllm/lib/python3.10/site-packages/pyarrow/pandas-shim.pxi new file mode 100644 index 0000000000000000000000000000000000000000..74f0d981b52f446c890b396b66d7b0a6e027db77 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/pandas-shim.pxi @@ -0,0 +1,261 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# pandas lazy-loading API shim that reduces API call and import overhead + +import warnings +from threading import Lock + + +cdef class _PandasAPIShim(object): + """ + Lazy pandas importer that isolates usages of pandas APIs and avoids + importing pandas until it's actually needed + """ + cdef: + bint _tried_importing_pandas + bint _have_pandas + + cdef readonly: + object _loose_version, _version + object _pd, _types_api, _compat_module + object _data_frame, _index, _series, _categorical_type + object _datetimetz_type, _extension_array, _extension_dtype + object _array_like_types, _is_extension_array_dtype, _lock + bint has_sparse + bint _pd024 + bint _is_v1, _is_ge_v21, _is_ge_v3 + + def __init__(self): + self._lock = Lock() + self._tried_importing_pandas = False + self._have_pandas = 0 + + cdef _import_pandas(self, bint raise_): + try: + import pandas as pd + import pyarrow.pandas_compat as pdcompat + except ImportError: + self._have_pandas = False + if raise_: + raise + else: + return + + from pyarrow.vendored.version import Version + + self._pd = pd + self._version = pd.__version__ + self._loose_version = Version(pd.__version__) + self._is_v1 = False + + if self._loose_version < Version('1.0.0'): + self._have_pandas = False + if raise_: + raise ImportError( + "pyarrow requires pandas 1.0.0 or above, pandas {} is " + "installed".format(self._version) + ) + else: + warnings.warn( + "pyarrow requires pandas 1.0.0 or above, pandas {} is " + "installed. Therefore, pandas-specific integration is not " + "used.".format(self._version), stacklevel=2) + return + + self._is_v1 = self._loose_version < Version('2.0.0') + self._is_ge_v21 = self._loose_version >= Version('2.1.0') + self._is_ge_v3 = self._loose_version >= Version('3.0.0.dev0') + + self._compat_module = pdcompat + self._data_frame = pd.DataFrame + self._index = pd.Index + self._categorical_type = pd.Categorical + self._series = pd.Series + self._extension_array = pd.api.extensions.ExtensionArray + self._array_like_types = ( + self._series, self._index, self._categorical_type, + self._extension_array) + self._extension_dtype = pd.api.extensions.ExtensionDtype + self._is_extension_array_dtype = ( + pd.api.types.is_extension_array_dtype) + self._types_api = pd.api.types + self._datetimetz_type = pd.api.types.DatetimeTZDtype + self._have_pandas = True + self.has_sparse = False + + cdef inline _check_import(self, bint raise_=True): + if not self._tried_importing_pandas: + with self._lock: + if not self._tried_importing_pandas: + try: + self._import_pandas(raise_) + finally: + self._tried_importing_pandas = True + return + + if not self._have_pandas and raise_: + self._import_pandas(raise_) + + def series(self, *args, **kwargs): + self._check_import() + return self._series(*args, **kwargs) + + def data_frame(self, *args, **kwargs): + self._check_import() + return self._data_frame(*args, **kwargs) + + cdef inline bint _have_pandas_internal(self): + if not self._tried_importing_pandas: + self._check_import(raise_=False) + return self._have_pandas + + @property + def have_pandas(self): + return self._have_pandas_internal() + + @property + def compat(self): + self._check_import() + return self._compat_module + + @property + def pd(self): + self._check_import() + return self._pd + + cpdef infer_dtype(self, obj): + self._check_import() + try: + return self._types_api.infer_dtype(obj, skipna=False) + except AttributeError: + return self._pd.lib.infer_dtype(obj) + + cpdef pandas_dtype(self, dtype): + self._check_import() + try: + return self._types_api.pandas_dtype(dtype) + except AttributeError: + return None + + @property + def loose_version(self): + self._check_import() + return self._loose_version + + @property + def version(self): + self._check_import() + return self._version + + def is_v1(self): + self._check_import() + return self._is_v1 + + def is_ge_v21(self): + self._check_import() + return self._is_ge_v21 + + def is_ge_v3(self): + self._check_import() + return self._is_ge_v3 + + @property + def categorical_type(self): + self._check_import() + return self._categorical_type + + @property + def datetimetz_type(self): + self._check_import() + return self._datetimetz_type + + @property + def extension_dtype(self): + self._check_import() + return self._extension_dtype + + cpdef is_array_like(self, obj): + self._check_import() + return isinstance(obj, self._array_like_types) + + cpdef is_categorical(self, obj): + if self._have_pandas_internal(): + return isinstance(obj, self._categorical_type) + else: + return False + + cpdef is_datetimetz(self, obj): + if self._have_pandas_internal(): + return isinstance(obj, self._datetimetz_type) + else: + return False + + cpdef is_extension_array_dtype(self, obj): + self._check_import() + if self._is_extension_array_dtype: + return self._is_extension_array_dtype(obj) + else: + return False + + cpdef is_sparse(self, obj): + if self._have_pandas_internal(): + return isinstance(obj.dtype, self.pd.SparseDtype) + else: + return False + + cpdef is_data_frame(self, obj): + if self._have_pandas_internal(): + return isinstance(obj, self._data_frame) + else: + return False + + cpdef is_series(self, obj): + if self._have_pandas_internal(): + return isinstance(obj, self._series) + else: + return False + + cpdef is_index(self, obj): + if self._have_pandas_internal(): + return isinstance(obj, self._index) + else: + return False + + cpdef get_values(self, obj): + """ + Get the underlying array values of a pandas Series or Index in the + format (np.ndarray or pandas ExtensionArray) as we need them. + + Assumes obj is a pandas Series or Index. + """ + self._check_import() + if isinstance(obj.dtype, (self.pd.api.types.IntervalDtype, + self.pd.api.types.PeriodDtype)): + return obj.array + return obj.values + + def get_rangeindex_attribute(self, level, name): + # public start/stop/step attributes added in pandas 0.25.0 + self._check_import() + if hasattr(level, name): + return getattr(level, name) + return getattr(level, '_' + name) + + +cdef _PandasAPIShim pandas_api = _PandasAPIShim() +_pandas_api = pandas_api diff --git a/vllm/lib/python3.10/site-packages/pyarrow/util.py b/vllm/lib/python3.10/site-packages/pyarrow/util.py new file mode 100644 index 0000000000000000000000000000000000000000..89780da10f7cfe682ce0b12a1a5bb99975e846e9 --- /dev/null +++ b/vllm/lib/python3.10/site-packages/pyarrow/util.py @@ -0,0 +1,258 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +# Miscellaneous utility code + +import os +import contextlib +import functools +import gc +import socket +import sys +import textwrap +import types +import warnings + + +_DEPR_MSG = ( + "pyarrow.{} is deprecated as of {}, please use pyarrow.{} instead." +) + + +def doc(*docstrings, **params): + """ + A decorator that takes docstring templates, concatenates them, and finally + performs string substitution on them. + This decorator will add a variable "_docstring_components" to the wrapped + callable to keep track of the original docstring template for potential future use. + If the docstring is a template, it will be saved as a string. + Otherwise, it will be saved as a callable and the docstring will be obtained via + the __doc__ attribute. + This decorator cannot be used on Cython classes due to a CPython constraint, + which enforces the __doc__ attribute to be read-only. + See https://github.com/python/cpython/issues/91309 + + Parameters + ---------- + *docstrings : None, str, or callable + The string / docstring / docstring template to be prepended in order + before the default docstring under the callable. + **params + The key/value pairs used to format the docstring template. + """ + + def decorator(decorated): + docstring_components = [] + + # collect docstrings and docstring templates + for docstring in docstrings: + if docstring is None: + continue + if hasattr(docstring, "_docstring_components"): + docstring_components.extend( + docstring._docstring_components + ) + elif isinstance(docstring, str) or docstring.__doc__: + docstring_components.append(docstring) + + # append the callable's docstring last + if decorated.__doc__: + docstring_components.append(textwrap.dedent(decorated.__doc__)) + + params_applied = [ + component.format(**params) + if isinstance(component, str) and len(params) > 0 + else component + for component in docstring_components + ] + + decorated.__doc__ = "".join( + [ + component + if isinstance(component, str) + else textwrap.dedent(component.__doc__ or "") + for component in params_applied + ] + ) + + decorated._docstring_components = ( + docstring_components + ) + return decorated + + return decorator + + +def _deprecate_api(old_name, new_name, api, next_version, type=FutureWarning): + msg = _DEPR_MSG.format(old_name, next_version, new_name) + + def wrapper(*args, **kwargs): + warnings.warn(msg, type) + return api(*args, **kwargs) + return wrapper + + +def _deprecate_class(old_name, new_class, next_version, + instancecheck=True): + """ + Raise warning if a deprecated class is used in an isinstance check. + """ + class _DeprecatedMeta(type): + def __instancecheck__(self, other): + warnings.warn( + _DEPR_MSG.format(old_name, next_version, new_class.__name__), + FutureWarning, + stacklevel=2 + ) + return isinstance(other, new_class) + + return _DeprecatedMeta(old_name, (new_class,), {}) + + +def _is_iterable(obj): + try: + iter(obj) + return True + except TypeError: + return False + + +def _is_path_like(path): + return isinstance(path, str) or hasattr(path, '__fspath__') + + +def _stringify_path(path): + """ + Convert *path* to a string or unicode path if possible. + """ + if isinstance(path, str): + return os.path.expanduser(path) + + # checking whether path implements the filesystem protocol + try: + return os.path.expanduser(path.__fspath__()) + except AttributeError: + pass + + raise TypeError("not a path-like object") + + +def product(seq): + """ + Return a product of sequence items. + """ + return functools.reduce(lambda a, b: a*b, seq, 1) + + +def get_contiguous_span(shape, strides, itemsize): + """ + Return a contiguous span of N-D array data. + + Parameters + ---------- + shape : tuple + strides : tuple + itemsize : int + Specify array shape data + + Returns + ------- + start, end : int + The span end points. + """ + if not strides: + start = 0 + end = itemsize * product(shape) + else: + start = 0 + end = itemsize + for i, dim in enumerate(shape): + if dim == 0: + start = end = 0 + break + stride = strides[i] + if stride > 0: + end += stride * (dim - 1) + elif stride < 0: + start += stride * (dim - 1) + if end - start != itemsize * product(shape): + raise ValueError('array data is non-contiguous') + return start, end + + +def find_free_port(): + sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM) + with contextlib.closing(sock) as sock: + sock.bind(('', 0)) + sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1) + return sock.getsockname()[1] + + +def guid(): + from uuid import uuid4 + return uuid4().hex + + +def _break_traceback_cycle_from_frame(frame): + # Clear local variables in all inner frames, so as to break the + # reference cycle. + this_frame = sys._getframe(0) + refs = gc.get_referrers(frame) + while refs: + for frame in refs: + if frame is not this_frame and isinstance(frame, types.FrameType): + break + else: + # No frame found in referrers (finished?) + break + refs = None + # Clear the frame locals, to try and break the cycle (it is + # somewhere along the chain of execution frames). + frame.clear() + # To visit the inner frame, we need to find it among the + # referrers of this frame (while `frame.f_back` would let + # us visit the outer frame). + refs = gc.get_referrers(frame) + refs = frame = this_frame = None + + +def download_tzdata_on_windows(): + r""" + Download and extract latest IANA timezone database into the + location expected by Arrow which is %USERPROFILE%\Downloads\tzdata. + """ + if sys.platform != 'win32': + raise TypeError(f"Timezone database is already provided by {sys.platform}") + + import tarfile + + tzdata_path = os.path.expandvars(r"%USERPROFILE%\Downloads\tzdata") + tzdata_compressed = os.path.join(tzdata_path, "tzdata.tar.gz") + os.makedirs(tzdata_path, exist_ok=True) + + from urllib.request import urlopen + with urlopen('https://data.iana.org/time-zones/tzdata-latest.tar.gz') as response: + with open(tzdata_compressed, 'wb') as f: + f.write(response.read()) + + assert os.path.exists(tzdata_compressed) + + tarfile.open(tzdata_compressed).extractall(tzdata_path) + + with urlopen('https://raw.githubusercontent.com/unicode-org/cldr/master/common/supplemental/windowsZones.xml') as response_zones: # noqa + with open(os.path.join(tzdata_path, "windowsZones.xml"), 'wb') as f: + f.write(response_zones.read())