Add files using upload-large-folder tool
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- .gitattributes +1 -0
- videollama2/lib/python3.10/site-packages/sklearn/__pycache__/naive_bayes.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/__init__.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/glm_distribution.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/link.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/loss.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/_loss/_loss.cpython-310-x86_64-linux-gnu.so +3 -0
- videollama2/lib/python3.10/site-packages/sklearn/_loss/tests/__pycache__/__init__.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/_loss/tests/__pycache__/test_glm_distribution.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/externals/__init__.py +5 -0
- videollama2/lib/python3.10/site-packages/sklearn/externals/_arff.py +1107 -0
- videollama2/lib/python3.10/site-packages/sklearn/impute/_knn.py +391 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/__init__.py +11 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/__pycache__/__init__.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/__pycache__/_label_propagation.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/_label_propagation.py +621 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/_self_training.py +400 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__init__.py +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__pycache__/__init__.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__pycache__/test_label_propagation.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__pycache__/test_self_training.cpython-310.pyc +0 -0
- videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/test_label_propagation.py +216 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_compute_linear_combination_native.h +22 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_cudnn_rnn_backward_native.h +22 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_empty_per_channel_affine_quantized_ops.h +39 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_floor_compositeexplicitautograd_dispatch.h +26 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_lerp_native.h +30 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_reciprocal_cuda_dispatch.h +24 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_grid_sampler_2d_cpu_fallback.h +39 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_lstm_mps_ops.h +39 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_make_per_channel_quantized_tensor_cpu_dispatch.h +23 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_compute_contiguous_strides_offsets_native.h +21 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_get_values_copy_compositeexplicitautogradnonfunctional_dispatch.h +23 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_tensor_from_mask_native.h +22 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_tensor_strides_compositeexplicitautograd_dispatch.h +24 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_prelu_kernel_ops.h +28 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_scaled_dot_product_flash_attention_backward.h +47 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_test_warn_in_autograd_native.h +22 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_to_dense_native.h +24 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_bilinear2d_aa_compositeexplicitautogradnonfunctional_dispatch.h +24 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact1d_meta_dispatch.h +28 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/addcmul_cuda_dispatch.h +26 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/and_native.h +24 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/as_strided_copy_ops.h +39 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/as_strided_native.h +24 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/avg_pool2d_ops.h +39 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/bitwise_or_cpu_dispatch.h +26 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/bucketize_ops.h +61 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/copy_meta_dispatch.h +23 -0
- vllm/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_convolution_transpose_compositeexplicitautograd_dispatch.h +26 -0
.gitattributes
CHANGED
|
@@ -1940,3 +1940,4 @@ videollama2/lib/python3.10/site-packages/sklearn/metrics/_pairwise_distances_red
|
|
| 1940 |
videollama2/lib/python3.10/site-packages/sklearn/metrics/_pairwise_fast.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
|
| 1941 |
videollama2/lib/python3.10/site-packages/sklearn/metrics/_dist_metrics.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
|
| 1942 |
videollama2/lib/python3.10/site-packages/sklearn/metrics/cluster/_expected_mutual_info_fast.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 1940 |
videollama2/lib/python3.10/site-packages/sklearn/metrics/_pairwise_fast.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
|
| 1941 |
videollama2/lib/python3.10/site-packages/sklearn/metrics/_dist_metrics.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
|
| 1942 |
videollama2/lib/python3.10/site-packages/sklearn/metrics/cluster/_expected_mutual_info_fast.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
|
| 1943 |
+
videollama2/lib/python3.10/site-packages/sklearn/_loss/_loss.cpython-310-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
|
videollama2/lib/python3.10/site-packages/sklearn/__pycache__/naive_bayes.cpython-310.pyc
ADDED
|
Binary file (50.3 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/__init__.cpython-310.pyc
ADDED
|
Binary file (586 Bytes). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/glm_distribution.cpython-310.pyc
ADDED
|
Binary file (10.5 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/link.cpython-310.pyc
ADDED
|
Binary file (8.45 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/_loss/__pycache__/loss.cpython-310.pyc
ADDED
|
Binary file (29.7 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/_loss/_loss.cpython-310-x86_64-linux-gnu.so
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:616237559467732cc220de2a62e942fb993125159cecf1a37a0289203e98697e
|
| 3 |
+
size 2109161
|
videollama2/lib/python3.10/site-packages/sklearn/_loss/tests/__pycache__/__init__.cpython-310.pyc
ADDED
|
Binary file (176 Bytes). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/_loss/tests/__pycache__/test_glm_distribution.cpython-310.pyc
ADDED
|
Binary file (3.49 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/externals/__init__.py
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
"""
|
| 3 |
+
External, bundled dependencies.
|
| 4 |
+
|
| 5 |
+
"""
|
videollama2/lib/python3.10/site-packages/sklearn/externals/_arff.py
ADDED
|
@@ -0,0 +1,1107 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# =============================================================================
|
| 2 |
+
# Federal University of Rio Grande do Sul (UFRGS)
|
| 3 |
+
# Connectionist Artificial Intelligence Laboratory (LIAC)
|
| 4 |
+
# Renato de Pontes Pereira - rppereira@inf.ufrgs.br
|
| 5 |
+
# =============================================================================
|
| 6 |
+
# Copyright (c) 2011 Renato de Pontes Pereira, renato.ppontes at gmail dot com
|
| 7 |
+
#
|
| 8 |
+
# Permission is hereby granted, free of charge, to any person obtaining a copy
|
| 9 |
+
# of this software and associated documentation files (the "Software"), to deal
|
| 10 |
+
# in the Software without restriction, including without limitation the rights
|
| 11 |
+
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
| 12 |
+
# copies of the Software, and to permit persons to whom the Software is
|
| 13 |
+
# furnished to do so, subject to the following conditions:
|
| 14 |
+
#
|
| 15 |
+
# The above copyright notice and this permission notice shall be included in
|
| 16 |
+
# all copies or substantial portions of the Software.
|
| 17 |
+
#
|
| 18 |
+
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
| 19 |
+
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
| 20 |
+
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
| 21 |
+
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
| 22 |
+
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
| 23 |
+
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
| 24 |
+
# SOFTWARE.
|
| 25 |
+
# =============================================================================
|
| 26 |
+
|
| 27 |
+
'''
|
| 28 |
+
The liac-arff module implements functions to read and write ARFF files in
|
| 29 |
+
Python. It was created in the Connectionist Artificial Intelligence Laboratory
|
| 30 |
+
(LIAC), which takes place at the Federal University of Rio Grande do Sul
|
| 31 |
+
(UFRGS), in Brazil.
|
| 32 |
+
|
| 33 |
+
ARFF (Attribute-Relation File Format) is an file format specially created for
|
| 34 |
+
describe datasets which are commonly used for machine learning experiments and
|
| 35 |
+
software. This file format was created to be used in Weka, the best
|
| 36 |
+
representative software for machine learning automated experiments.
|
| 37 |
+
|
| 38 |
+
An ARFF file can be divided into two sections: header and data. The Header
|
| 39 |
+
describes the metadata of the dataset, including a general description of the
|
| 40 |
+
dataset, its name and its attributes. The source below is an example of a
|
| 41 |
+
header section in a XOR dataset::
|
| 42 |
+
|
| 43 |
+
%
|
| 44 |
+
% XOR Dataset
|
| 45 |
+
%
|
| 46 |
+
% Created by Renato Pereira
|
| 47 |
+
% rppereira@inf.ufrgs.br
|
| 48 |
+
% http://inf.ufrgs.br/~rppereira
|
| 49 |
+
%
|
| 50 |
+
%
|
| 51 |
+
@RELATION XOR
|
| 52 |
+
|
| 53 |
+
@ATTRIBUTE input1 REAL
|
| 54 |
+
@ATTRIBUTE input2 REAL
|
| 55 |
+
@ATTRIBUTE y REAL
|
| 56 |
+
|
| 57 |
+
The Data section of an ARFF file describes the observations of the dataset, in
|
| 58 |
+
the case of XOR dataset::
|
| 59 |
+
|
| 60 |
+
@DATA
|
| 61 |
+
0.0,0.0,0.0
|
| 62 |
+
0.0,1.0,1.0
|
| 63 |
+
1.0,0.0,1.0
|
| 64 |
+
1.0,1.0,0.0
|
| 65 |
+
%
|
| 66 |
+
%
|
| 67 |
+
%
|
| 68 |
+
|
| 69 |
+
Notice that several lines are starting with an ``%`` symbol, denoting a
|
| 70 |
+
comment, thus, lines with ``%`` at the beginning will be ignored, except by the
|
| 71 |
+
description part at the beginning of the file. The declarations ``@RELATION``,
|
| 72 |
+
``@ATTRIBUTE``, and ``@DATA`` are all case insensitive and obligatory.
|
| 73 |
+
|
| 74 |
+
For more information and details about the ARFF file description, consult
|
| 75 |
+
http://www.cs.waikato.ac.nz/~ml/weka/arff.html
|
| 76 |
+
|
| 77 |
+
|
| 78 |
+
ARFF Files in Python
|
| 79 |
+
~~~~~~~~~~~~~~~~~~~~
|
| 80 |
+
|
| 81 |
+
This module uses built-ins python objects to represent a deserialized ARFF
|
| 82 |
+
file. A dictionary is used as the container of the data and metadata of ARFF,
|
| 83 |
+
and have the following keys:
|
| 84 |
+
|
| 85 |
+
- **description**: (OPTIONAL) a string with the description of the dataset.
|
| 86 |
+
- **relation**: (OBLIGATORY) a string with the name of the dataset.
|
| 87 |
+
- **attributes**: (OBLIGATORY) a list of attributes with the following
|
| 88 |
+
template::
|
| 89 |
+
|
| 90 |
+
(attribute_name, attribute_type)
|
| 91 |
+
|
| 92 |
+
the attribute_name is a string, and attribute_type must be an string
|
| 93 |
+
or a list of strings.
|
| 94 |
+
- **data**: (OBLIGATORY) a list of data instances. Each data instance must be
|
| 95 |
+
a list with values, depending on the attributes.
|
| 96 |
+
|
| 97 |
+
The above keys must follow the case which were described, i.e., the keys are
|
| 98 |
+
case sensitive. The attribute type ``attribute_type`` must be one of these
|
| 99 |
+
strings (they are not case sensitive): ``NUMERIC``, ``INTEGER``, ``REAL`` or
|
| 100 |
+
``STRING``. For nominal attributes, the ``atribute_type`` must be a list of
|
| 101 |
+
strings.
|
| 102 |
+
|
| 103 |
+
In this format, the XOR dataset presented above can be represented as a python
|
| 104 |
+
object as::
|
| 105 |
+
|
| 106 |
+
xor_dataset = {
|
| 107 |
+
'description': 'XOR Dataset',
|
| 108 |
+
'relation': 'XOR',
|
| 109 |
+
'attributes': [
|
| 110 |
+
('input1', 'REAL'),
|
| 111 |
+
('input2', 'REAL'),
|
| 112 |
+
('y', 'REAL'),
|
| 113 |
+
],
|
| 114 |
+
'data': [
|
| 115 |
+
[0.0, 0.0, 0.0],
|
| 116 |
+
[0.0, 1.0, 1.0],
|
| 117 |
+
[1.0, 0.0, 1.0],
|
| 118 |
+
[1.0, 1.0, 0.0]
|
| 119 |
+
]
|
| 120 |
+
}
|
| 121 |
+
|
| 122 |
+
|
| 123 |
+
Features
|
| 124 |
+
~~~~~~~~
|
| 125 |
+
|
| 126 |
+
This module provides several features, including:
|
| 127 |
+
|
| 128 |
+
- Read and write ARFF files using python built-in structures, such dictionaries
|
| 129 |
+
and lists;
|
| 130 |
+
- Supports `scipy.sparse.coo <http://docs.scipy
|
| 131 |
+
.org/doc/scipy/reference/generated/scipy.sparse.coo_matrix.html#scipy.sparse.coo_matrix>`_
|
| 132 |
+
and lists of dictionaries as used by SVMLight
|
| 133 |
+
- Supports the following attribute types: NUMERIC, REAL, INTEGER, STRING, and
|
| 134 |
+
NOMINAL;
|
| 135 |
+
- Has an interface similar to other built-in modules such as ``json``, or
|
| 136 |
+
``zipfile``;
|
| 137 |
+
- Supports read and write the descriptions of files;
|
| 138 |
+
- Supports missing values and names with spaces;
|
| 139 |
+
- Supports unicode values and names;
|
| 140 |
+
- Fully compatible with Python 2.7+, Python 3.5+, pypy and pypy3;
|
| 141 |
+
- Under `MIT License <http://opensource.org/licenses/MIT>`_
|
| 142 |
+
|
| 143 |
+
'''
|
| 144 |
+
__author__ = 'Renato de Pontes Pereira, Matthias Feurer, Joel Nothman'
|
| 145 |
+
__author_email__ = ('renato.ppontes@gmail.com, '
|
| 146 |
+
'feurerm@informatik.uni-freiburg.de, '
|
| 147 |
+
'joel.nothman@gmail.com')
|
| 148 |
+
__version__ = '2.4.0'
|
| 149 |
+
|
| 150 |
+
import re
|
| 151 |
+
import csv
|
| 152 |
+
from typing import TYPE_CHECKING
|
| 153 |
+
from typing import Optional, List, Dict, Any, Iterator, Union, Tuple
|
| 154 |
+
|
| 155 |
+
# CONSTANTS ===================================================================
|
| 156 |
+
_SIMPLE_TYPES = ['NUMERIC', 'REAL', 'INTEGER', 'STRING']
|
| 157 |
+
|
| 158 |
+
_TK_DESCRIPTION = '%'
|
| 159 |
+
_TK_COMMENT = '%'
|
| 160 |
+
_TK_RELATION = '@RELATION'
|
| 161 |
+
_TK_ATTRIBUTE = '@ATTRIBUTE'
|
| 162 |
+
_TK_DATA = '@DATA'
|
| 163 |
+
|
| 164 |
+
_RE_RELATION = re.compile(r'^([^\{\}%,\s]*|\".*\"|\'.*\')$', re.UNICODE)
|
| 165 |
+
_RE_ATTRIBUTE = re.compile(r'^(\".*\"|\'.*\'|[^\{\}%,\s]*)\s+(.+)$', re.UNICODE)
|
| 166 |
+
_RE_QUOTE_CHARS = re.compile(r'["\'\\\s%,\000-\031]', re.UNICODE)
|
| 167 |
+
_RE_ESCAPE_CHARS = re.compile(r'(?=["\'\\%])|[\n\r\t\000-\031]')
|
| 168 |
+
_RE_SPARSE_LINE = re.compile(r'^\s*\{.*\}\s*$', re.UNICODE)
|
| 169 |
+
_RE_NONTRIVIAL_DATA = re.compile('["\'{}\\s]', re.UNICODE)
|
| 170 |
+
|
| 171 |
+
ArffDenseDataType = Iterator[List]
|
| 172 |
+
ArffSparseDataType = Tuple[List, ...]
|
| 173 |
+
|
| 174 |
+
|
| 175 |
+
if TYPE_CHECKING:
|
| 176 |
+
# typing_extensions is available when mypy is installed
|
| 177 |
+
from typing_extensions import TypedDict
|
| 178 |
+
|
| 179 |
+
class ArffContainerType(TypedDict):
|
| 180 |
+
description: str
|
| 181 |
+
relation: str
|
| 182 |
+
attributes: List
|
| 183 |
+
data: Union[ArffDenseDataType, ArffSparseDataType]
|
| 184 |
+
|
| 185 |
+
else:
|
| 186 |
+
ArffContainerType = Dict[str, Any]
|
| 187 |
+
|
| 188 |
+
|
| 189 |
+
def _build_re_values():
|
| 190 |
+
quoted_re = r'''
|
| 191 |
+
" # open quote followed by zero or more of:
|
| 192 |
+
(?:
|
| 193 |
+
(?<!\\) # no additional backslash
|
| 194 |
+
(?:\\\\)* # maybe escaped backslashes
|
| 195 |
+
\\" # escaped quote
|
| 196 |
+
|
|
| 197 |
+
\\[^"] # escaping a non-quote
|
| 198 |
+
|
|
| 199 |
+
[^"\\] # non-quote char
|
| 200 |
+
)*
|
| 201 |
+
" # close quote
|
| 202 |
+
'''
|
| 203 |
+
# a value is surrounded by " or by ' or contains no quotables
|
| 204 |
+
value_re = r'''(?:
|
| 205 |
+
%s| # a value may be surrounded by "
|
| 206 |
+
%s| # or by '
|
| 207 |
+
[^,\s"'{}]+ # or may contain no characters requiring quoting
|
| 208 |
+
)''' % (quoted_re,
|
| 209 |
+
quoted_re.replace('"', "'"))
|
| 210 |
+
|
| 211 |
+
# This captures (value, error) groups. Because empty values are allowed,
|
| 212 |
+
# we cannot just look for empty values to handle syntax errors.
|
| 213 |
+
# We presume the line has had ',' prepended...
|
| 214 |
+
dense = re.compile(r'''(?x)
|
| 215 |
+
, # may follow ','
|
| 216 |
+
\s*
|
| 217 |
+
((?=,)|$|{value_re}) # empty or value
|
| 218 |
+
|
|
| 219 |
+
(\S.*) # error
|
| 220 |
+
'''.format(value_re=value_re))
|
| 221 |
+
|
| 222 |
+
# This captures (key, value) groups and will have an empty key/value
|
| 223 |
+
# in case of syntax errors.
|
| 224 |
+
# It does not ensure that the line starts with '{' or ends with '}'.
|
| 225 |
+
sparse = re.compile(r'''(?x)
|
| 226 |
+
(?:^\s*\{|,) # may follow ',', or '{' at line start
|
| 227 |
+
\s*
|
| 228 |
+
(\d+) # attribute key
|
| 229 |
+
\s+
|
| 230 |
+
(%(value_re)s) # value
|
| 231 |
+
|
|
| 232 |
+
(?!}\s*$) # not an error if it's }$
|
| 233 |
+
(?!^\s*{\s*}\s*$) # not an error if it's ^{}$
|
| 234 |
+
\S.* # error
|
| 235 |
+
''' % {'value_re': value_re})
|
| 236 |
+
return dense, sparse
|
| 237 |
+
|
| 238 |
+
|
| 239 |
+
|
| 240 |
+
_RE_DENSE_VALUES, _RE_SPARSE_KEY_VALUES = _build_re_values()
|
| 241 |
+
|
| 242 |
+
|
| 243 |
+
_ESCAPE_SUB_MAP = {
|
| 244 |
+
'\\\\': '\\',
|
| 245 |
+
'\\"': '"',
|
| 246 |
+
"\\'": "'",
|
| 247 |
+
'\\t': '\t',
|
| 248 |
+
'\\n': '\n',
|
| 249 |
+
'\\r': '\r',
|
| 250 |
+
'\\b': '\b',
|
| 251 |
+
'\\f': '\f',
|
| 252 |
+
'\\%': '%',
|
| 253 |
+
}
|
| 254 |
+
_UNESCAPE_SUB_MAP = {chr(i): '\\%03o' % i for i in range(32)}
|
| 255 |
+
_UNESCAPE_SUB_MAP.update({v: k for k, v in _ESCAPE_SUB_MAP.items()})
|
| 256 |
+
_UNESCAPE_SUB_MAP[''] = '\\'
|
| 257 |
+
_ESCAPE_SUB_MAP.update({'\\%d' % i: chr(i) for i in range(10)})
|
| 258 |
+
|
| 259 |
+
|
| 260 |
+
def _escape_sub_callback(match):
|
| 261 |
+
s = match.group()
|
| 262 |
+
if len(s) == 2:
|
| 263 |
+
try:
|
| 264 |
+
return _ESCAPE_SUB_MAP[s]
|
| 265 |
+
except KeyError:
|
| 266 |
+
raise ValueError('Unsupported escape sequence: %s' % s)
|
| 267 |
+
if s[1] == 'u':
|
| 268 |
+
return chr(int(s[2:], 16))
|
| 269 |
+
else:
|
| 270 |
+
return chr(int(s[1:], 8))
|
| 271 |
+
|
| 272 |
+
|
| 273 |
+
def _unquote(v):
|
| 274 |
+
if v[:1] in ('"', "'"):
|
| 275 |
+
return re.sub(r'\\([0-9]{1,3}|u[0-9a-f]{4}|.)', _escape_sub_callback,
|
| 276 |
+
v[1:-1])
|
| 277 |
+
elif v in ('?', ''):
|
| 278 |
+
return None
|
| 279 |
+
else:
|
| 280 |
+
return v
|
| 281 |
+
|
| 282 |
+
|
| 283 |
+
def _parse_values(s):
|
| 284 |
+
'''(INTERNAL) Split a line into a list of values'''
|
| 285 |
+
if not _RE_NONTRIVIAL_DATA.search(s):
|
| 286 |
+
# Fast path for trivial cases (unfortunately we have to handle missing
|
| 287 |
+
# values because of the empty string case :(.)
|
| 288 |
+
return [None if s in ('?', '') else s
|
| 289 |
+
for s in next(csv.reader([s]))]
|
| 290 |
+
|
| 291 |
+
# _RE_DENSE_VALUES tokenizes despite quoting, whitespace, etc.
|
| 292 |
+
values, errors = zip(*_RE_DENSE_VALUES.findall(',' + s))
|
| 293 |
+
if not any(errors):
|
| 294 |
+
return [_unquote(v) for v in values]
|
| 295 |
+
if _RE_SPARSE_LINE.match(s):
|
| 296 |
+
try:
|
| 297 |
+
return {int(k): _unquote(v)
|
| 298 |
+
for k, v in _RE_SPARSE_KEY_VALUES.findall(s)}
|
| 299 |
+
except ValueError:
|
| 300 |
+
# an ARFF syntax error in sparse data
|
| 301 |
+
for match in _RE_SPARSE_KEY_VALUES.finditer(s):
|
| 302 |
+
if not match.group(1):
|
| 303 |
+
raise BadLayout('Error parsing %r' % match.group())
|
| 304 |
+
raise BadLayout('Unknown parsing error')
|
| 305 |
+
else:
|
| 306 |
+
# an ARFF syntax error
|
| 307 |
+
for match in _RE_DENSE_VALUES.finditer(s):
|
| 308 |
+
if match.group(2):
|
| 309 |
+
raise BadLayout('Error parsing %r' % match.group())
|
| 310 |
+
raise BadLayout('Unknown parsing error')
|
| 311 |
+
|
| 312 |
+
|
| 313 |
+
DENSE = 0 # Constant value representing a dense matrix
|
| 314 |
+
COO = 1 # Constant value representing a sparse matrix in coordinate format
|
| 315 |
+
LOD = 2 # Constant value representing a sparse matrix in list of
|
| 316 |
+
# dictionaries format
|
| 317 |
+
DENSE_GEN = 3 # Generator of dictionaries
|
| 318 |
+
LOD_GEN = 4 # Generator of dictionaries
|
| 319 |
+
_SUPPORTED_DATA_STRUCTURES = [DENSE, COO, LOD, DENSE_GEN, LOD_GEN]
|
| 320 |
+
|
| 321 |
+
|
| 322 |
+
# EXCEPTIONS ==================================================================
|
| 323 |
+
class ArffException(Exception):
|
| 324 |
+
message: Optional[str] = None
|
| 325 |
+
|
| 326 |
+
def __init__(self):
|
| 327 |
+
self.line = -1
|
| 328 |
+
|
| 329 |
+
def __str__(self):
|
| 330 |
+
return self.message%self.line
|
| 331 |
+
|
| 332 |
+
class BadRelationFormat(ArffException):
|
| 333 |
+
'''Error raised when the relation declaration is in an invalid format.'''
|
| 334 |
+
message = 'Bad @RELATION format, at line %d.'
|
| 335 |
+
|
| 336 |
+
class BadAttributeFormat(ArffException):
|
| 337 |
+
'''Error raised when some attribute declaration is in an invalid format.'''
|
| 338 |
+
message = 'Bad @ATTRIBUTE format, at line %d.'
|
| 339 |
+
|
| 340 |
+
class BadDataFormat(ArffException):
|
| 341 |
+
'''Error raised when some data instance is in an invalid format.'''
|
| 342 |
+
def __init__(self, value):
|
| 343 |
+
super().__init__()
|
| 344 |
+
self.message = (
|
| 345 |
+
'Bad @DATA instance format in line %d: ' +
|
| 346 |
+
('%s' % value)
|
| 347 |
+
)
|
| 348 |
+
|
| 349 |
+
class BadAttributeType(ArffException):
|
| 350 |
+
'''Error raised when some invalid type is provided into the attribute
|
| 351 |
+
declaration.'''
|
| 352 |
+
message = 'Bad @ATTRIBUTE type, at line %d.'
|
| 353 |
+
|
| 354 |
+
class BadAttributeName(ArffException):
|
| 355 |
+
'''Error raised when an attribute name is provided twice the attribute
|
| 356 |
+
declaration.'''
|
| 357 |
+
|
| 358 |
+
def __init__(self, value, value2):
|
| 359 |
+
super().__init__()
|
| 360 |
+
self.message = (
|
| 361 |
+
('Bad @ATTRIBUTE name %s at line' % value) +
|
| 362 |
+
' %d, this name is already in use in line' +
|
| 363 |
+
(' %d.' % value2)
|
| 364 |
+
)
|
| 365 |
+
|
| 366 |
+
class BadNominalValue(ArffException):
|
| 367 |
+
'''Error raised when a value in used in some data instance but is not
|
| 368 |
+
declared into it respective attribute declaration.'''
|
| 369 |
+
|
| 370 |
+
def __init__(self, value):
|
| 371 |
+
super().__init__()
|
| 372 |
+
self.message = (
|
| 373 |
+
('Data value %s not found in nominal declaration, ' % value)
|
| 374 |
+
+ 'at line %d.'
|
| 375 |
+
)
|
| 376 |
+
|
| 377 |
+
class BadNominalFormatting(ArffException):
|
| 378 |
+
'''Error raised when a nominal value with space is not properly quoted.'''
|
| 379 |
+
def __init__(self, value):
|
| 380 |
+
super().__init__()
|
| 381 |
+
self.message = (
|
| 382 |
+
('Nominal data value "%s" not properly quoted in line ' % value) +
|
| 383 |
+
'%d.'
|
| 384 |
+
)
|
| 385 |
+
|
| 386 |
+
class BadNumericalValue(ArffException):
|
| 387 |
+
'''Error raised when and invalid numerical value is used in some data
|
| 388 |
+
instance.'''
|
| 389 |
+
message = 'Invalid numerical value, at line %d.'
|
| 390 |
+
|
| 391 |
+
class BadStringValue(ArffException):
|
| 392 |
+
'''Error raise when a string contains space but is not quoted.'''
|
| 393 |
+
message = 'Invalid string value at line %d.'
|
| 394 |
+
|
| 395 |
+
class BadLayout(ArffException):
|
| 396 |
+
'''Error raised when the layout of the ARFF file has something wrong.'''
|
| 397 |
+
message = 'Invalid layout of the ARFF file, at line %d.'
|
| 398 |
+
|
| 399 |
+
def __init__(self, msg=''):
|
| 400 |
+
super().__init__()
|
| 401 |
+
if msg:
|
| 402 |
+
self.message = BadLayout.message + ' ' + msg.replace('%', '%%')
|
| 403 |
+
|
| 404 |
+
|
| 405 |
+
class BadObject(ArffException):
|
| 406 |
+
'''Error raised when the object representing the ARFF file has something
|
| 407 |
+
wrong.'''
|
| 408 |
+
def __init__(self, msg='Invalid object.'):
|
| 409 |
+
self.msg = msg
|
| 410 |
+
|
| 411 |
+
def __str__(self):
|
| 412 |
+
return '%s' % self.msg
|
| 413 |
+
|
| 414 |
+
# =============================================================================
|
| 415 |
+
|
| 416 |
+
# INTERNAL ====================================================================
|
| 417 |
+
def _unescape_sub_callback(match):
|
| 418 |
+
return _UNESCAPE_SUB_MAP[match.group()]
|
| 419 |
+
|
| 420 |
+
|
| 421 |
+
def encode_string(s):
|
| 422 |
+
if _RE_QUOTE_CHARS.search(s):
|
| 423 |
+
return "'%s'" % _RE_ESCAPE_CHARS.sub(_unescape_sub_callback, s)
|
| 424 |
+
return s
|
| 425 |
+
|
| 426 |
+
|
| 427 |
+
class EncodedNominalConversor:
|
| 428 |
+
def __init__(self, values):
|
| 429 |
+
self.values = {v: i for i, v in enumerate(values)}
|
| 430 |
+
self.values[0] = 0
|
| 431 |
+
|
| 432 |
+
def __call__(self, value):
|
| 433 |
+
try:
|
| 434 |
+
return self.values[value]
|
| 435 |
+
except KeyError:
|
| 436 |
+
raise BadNominalValue(value)
|
| 437 |
+
|
| 438 |
+
|
| 439 |
+
class NominalConversor:
|
| 440 |
+
def __init__(self, values):
|
| 441 |
+
self.values = set(values)
|
| 442 |
+
self.zero_value = values[0]
|
| 443 |
+
|
| 444 |
+
def __call__(self, value):
|
| 445 |
+
if value not in self.values:
|
| 446 |
+
if value == 0:
|
| 447 |
+
# Sparse decode
|
| 448 |
+
# See issue #52: nominals should take their first value when
|
| 449 |
+
# unspecified in a sparse matrix. Naturally, this is consistent
|
| 450 |
+
# with EncodedNominalConversor.
|
| 451 |
+
return self.zero_value
|
| 452 |
+
raise BadNominalValue(value)
|
| 453 |
+
return str(value)
|
| 454 |
+
|
| 455 |
+
|
| 456 |
+
class DenseGeneratorData:
|
| 457 |
+
'''Internal helper class to allow for different matrix types without
|
| 458 |
+
making the code a huge collection of if statements.'''
|
| 459 |
+
|
| 460 |
+
def decode_rows(self, stream, conversors):
|
| 461 |
+
for row in stream:
|
| 462 |
+
values = _parse_values(row)
|
| 463 |
+
|
| 464 |
+
if isinstance(values, dict):
|
| 465 |
+
if values and max(values) >= len(conversors):
|
| 466 |
+
raise BadDataFormat(row)
|
| 467 |
+
# XXX: int 0 is used for implicit values, not '0'
|
| 468 |
+
values = [values[i] if i in values else 0 for i in
|
| 469 |
+
range(len(conversors))]
|
| 470 |
+
else:
|
| 471 |
+
if len(values) != len(conversors):
|
| 472 |
+
raise BadDataFormat(row)
|
| 473 |
+
|
| 474 |
+
yield self._decode_values(values, conversors)
|
| 475 |
+
|
| 476 |
+
@staticmethod
|
| 477 |
+
def _decode_values(values, conversors):
|
| 478 |
+
try:
|
| 479 |
+
values = [None if value is None else conversor(value)
|
| 480 |
+
for conversor, value
|
| 481 |
+
in zip(conversors, values)]
|
| 482 |
+
except ValueError as exc:
|
| 483 |
+
if 'float: ' in str(exc):
|
| 484 |
+
raise BadNumericalValue()
|
| 485 |
+
return values
|
| 486 |
+
|
| 487 |
+
def encode_data(self, data, attributes):
|
| 488 |
+
'''(INTERNAL) Encodes a line of data.
|
| 489 |
+
|
| 490 |
+
Data instances follow the csv format, i.e, attribute values are
|
| 491 |
+
delimited by commas. After converted from csv.
|
| 492 |
+
|
| 493 |
+
:param data: a list of values.
|
| 494 |
+
:param attributes: a list of attributes. Used to check if data is valid.
|
| 495 |
+
:return: a string with the encoded data line.
|
| 496 |
+
'''
|
| 497 |
+
current_row = 0
|
| 498 |
+
|
| 499 |
+
for inst in data:
|
| 500 |
+
if len(inst) != len(attributes):
|
| 501 |
+
raise BadObject(
|
| 502 |
+
'Instance %d has %d attributes, expected %d' %
|
| 503 |
+
(current_row, len(inst), len(attributes))
|
| 504 |
+
)
|
| 505 |
+
|
| 506 |
+
new_data = []
|
| 507 |
+
for value in inst:
|
| 508 |
+
if value is None or value == '' or value != value:
|
| 509 |
+
s = '?'
|
| 510 |
+
else:
|
| 511 |
+
s = encode_string(str(value))
|
| 512 |
+
new_data.append(s)
|
| 513 |
+
|
| 514 |
+
current_row += 1
|
| 515 |
+
yield ','.join(new_data)
|
| 516 |
+
|
| 517 |
+
|
| 518 |
+
class _DataListMixin:
|
| 519 |
+
"""Mixin to return a list from decode_rows instead of a generator"""
|
| 520 |
+
def decode_rows(self, stream, conversors):
|
| 521 |
+
return list(super().decode_rows(stream, conversors))
|
| 522 |
+
|
| 523 |
+
|
| 524 |
+
class Data(_DataListMixin, DenseGeneratorData):
|
| 525 |
+
pass
|
| 526 |
+
|
| 527 |
+
|
| 528 |
+
class COOData:
|
| 529 |
+
def decode_rows(self, stream, conversors):
|
| 530 |
+
data, rows, cols = [], [], []
|
| 531 |
+
for i, row in enumerate(stream):
|
| 532 |
+
values = _parse_values(row)
|
| 533 |
+
if not isinstance(values, dict):
|
| 534 |
+
raise BadLayout()
|
| 535 |
+
if not values:
|
| 536 |
+
continue
|
| 537 |
+
row_cols, values = zip(*sorted(values.items()))
|
| 538 |
+
try:
|
| 539 |
+
values = [value if value is None else conversors[key](value)
|
| 540 |
+
for key, value in zip(row_cols, values)]
|
| 541 |
+
except ValueError as exc:
|
| 542 |
+
if 'float: ' in str(exc):
|
| 543 |
+
raise BadNumericalValue()
|
| 544 |
+
raise
|
| 545 |
+
except IndexError:
|
| 546 |
+
# conversor out of range
|
| 547 |
+
raise BadDataFormat(row)
|
| 548 |
+
|
| 549 |
+
data.extend(values)
|
| 550 |
+
rows.extend([i] * len(values))
|
| 551 |
+
cols.extend(row_cols)
|
| 552 |
+
|
| 553 |
+
return data, rows, cols
|
| 554 |
+
|
| 555 |
+
def encode_data(self, data, attributes):
|
| 556 |
+
num_attributes = len(attributes)
|
| 557 |
+
new_data = []
|
| 558 |
+
current_row = 0
|
| 559 |
+
|
| 560 |
+
row = data.row
|
| 561 |
+
col = data.col
|
| 562 |
+
data = data.data
|
| 563 |
+
|
| 564 |
+
# Check if the rows are sorted
|
| 565 |
+
if not all(row[i] <= row[i + 1] for i in range(len(row) - 1)):
|
| 566 |
+
raise ValueError("liac-arff can only output COO matrices with "
|
| 567 |
+
"sorted rows.")
|
| 568 |
+
|
| 569 |
+
for v, col, row in zip(data, col, row):
|
| 570 |
+
if row > current_row:
|
| 571 |
+
# Add empty rows if necessary
|
| 572 |
+
while current_row < row:
|
| 573 |
+
yield " ".join(["{", ','.join(new_data), "}"])
|
| 574 |
+
new_data = []
|
| 575 |
+
current_row += 1
|
| 576 |
+
|
| 577 |
+
if col >= num_attributes:
|
| 578 |
+
raise BadObject(
|
| 579 |
+
'Instance %d has at least %d attributes, expected %d' %
|
| 580 |
+
(current_row, col + 1, num_attributes)
|
| 581 |
+
)
|
| 582 |
+
|
| 583 |
+
if v is None or v == '' or v != v:
|
| 584 |
+
s = '?'
|
| 585 |
+
else:
|
| 586 |
+
s = encode_string(str(v))
|
| 587 |
+
new_data.append("%d %s" % (col, s))
|
| 588 |
+
|
| 589 |
+
yield " ".join(["{", ','.join(new_data), "}"])
|
| 590 |
+
|
| 591 |
+
class LODGeneratorData:
|
| 592 |
+
def decode_rows(self, stream, conversors):
|
| 593 |
+
for row in stream:
|
| 594 |
+
values = _parse_values(row)
|
| 595 |
+
|
| 596 |
+
if not isinstance(values, dict):
|
| 597 |
+
raise BadLayout()
|
| 598 |
+
try:
|
| 599 |
+
yield {key: None if value is None else conversors[key](value)
|
| 600 |
+
for key, value in values.items()}
|
| 601 |
+
except ValueError as exc:
|
| 602 |
+
if 'float: ' in str(exc):
|
| 603 |
+
raise BadNumericalValue()
|
| 604 |
+
raise
|
| 605 |
+
except IndexError:
|
| 606 |
+
# conversor out of range
|
| 607 |
+
raise BadDataFormat(row)
|
| 608 |
+
|
| 609 |
+
def encode_data(self, data, attributes):
|
| 610 |
+
current_row = 0
|
| 611 |
+
|
| 612 |
+
num_attributes = len(attributes)
|
| 613 |
+
for row in data:
|
| 614 |
+
new_data = []
|
| 615 |
+
|
| 616 |
+
if len(row) > 0 and max(row) >= num_attributes:
|
| 617 |
+
raise BadObject(
|
| 618 |
+
'Instance %d has %d attributes, expected %d' %
|
| 619 |
+
(current_row, max(row) + 1, num_attributes)
|
| 620 |
+
)
|
| 621 |
+
|
| 622 |
+
for col in sorted(row):
|
| 623 |
+
v = row[col]
|
| 624 |
+
if v is None or v == '' or v != v:
|
| 625 |
+
s = '?'
|
| 626 |
+
else:
|
| 627 |
+
s = encode_string(str(v))
|
| 628 |
+
new_data.append("%d %s" % (col, s))
|
| 629 |
+
|
| 630 |
+
current_row += 1
|
| 631 |
+
yield " ".join(["{", ','.join(new_data), "}"])
|
| 632 |
+
|
| 633 |
+
class LODData(_DataListMixin, LODGeneratorData):
|
| 634 |
+
pass
|
| 635 |
+
|
| 636 |
+
|
| 637 |
+
def _get_data_object_for_decoding(matrix_type):
|
| 638 |
+
if matrix_type == DENSE:
|
| 639 |
+
return Data()
|
| 640 |
+
elif matrix_type == COO:
|
| 641 |
+
return COOData()
|
| 642 |
+
elif matrix_type == LOD:
|
| 643 |
+
return LODData()
|
| 644 |
+
elif matrix_type == DENSE_GEN:
|
| 645 |
+
return DenseGeneratorData()
|
| 646 |
+
elif matrix_type == LOD_GEN:
|
| 647 |
+
return LODGeneratorData()
|
| 648 |
+
else:
|
| 649 |
+
raise ValueError("Matrix type %s not supported." % str(matrix_type))
|
| 650 |
+
|
| 651 |
+
def _get_data_object_for_encoding(matrix):
|
| 652 |
+
# Probably a scipy.sparse
|
| 653 |
+
if hasattr(matrix, 'format'):
|
| 654 |
+
if matrix.format == 'coo':
|
| 655 |
+
return COOData()
|
| 656 |
+
else:
|
| 657 |
+
raise ValueError('Cannot guess matrix format!')
|
| 658 |
+
elif isinstance(matrix[0], dict):
|
| 659 |
+
return LODData()
|
| 660 |
+
else:
|
| 661 |
+
return Data()
|
| 662 |
+
|
| 663 |
+
# =============================================================================
|
| 664 |
+
|
| 665 |
+
# ADVANCED INTERFACE ==========================================================
|
| 666 |
+
class ArffDecoder:
|
| 667 |
+
'''An ARFF decoder.'''
|
| 668 |
+
|
| 669 |
+
def __init__(self):
|
| 670 |
+
'''Constructor.'''
|
| 671 |
+
self._conversors = []
|
| 672 |
+
self._current_line = 0
|
| 673 |
+
|
| 674 |
+
def _decode_comment(self, s):
|
| 675 |
+
'''(INTERNAL) Decodes a comment line.
|
| 676 |
+
|
| 677 |
+
Comments are single line strings starting, obligatorily, with the ``%``
|
| 678 |
+
character, and can have any symbol, including whitespaces or special
|
| 679 |
+
characters.
|
| 680 |
+
|
| 681 |
+
This method must receive a normalized string, i.e., a string without
|
| 682 |
+
padding, including the "\r\n" characters.
|
| 683 |
+
|
| 684 |
+
:param s: a normalized string.
|
| 685 |
+
:return: a string with the decoded comment.
|
| 686 |
+
'''
|
| 687 |
+
res = re.sub(r'^\%( )?', '', s)
|
| 688 |
+
return res
|
| 689 |
+
|
| 690 |
+
def _decode_relation(self, s):
|
| 691 |
+
'''(INTERNAL) Decodes a relation line.
|
| 692 |
+
|
| 693 |
+
The relation declaration is a line with the format ``@RELATION
|
| 694 |
+
<relation-name>``, where ``relation-name`` is a string. The string must
|
| 695 |
+
start with alphabetic character and must be quoted if the name includes
|
| 696 |
+
spaces, otherwise this method will raise a `BadRelationFormat` exception.
|
| 697 |
+
|
| 698 |
+
This method must receive a normalized string, i.e., a string without
|
| 699 |
+
padding, including the "\r\n" characters.
|
| 700 |
+
|
| 701 |
+
:param s: a normalized string.
|
| 702 |
+
:return: a string with the decoded relation name.
|
| 703 |
+
'''
|
| 704 |
+
_, v = s.split(' ', 1)
|
| 705 |
+
v = v.strip()
|
| 706 |
+
|
| 707 |
+
if not _RE_RELATION.match(v):
|
| 708 |
+
raise BadRelationFormat()
|
| 709 |
+
|
| 710 |
+
res = str(v.strip('"\''))
|
| 711 |
+
return res
|
| 712 |
+
|
| 713 |
+
def _decode_attribute(self, s):
|
| 714 |
+
'''(INTERNAL) Decodes an attribute line.
|
| 715 |
+
|
| 716 |
+
The attribute is the most complex declaration in an arff file. All
|
| 717 |
+
attributes must follow the template::
|
| 718 |
+
|
| 719 |
+
@attribute <attribute-name> <datatype>
|
| 720 |
+
|
| 721 |
+
where ``attribute-name`` is a string, quoted if the name contains any
|
| 722 |
+
whitespace, and ``datatype`` can be:
|
| 723 |
+
|
| 724 |
+
- Numerical attributes as ``NUMERIC``, ``INTEGER`` or ``REAL``.
|
| 725 |
+
- Strings as ``STRING``.
|
| 726 |
+
- Dates (NOT IMPLEMENTED).
|
| 727 |
+
- Nominal attributes with format:
|
| 728 |
+
|
| 729 |
+
{<nominal-name1>, <nominal-name2>, <nominal-name3>, ...}
|
| 730 |
+
|
| 731 |
+
The nominal names follow the rules for the attribute names, i.e., they
|
| 732 |
+
must be quoted if the name contains whitespaces.
|
| 733 |
+
|
| 734 |
+
This method must receive a normalized string, i.e., a string without
|
| 735 |
+
padding, including the "\r\n" characters.
|
| 736 |
+
|
| 737 |
+
:param s: a normalized string.
|
| 738 |
+
:return: a tuple (ATTRIBUTE_NAME, TYPE_OR_VALUES).
|
| 739 |
+
'''
|
| 740 |
+
_, v = s.split(' ', 1)
|
| 741 |
+
v = v.strip()
|
| 742 |
+
|
| 743 |
+
# Verify the general structure of declaration
|
| 744 |
+
m = _RE_ATTRIBUTE.match(v)
|
| 745 |
+
if not m:
|
| 746 |
+
raise BadAttributeFormat()
|
| 747 |
+
|
| 748 |
+
# Extracts the raw name and type
|
| 749 |
+
name, type_ = m.groups()
|
| 750 |
+
|
| 751 |
+
# Extracts the final name
|
| 752 |
+
name = str(name.strip('"\''))
|
| 753 |
+
|
| 754 |
+
# Extracts the final type
|
| 755 |
+
if type_[:1] == "{" and type_[-1:] == "}":
|
| 756 |
+
try:
|
| 757 |
+
type_ = _parse_values(type_.strip('{} '))
|
| 758 |
+
except Exception:
|
| 759 |
+
raise BadAttributeType()
|
| 760 |
+
if isinstance(type_, dict):
|
| 761 |
+
raise BadAttributeType()
|
| 762 |
+
|
| 763 |
+
else:
|
| 764 |
+
# If not nominal, verify the type name
|
| 765 |
+
type_ = str(type_).upper()
|
| 766 |
+
if type_ not in ['NUMERIC', 'REAL', 'INTEGER', 'STRING']:
|
| 767 |
+
raise BadAttributeType()
|
| 768 |
+
|
| 769 |
+
return (name, type_)
|
| 770 |
+
|
| 771 |
+
def _decode(self, s, encode_nominal=False, matrix_type=DENSE):
|
| 772 |
+
'''Do the job the ``encode``.'''
|
| 773 |
+
|
| 774 |
+
# Make sure this method is idempotent
|
| 775 |
+
self._current_line = 0
|
| 776 |
+
|
| 777 |
+
# If string, convert to a list of lines
|
| 778 |
+
if isinstance(s, str):
|
| 779 |
+
s = s.strip('\r\n ').replace('\r\n', '\n').split('\n')
|
| 780 |
+
|
| 781 |
+
# Create the return object
|
| 782 |
+
obj: ArffContainerType = {
|
| 783 |
+
'description': '',
|
| 784 |
+
'relation': '',
|
| 785 |
+
'attributes': [],
|
| 786 |
+
'data': []
|
| 787 |
+
}
|
| 788 |
+
attribute_names = {}
|
| 789 |
+
|
| 790 |
+
# Create the data helper object
|
| 791 |
+
data = _get_data_object_for_decoding(matrix_type)
|
| 792 |
+
|
| 793 |
+
# Read all lines
|
| 794 |
+
STATE = _TK_DESCRIPTION
|
| 795 |
+
s = iter(s)
|
| 796 |
+
for row in s:
|
| 797 |
+
self._current_line += 1
|
| 798 |
+
# Ignore empty lines
|
| 799 |
+
row = row.strip(' \r\n')
|
| 800 |
+
if not row: continue
|
| 801 |
+
|
| 802 |
+
u_row = row.upper()
|
| 803 |
+
|
| 804 |
+
# DESCRIPTION -----------------------------------------------------
|
| 805 |
+
if u_row.startswith(_TK_DESCRIPTION) and STATE == _TK_DESCRIPTION:
|
| 806 |
+
obj['description'] += self._decode_comment(row) + '\n'
|
| 807 |
+
# -----------------------------------------------------------------
|
| 808 |
+
|
| 809 |
+
# RELATION --------------------------------------------------------
|
| 810 |
+
elif u_row.startswith(_TK_RELATION):
|
| 811 |
+
if STATE != _TK_DESCRIPTION:
|
| 812 |
+
raise BadLayout()
|
| 813 |
+
|
| 814 |
+
STATE = _TK_RELATION
|
| 815 |
+
obj['relation'] = self._decode_relation(row)
|
| 816 |
+
# -----------------------------------------------------------------
|
| 817 |
+
|
| 818 |
+
# ATTRIBUTE -------------------------------------------------------
|
| 819 |
+
elif u_row.startswith(_TK_ATTRIBUTE):
|
| 820 |
+
if STATE != _TK_RELATION and STATE != _TK_ATTRIBUTE:
|
| 821 |
+
raise BadLayout()
|
| 822 |
+
|
| 823 |
+
STATE = _TK_ATTRIBUTE
|
| 824 |
+
|
| 825 |
+
attr = self._decode_attribute(row)
|
| 826 |
+
if attr[0] in attribute_names:
|
| 827 |
+
raise BadAttributeName(attr[0], attribute_names[attr[0]])
|
| 828 |
+
else:
|
| 829 |
+
attribute_names[attr[0]] = self._current_line
|
| 830 |
+
obj['attributes'].append(attr)
|
| 831 |
+
|
| 832 |
+
if isinstance(attr[1], (list, tuple)):
|
| 833 |
+
if encode_nominal:
|
| 834 |
+
conversor = EncodedNominalConversor(attr[1])
|
| 835 |
+
else:
|
| 836 |
+
conversor = NominalConversor(attr[1])
|
| 837 |
+
else:
|
| 838 |
+
CONVERSOR_MAP = {'STRING': str,
|
| 839 |
+
'INTEGER': lambda x: int(float(x)),
|
| 840 |
+
'NUMERIC': float,
|
| 841 |
+
'REAL': float}
|
| 842 |
+
conversor = CONVERSOR_MAP[attr[1]]
|
| 843 |
+
|
| 844 |
+
self._conversors.append(conversor)
|
| 845 |
+
# -----------------------------------------------------------------
|
| 846 |
+
|
| 847 |
+
# DATA ------------------------------------------------------------
|
| 848 |
+
elif u_row.startswith(_TK_DATA):
|
| 849 |
+
if STATE != _TK_ATTRIBUTE:
|
| 850 |
+
raise BadLayout()
|
| 851 |
+
|
| 852 |
+
break
|
| 853 |
+
# -----------------------------------------------------------------
|
| 854 |
+
|
| 855 |
+
# COMMENT ---------------------------------------------------------
|
| 856 |
+
elif u_row.startswith(_TK_COMMENT):
|
| 857 |
+
pass
|
| 858 |
+
# -----------------------------------------------------------------
|
| 859 |
+
else:
|
| 860 |
+
# Never found @DATA
|
| 861 |
+
raise BadLayout()
|
| 862 |
+
|
| 863 |
+
def stream():
|
| 864 |
+
for row in s:
|
| 865 |
+
self._current_line += 1
|
| 866 |
+
row = row.strip()
|
| 867 |
+
# Ignore empty lines and comment lines.
|
| 868 |
+
if row and not row.startswith(_TK_COMMENT):
|
| 869 |
+
yield row
|
| 870 |
+
|
| 871 |
+
# Alter the data object
|
| 872 |
+
obj['data'] = data.decode_rows(stream(), self._conversors)
|
| 873 |
+
if obj['description'].endswith('\n'):
|
| 874 |
+
obj['description'] = obj['description'][:-1]
|
| 875 |
+
|
| 876 |
+
return obj
|
| 877 |
+
|
| 878 |
+
def decode(self, s, encode_nominal=False, return_type=DENSE):
|
| 879 |
+
'''Returns the Python representation of a given ARFF file.
|
| 880 |
+
|
| 881 |
+
When a file object is passed as an argument, this method reads lines
|
| 882 |
+
iteratively, avoiding to load unnecessary information to the memory.
|
| 883 |
+
|
| 884 |
+
:param s: a string or file object with the ARFF file.
|
| 885 |
+
:param encode_nominal: boolean, if True perform a label encoding
|
| 886 |
+
while reading the .arff file.
|
| 887 |
+
:param return_type: determines the data structure used to store the
|
| 888 |
+
dataset. Can be one of `arff.DENSE`, `arff.COO`, `arff.LOD`,
|
| 889 |
+
`arff.DENSE_GEN` or `arff.LOD_GEN`.
|
| 890 |
+
Consult the sections on `working with sparse data`_ and `loading
|
| 891 |
+
progressively`_.
|
| 892 |
+
'''
|
| 893 |
+
try:
|
| 894 |
+
return self._decode(s, encode_nominal=encode_nominal,
|
| 895 |
+
matrix_type=return_type)
|
| 896 |
+
except ArffException as e:
|
| 897 |
+
e.line = self._current_line
|
| 898 |
+
raise e
|
| 899 |
+
|
| 900 |
+
|
| 901 |
+
class ArffEncoder:
|
| 902 |
+
'''An ARFF encoder.'''
|
| 903 |
+
|
| 904 |
+
def _encode_comment(self, s=''):
|
| 905 |
+
'''(INTERNAL) Encodes a comment line.
|
| 906 |
+
|
| 907 |
+
Comments are single line strings starting, obligatorily, with the ``%``
|
| 908 |
+
character, and can have any symbol, including whitespaces or special
|
| 909 |
+
characters.
|
| 910 |
+
|
| 911 |
+
If ``s`` is None, this method will simply return an empty comment.
|
| 912 |
+
|
| 913 |
+
:param s: (OPTIONAL) string.
|
| 914 |
+
:return: a string with the encoded comment line.
|
| 915 |
+
'''
|
| 916 |
+
if s:
|
| 917 |
+
return '%s %s'%(_TK_COMMENT, s)
|
| 918 |
+
else:
|
| 919 |
+
return '%s' % _TK_COMMENT
|
| 920 |
+
|
| 921 |
+
def _encode_relation(self, name):
|
| 922 |
+
'''(INTERNAL) Decodes a relation line.
|
| 923 |
+
|
| 924 |
+
The relation declaration is a line with the format ``@RELATION
|
| 925 |
+
<relation-name>``, where ``relation-name`` is a string.
|
| 926 |
+
|
| 927 |
+
:param name: a string.
|
| 928 |
+
:return: a string with the encoded relation declaration.
|
| 929 |
+
'''
|
| 930 |
+
for char in ' %{},':
|
| 931 |
+
if char in name:
|
| 932 |
+
name = '"%s"'%name
|
| 933 |
+
break
|
| 934 |
+
|
| 935 |
+
return '%s %s'%(_TK_RELATION, name)
|
| 936 |
+
|
| 937 |
+
def _encode_attribute(self, name, type_):
|
| 938 |
+
'''(INTERNAL) Encodes an attribute line.
|
| 939 |
+
|
| 940 |
+
The attribute follow the template::
|
| 941 |
+
|
| 942 |
+
@attribute <attribute-name> <datatype>
|
| 943 |
+
|
| 944 |
+
where ``attribute-name`` is a string, and ``datatype`` can be:
|
| 945 |
+
|
| 946 |
+
- Numerical attributes as ``NUMERIC``, ``INTEGER`` or ``REAL``.
|
| 947 |
+
- Strings as ``STRING``.
|
| 948 |
+
- Dates (NOT IMPLEMENTED).
|
| 949 |
+
- Nominal attributes with format:
|
| 950 |
+
|
| 951 |
+
{<nominal-name1>, <nominal-name2>, <nominal-name3>, ...}
|
| 952 |
+
|
| 953 |
+
This method must receive a the name of the attribute and its type, if
|
| 954 |
+
the attribute type is nominal, ``type`` must be a list of values.
|
| 955 |
+
|
| 956 |
+
:param name: a string.
|
| 957 |
+
:param type_: a string or a list of string.
|
| 958 |
+
:return: a string with the encoded attribute declaration.
|
| 959 |
+
'''
|
| 960 |
+
for char in ' %{},':
|
| 961 |
+
if char in name:
|
| 962 |
+
name = '"%s"'%name
|
| 963 |
+
break
|
| 964 |
+
|
| 965 |
+
if isinstance(type_, (tuple, list)):
|
| 966 |
+
type_tmp = ['%s' % encode_string(type_k) for type_k in type_]
|
| 967 |
+
type_ = '{%s}'%(', '.join(type_tmp))
|
| 968 |
+
|
| 969 |
+
return '%s %s %s'%(_TK_ATTRIBUTE, name, type_)
|
| 970 |
+
|
| 971 |
+
def encode(self, obj):
|
| 972 |
+
'''Encodes a given object to an ARFF file.
|
| 973 |
+
|
| 974 |
+
:param obj: the object containing the ARFF information.
|
| 975 |
+
:return: the ARFF file as an string.
|
| 976 |
+
'''
|
| 977 |
+
data = [row for row in self.iter_encode(obj)]
|
| 978 |
+
|
| 979 |
+
return '\n'.join(data)
|
| 980 |
+
|
| 981 |
+
def iter_encode(self, obj):
|
| 982 |
+
'''The iterative version of `arff.ArffEncoder.encode`.
|
| 983 |
+
|
| 984 |
+
This encodes iteratively a given object and return, one-by-one, the
|
| 985 |
+
lines of the ARFF file.
|
| 986 |
+
|
| 987 |
+
:param obj: the object containing the ARFF information.
|
| 988 |
+
:return: (yields) the ARFF file as strings.
|
| 989 |
+
'''
|
| 990 |
+
# DESCRIPTION
|
| 991 |
+
if obj.get('description', None):
|
| 992 |
+
for row in obj['description'].split('\n'):
|
| 993 |
+
yield self._encode_comment(row)
|
| 994 |
+
|
| 995 |
+
# RELATION
|
| 996 |
+
if not obj.get('relation'):
|
| 997 |
+
raise BadObject('Relation name not found or with invalid value.')
|
| 998 |
+
|
| 999 |
+
yield self._encode_relation(obj['relation'])
|
| 1000 |
+
yield ''
|
| 1001 |
+
|
| 1002 |
+
# ATTRIBUTES
|
| 1003 |
+
if not obj.get('attributes'):
|
| 1004 |
+
raise BadObject('Attributes not found.')
|
| 1005 |
+
|
| 1006 |
+
attribute_names = set()
|
| 1007 |
+
for attr in obj['attributes']:
|
| 1008 |
+
# Verify for bad object format
|
| 1009 |
+
if not isinstance(attr, (tuple, list)) or \
|
| 1010 |
+
len(attr) != 2 or \
|
| 1011 |
+
not isinstance(attr[0], str):
|
| 1012 |
+
raise BadObject('Invalid attribute declaration "%s"'%str(attr))
|
| 1013 |
+
|
| 1014 |
+
if isinstance(attr[1], str):
|
| 1015 |
+
# Verify for invalid types
|
| 1016 |
+
if attr[1] not in _SIMPLE_TYPES:
|
| 1017 |
+
raise BadObject('Invalid attribute type "%s"'%str(attr))
|
| 1018 |
+
|
| 1019 |
+
# Verify for bad object format
|
| 1020 |
+
elif not isinstance(attr[1], (tuple, list)):
|
| 1021 |
+
raise BadObject('Invalid attribute type "%s"'%str(attr))
|
| 1022 |
+
|
| 1023 |
+
# Verify attribute name is not used twice
|
| 1024 |
+
if attr[0] in attribute_names:
|
| 1025 |
+
raise BadObject('Trying to use attribute name "%s" for the '
|
| 1026 |
+
'second time.' % str(attr[0]))
|
| 1027 |
+
else:
|
| 1028 |
+
attribute_names.add(attr[0])
|
| 1029 |
+
|
| 1030 |
+
yield self._encode_attribute(attr[0], attr[1])
|
| 1031 |
+
yield ''
|
| 1032 |
+
attributes = obj['attributes']
|
| 1033 |
+
|
| 1034 |
+
# DATA
|
| 1035 |
+
yield _TK_DATA
|
| 1036 |
+
if 'data' in obj:
|
| 1037 |
+
data = _get_data_object_for_encoding(obj.get('data'))
|
| 1038 |
+
yield from data.encode_data(obj.get('data'), attributes)
|
| 1039 |
+
|
| 1040 |
+
yield ''
|
| 1041 |
+
|
| 1042 |
+
# =============================================================================
|
| 1043 |
+
|
| 1044 |
+
# BASIC INTERFACE =============================================================
|
| 1045 |
+
def load(fp, encode_nominal=False, return_type=DENSE):
|
| 1046 |
+
'''Load a file-like object containing the ARFF document and convert it into
|
| 1047 |
+
a Python object.
|
| 1048 |
+
|
| 1049 |
+
:param fp: a file-like object.
|
| 1050 |
+
:param encode_nominal: boolean, if True perform a label encoding
|
| 1051 |
+
while reading the .arff file.
|
| 1052 |
+
:param return_type: determines the data structure used to store the
|
| 1053 |
+
dataset. Can be one of `arff.DENSE`, `arff.COO`, `arff.LOD`,
|
| 1054 |
+
`arff.DENSE_GEN` or `arff.LOD_GEN`.
|
| 1055 |
+
Consult the sections on `working with sparse data`_ and `loading
|
| 1056 |
+
progressively`_.
|
| 1057 |
+
:return: a dictionary.
|
| 1058 |
+
'''
|
| 1059 |
+
decoder = ArffDecoder()
|
| 1060 |
+
return decoder.decode(fp, encode_nominal=encode_nominal,
|
| 1061 |
+
return_type=return_type)
|
| 1062 |
+
|
| 1063 |
+
def loads(s, encode_nominal=False, return_type=DENSE):
|
| 1064 |
+
'''Convert a string instance containing the ARFF document into a Python
|
| 1065 |
+
object.
|
| 1066 |
+
|
| 1067 |
+
:param s: a string object.
|
| 1068 |
+
:param encode_nominal: boolean, if True perform a label encoding
|
| 1069 |
+
while reading the .arff file.
|
| 1070 |
+
:param return_type: determines the data structure used to store the
|
| 1071 |
+
dataset. Can be one of `arff.DENSE`, `arff.COO`, `arff.LOD`,
|
| 1072 |
+
`arff.DENSE_GEN` or `arff.LOD_GEN`.
|
| 1073 |
+
Consult the sections on `working with sparse data`_ and `loading
|
| 1074 |
+
progressively`_.
|
| 1075 |
+
:return: a dictionary.
|
| 1076 |
+
'''
|
| 1077 |
+
decoder = ArffDecoder()
|
| 1078 |
+
return decoder.decode(s, encode_nominal=encode_nominal,
|
| 1079 |
+
return_type=return_type)
|
| 1080 |
+
|
| 1081 |
+
def dump(obj, fp):
|
| 1082 |
+
'''Serialize an object representing the ARFF document to a given file-like
|
| 1083 |
+
object.
|
| 1084 |
+
|
| 1085 |
+
:param obj: a dictionary.
|
| 1086 |
+
:param fp: a file-like object.
|
| 1087 |
+
'''
|
| 1088 |
+
encoder = ArffEncoder()
|
| 1089 |
+
generator = encoder.iter_encode(obj)
|
| 1090 |
+
|
| 1091 |
+
last_row = next(generator)
|
| 1092 |
+
for row in generator:
|
| 1093 |
+
fp.write(last_row + '\n')
|
| 1094 |
+
last_row = row
|
| 1095 |
+
fp.write(last_row)
|
| 1096 |
+
|
| 1097 |
+
return fp
|
| 1098 |
+
|
| 1099 |
+
def dumps(obj):
|
| 1100 |
+
'''Serialize an object representing the ARFF document, returning a string.
|
| 1101 |
+
|
| 1102 |
+
:param obj: a dictionary.
|
| 1103 |
+
:return: a string with the ARFF document.
|
| 1104 |
+
'''
|
| 1105 |
+
encoder = ArffEncoder()
|
| 1106 |
+
return encoder.encode(obj)
|
| 1107 |
+
# =============================================================================
|
videollama2/lib/python3.10/site-packages/sklearn/impute/_knn.py
ADDED
|
@@ -0,0 +1,391 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Authors: Ashim Bhattarai <ashimb9@gmail.com>
|
| 2 |
+
# Thomas J Fan <thomasjpfan@gmail.com>
|
| 3 |
+
# License: BSD 3 clause
|
| 4 |
+
|
| 5 |
+
from numbers import Integral
|
| 6 |
+
import numpy as np
|
| 7 |
+
|
| 8 |
+
from ._base import _BaseImputer
|
| 9 |
+
from ..utils.validation import FLOAT_DTYPES
|
| 10 |
+
from ..metrics import pairwise_distances_chunked
|
| 11 |
+
from ..metrics.pairwise import _NAN_METRICS
|
| 12 |
+
from ..neighbors._base import _get_weights
|
| 13 |
+
from ..utils import is_scalar_nan
|
| 14 |
+
from ..utils._mask import _get_mask
|
| 15 |
+
from ..utils.validation import check_is_fitted
|
| 16 |
+
from ..utils.validation import _check_feature_names_in
|
| 17 |
+
from ..utils._param_validation import Hidden, Interval, StrOptions
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
class KNNImputer(_BaseImputer):
|
| 21 |
+
"""Imputation for completing missing values using k-Nearest Neighbors.
|
| 22 |
+
|
| 23 |
+
Each sample's missing values are imputed using the mean value from
|
| 24 |
+
`n_neighbors` nearest neighbors found in the training set. Two samples are
|
| 25 |
+
close if the features that neither is missing are close.
|
| 26 |
+
|
| 27 |
+
Read more in the :ref:`User Guide <knnimpute>`.
|
| 28 |
+
|
| 29 |
+
.. versionadded:: 0.22
|
| 30 |
+
|
| 31 |
+
Parameters
|
| 32 |
+
----------
|
| 33 |
+
missing_values : int, float, str, np.nan or None, default=np.nan
|
| 34 |
+
The placeholder for the missing values. All occurrences of
|
| 35 |
+
`missing_values` will be imputed. For pandas' dataframes with
|
| 36 |
+
nullable integer dtypes with missing values, `missing_values`
|
| 37 |
+
should be set to np.nan, since `pd.NA` will be converted to np.nan.
|
| 38 |
+
|
| 39 |
+
n_neighbors : int, default=5
|
| 40 |
+
Number of neighboring samples to use for imputation.
|
| 41 |
+
|
| 42 |
+
weights : {'uniform', 'distance'} or callable, default='uniform'
|
| 43 |
+
Weight function used in prediction. Possible values:
|
| 44 |
+
|
| 45 |
+
- 'uniform' : uniform weights. All points in each neighborhood are
|
| 46 |
+
weighted equally.
|
| 47 |
+
- 'distance' : weight points by the inverse of their distance.
|
| 48 |
+
in this case, closer neighbors of a query point will have a
|
| 49 |
+
greater influence than neighbors which are further away.
|
| 50 |
+
- callable : a user-defined function which accepts an
|
| 51 |
+
array of distances, and returns an array of the same shape
|
| 52 |
+
containing the weights.
|
| 53 |
+
|
| 54 |
+
metric : {'nan_euclidean'} or callable, default='nan_euclidean'
|
| 55 |
+
Distance metric for searching neighbors. Possible values:
|
| 56 |
+
|
| 57 |
+
- 'nan_euclidean'
|
| 58 |
+
- callable : a user-defined function which conforms to the definition
|
| 59 |
+
of ``_pairwise_callable(X, Y, metric, **kwds)``. The function
|
| 60 |
+
accepts two arrays, X and Y, and a `missing_values` keyword in
|
| 61 |
+
`kwds` and returns a scalar distance value.
|
| 62 |
+
|
| 63 |
+
copy : bool, default=True
|
| 64 |
+
If True, a copy of X will be created. If False, imputation will
|
| 65 |
+
be done in-place whenever possible.
|
| 66 |
+
|
| 67 |
+
add_indicator : bool, default=False
|
| 68 |
+
If True, a :class:`MissingIndicator` transform will stack onto the
|
| 69 |
+
output of the imputer's transform. This allows a predictive estimator
|
| 70 |
+
to account for missingness despite imputation. If a feature has no
|
| 71 |
+
missing values at fit/train time, the feature won't appear on the
|
| 72 |
+
missing indicator even if there are missing values at transform/test
|
| 73 |
+
time.
|
| 74 |
+
|
| 75 |
+
keep_empty_features : bool, default=False
|
| 76 |
+
If True, features that consist exclusively of missing values when
|
| 77 |
+
`fit` is called are returned in results when `transform` is called.
|
| 78 |
+
The imputed value is always `0`.
|
| 79 |
+
|
| 80 |
+
.. versionadded:: 1.2
|
| 81 |
+
|
| 82 |
+
Attributes
|
| 83 |
+
----------
|
| 84 |
+
indicator_ : :class:`~sklearn.impute.MissingIndicator`
|
| 85 |
+
Indicator used to add binary indicators for missing values.
|
| 86 |
+
``None`` if add_indicator is False.
|
| 87 |
+
|
| 88 |
+
n_features_in_ : int
|
| 89 |
+
Number of features seen during :term:`fit`.
|
| 90 |
+
|
| 91 |
+
.. versionadded:: 0.24
|
| 92 |
+
|
| 93 |
+
feature_names_in_ : ndarray of shape (`n_features_in_`,)
|
| 94 |
+
Names of features seen during :term:`fit`. Defined only when `X`
|
| 95 |
+
has feature names that are all strings.
|
| 96 |
+
|
| 97 |
+
.. versionadded:: 1.0
|
| 98 |
+
|
| 99 |
+
See Also
|
| 100 |
+
--------
|
| 101 |
+
SimpleImputer : Univariate imputer for completing missing values
|
| 102 |
+
with simple strategies.
|
| 103 |
+
IterativeImputer : Multivariate imputer that estimates values to impute for
|
| 104 |
+
each feature with missing values from all the others.
|
| 105 |
+
|
| 106 |
+
References
|
| 107 |
+
----------
|
| 108 |
+
* Olga Troyanskaya, Michael Cantor, Gavin Sherlock, Pat Brown, Trevor
|
| 109 |
+
Hastie, Robert Tibshirani, David Botstein and Russ B. Altman, Missing
|
| 110 |
+
value estimation methods for DNA microarrays, BIOINFORMATICS Vol. 17
|
| 111 |
+
no. 6, 2001 Pages 520-525.
|
| 112 |
+
|
| 113 |
+
Examples
|
| 114 |
+
--------
|
| 115 |
+
>>> import numpy as np
|
| 116 |
+
>>> from sklearn.impute import KNNImputer
|
| 117 |
+
>>> X = [[1, 2, np.nan], [3, 4, 3], [np.nan, 6, 5], [8, 8, 7]]
|
| 118 |
+
>>> imputer = KNNImputer(n_neighbors=2)
|
| 119 |
+
>>> imputer.fit_transform(X)
|
| 120 |
+
array([[1. , 2. , 4. ],
|
| 121 |
+
[3. , 4. , 3. ],
|
| 122 |
+
[5.5, 6. , 5. ],
|
| 123 |
+
[8. , 8. , 7. ]])
|
| 124 |
+
"""
|
| 125 |
+
|
| 126 |
+
_parameter_constraints: dict = {
|
| 127 |
+
**_BaseImputer._parameter_constraints,
|
| 128 |
+
"n_neighbors": [Interval(Integral, 1, None, closed="left")],
|
| 129 |
+
"weights": [StrOptions({"uniform", "distance"}), callable, Hidden(None)],
|
| 130 |
+
"metric": [StrOptions(set(_NAN_METRICS)), callable],
|
| 131 |
+
"copy": ["boolean"],
|
| 132 |
+
}
|
| 133 |
+
|
| 134 |
+
def __init__(
|
| 135 |
+
self,
|
| 136 |
+
*,
|
| 137 |
+
missing_values=np.nan,
|
| 138 |
+
n_neighbors=5,
|
| 139 |
+
weights="uniform",
|
| 140 |
+
metric="nan_euclidean",
|
| 141 |
+
copy=True,
|
| 142 |
+
add_indicator=False,
|
| 143 |
+
keep_empty_features=False,
|
| 144 |
+
):
|
| 145 |
+
super().__init__(
|
| 146 |
+
missing_values=missing_values,
|
| 147 |
+
add_indicator=add_indicator,
|
| 148 |
+
keep_empty_features=keep_empty_features,
|
| 149 |
+
)
|
| 150 |
+
self.n_neighbors = n_neighbors
|
| 151 |
+
self.weights = weights
|
| 152 |
+
self.metric = metric
|
| 153 |
+
self.copy = copy
|
| 154 |
+
|
| 155 |
+
def _calc_impute(self, dist_pot_donors, n_neighbors, fit_X_col, mask_fit_X_col):
|
| 156 |
+
"""Helper function to impute a single column.
|
| 157 |
+
|
| 158 |
+
Parameters
|
| 159 |
+
----------
|
| 160 |
+
dist_pot_donors : ndarray of shape (n_receivers, n_potential_donors)
|
| 161 |
+
Distance matrix between the receivers and potential donors from
|
| 162 |
+
training set. There must be at least one non-nan distance between
|
| 163 |
+
a receiver and a potential donor.
|
| 164 |
+
|
| 165 |
+
n_neighbors : int
|
| 166 |
+
Number of neighbors to consider.
|
| 167 |
+
|
| 168 |
+
fit_X_col : ndarray of shape (n_potential_donors,)
|
| 169 |
+
Column of potential donors from training set.
|
| 170 |
+
|
| 171 |
+
mask_fit_X_col : ndarray of shape (n_potential_donors,)
|
| 172 |
+
Missing mask for fit_X_col.
|
| 173 |
+
|
| 174 |
+
Returns
|
| 175 |
+
-------
|
| 176 |
+
imputed_values: ndarray of shape (n_receivers,)
|
| 177 |
+
Imputed values for receiver.
|
| 178 |
+
"""
|
| 179 |
+
# Get donors
|
| 180 |
+
donors_idx = np.argpartition(dist_pot_donors, n_neighbors - 1, axis=1)[
|
| 181 |
+
:, :n_neighbors
|
| 182 |
+
]
|
| 183 |
+
|
| 184 |
+
# Get weight matrix from distance matrix
|
| 185 |
+
donors_dist = dist_pot_donors[
|
| 186 |
+
np.arange(donors_idx.shape[0])[:, None], donors_idx
|
| 187 |
+
]
|
| 188 |
+
|
| 189 |
+
weight_matrix = _get_weights(donors_dist, self.weights)
|
| 190 |
+
|
| 191 |
+
# fill nans with zeros
|
| 192 |
+
if weight_matrix is not None:
|
| 193 |
+
weight_matrix[np.isnan(weight_matrix)] = 0.0
|
| 194 |
+
|
| 195 |
+
# Retrieve donor values and calculate kNN average
|
| 196 |
+
donors = fit_X_col.take(donors_idx)
|
| 197 |
+
donors_mask = mask_fit_X_col.take(donors_idx)
|
| 198 |
+
donors = np.ma.array(donors, mask=donors_mask)
|
| 199 |
+
|
| 200 |
+
return np.ma.average(donors, axis=1, weights=weight_matrix).data
|
| 201 |
+
|
| 202 |
+
def fit(self, X, y=None):
|
| 203 |
+
"""Fit the imputer on X.
|
| 204 |
+
|
| 205 |
+
Parameters
|
| 206 |
+
----------
|
| 207 |
+
X : array-like shape of (n_samples, n_features)
|
| 208 |
+
Input data, where `n_samples` is the number of samples and
|
| 209 |
+
`n_features` is the number of features.
|
| 210 |
+
|
| 211 |
+
y : Ignored
|
| 212 |
+
Not used, present here for API consistency by convention.
|
| 213 |
+
|
| 214 |
+
Returns
|
| 215 |
+
-------
|
| 216 |
+
self : object
|
| 217 |
+
The fitted `KNNImputer` class instance.
|
| 218 |
+
"""
|
| 219 |
+
self._validate_params()
|
| 220 |
+
# Check data integrity and calling arguments
|
| 221 |
+
if not is_scalar_nan(self.missing_values):
|
| 222 |
+
force_all_finite = True
|
| 223 |
+
else:
|
| 224 |
+
force_all_finite = "allow-nan"
|
| 225 |
+
|
| 226 |
+
X = self._validate_data(
|
| 227 |
+
X,
|
| 228 |
+
accept_sparse=False,
|
| 229 |
+
dtype=FLOAT_DTYPES,
|
| 230 |
+
force_all_finite=force_all_finite,
|
| 231 |
+
copy=self.copy,
|
| 232 |
+
)
|
| 233 |
+
|
| 234 |
+
self._fit_X = X
|
| 235 |
+
self._mask_fit_X = _get_mask(self._fit_X, self.missing_values)
|
| 236 |
+
self._valid_mask = ~np.all(self._mask_fit_X, axis=0)
|
| 237 |
+
|
| 238 |
+
super()._fit_indicator(self._mask_fit_X)
|
| 239 |
+
|
| 240 |
+
return self
|
| 241 |
+
|
| 242 |
+
def transform(self, X):
|
| 243 |
+
"""Impute all missing values in X.
|
| 244 |
+
|
| 245 |
+
Parameters
|
| 246 |
+
----------
|
| 247 |
+
X : array-like of shape (n_samples, n_features)
|
| 248 |
+
The input data to complete.
|
| 249 |
+
|
| 250 |
+
Returns
|
| 251 |
+
-------
|
| 252 |
+
X : array-like of shape (n_samples, n_output_features)
|
| 253 |
+
The imputed dataset. `n_output_features` is the number of features
|
| 254 |
+
that is not always missing during `fit`.
|
| 255 |
+
"""
|
| 256 |
+
|
| 257 |
+
check_is_fitted(self)
|
| 258 |
+
if not is_scalar_nan(self.missing_values):
|
| 259 |
+
force_all_finite = True
|
| 260 |
+
else:
|
| 261 |
+
force_all_finite = "allow-nan"
|
| 262 |
+
X = self._validate_data(
|
| 263 |
+
X,
|
| 264 |
+
accept_sparse=False,
|
| 265 |
+
dtype=FLOAT_DTYPES,
|
| 266 |
+
force_all_finite=force_all_finite,
|
| 267 |
+
copy=self.copy,
|
| 268 |
+
reset=False,
|
| 269 |
+
)
|
| 270 |
+
|
| 271 |
+
mask = _get_mask(X, self.missing_values)
|
| 272 |
+
mask_fit_X = self._mask_fit_X
|
| 273 |
+
valid_mask = self._valid_mask
|
| 274 |
+
|
| 275 |
+
X_indicator = super()._transform_indicator(mask)
|
| 276 |
+
|
| 277 |
+
# Removes columns where the training data is all nan
|
| 278 |
+
if not np.any(mask):
|
| 279 |
+
# No missing values in X
|
| 280 |
+
if self.keep_empty_features:
|
| 281 |
+
Xc = X
|
| 282 |
+
Xc[:, ~valid_mask] = 0
|
| 283 |
+
else:
|
| 284 |
+
Xc = X[:, valid_mask]
|
| 285 |
+
return Xc
|
| 286 |
+
|
| 287 |
+
row_missing_idx = np.flatnonzero(mask.any(axis=1))
|
| 288 |
+
|
| 289 |
+
non_missing_fix_X = np.logical_not(mask_fit_X)
|
| 290 |
+
|
| 291 |
+
# Maps from indices from X to indices in dist matrix
|
| 292 |
+
dist_idx_map = np.zeros(X.shape[0], dtype=int)
|
| 293 |
+
dist_idx_map[row_missing_idx] = np.arange(row_missing_idx.shape[0])
|
| 294 |
+
|
| 295 |
+
def process_chunk(dist_chunk, start):
|
| 296 |
+
row_missing_chunk = row_missing_idx[start : start + len(dist_chunk)]
|
| 297 |
+
|
| 298 |
+
# Find and impute missing by column
|
| 299 |
+
for col in range(X.shape[1]):
|
| 300 |
+
if not valid_mask[col]:
|
| 301 |
+
# column was all missing during training
|
| 302 |
+
continue
|
| 303 |
+
|
| 304 |
+
col_mask = mask[row_missing_chunk, col]
|
| 305 |
+
if not np.any(col_mask):
|
| 306 |
+
# column has no missing values
|
| 307 |
+
continue
|
| 308 |
+
|
| 309 |
+
(potential_donors_idx,) = np.nonzero(non_missing_fix_X[:, col])
|
| 310 |
+
|
| 311 |
+
# receivers_idx are indices in X
|
| 312 |
+
receivers_idx = row_missing_chunk[np.flatnonzero(col_mask)]
|
| 313 |
+
|
| 314 |
+
# distances for samples that needed imputation for column
|
| 315 |
+
dist_subset = dist_chunk[dist_idx_map[receivers_idx] - start][
|
| 316 |
+
:, potential_donors_idx
|
| 317 |
+
]
|
| 318 |
+
|
| 319 |
+
# receivers with all nan distances impute with mean
|
| 320 |
+
all_nan_dist_mask = np.isnan(dist_subset).all(axis=1)
|
| 321 |
+
all_nan_receivers_idx = receivers_idx[all_nan_dist_mask]
|
| 322 |
+
|
| 323 |
+
if all_nan_receivers_idx.size:
|
| 324 |
+
col_mean = np.ma.array(
|
| 325 |
+
self._fit_X[:, col], mask=mask_fit_X[:, col]
|
| 326 |
+
).mean()
|
| 327 |
+
X[all_nan_receivers_idx, col] = col_mean
|
| 328 |
+
|
| 329 |
+
if len(all_nan_receivers_idx) == len(receivers_idx):
|
| 330 |
+
# all receivers imputed with mean
|
| 331 |
+
continue
|
| 332 |
+
|
| 333 |
+
# receivers with at least one defined distance
|
| 334 |
+
receivers_idx = receivers_idx[~all_nan_dist_mask]
|
| 335 |
+
dist_subset = dist_chunk[dist_idx_map[receivers_idx] - start][
|
| 336 |
+
:, potential_donors_idx
|
| 337 |
+
]
|
| 338 |
+
|
| 339 |
+
n_neighbors = min(self.n_neighbors, len(potential_donors_idx))
|
| 340 |
+
value = self._calc_impute(
|
| 341 |
+
dist_subset,
|
| 342 |
+
n_neighbors,
|
| 343 |
+
self._fit_X[potential_donors_idx, col],
|
| 344 |
+
mask_fit_X[potential_donors_idx, col],
|
| 345 |
+
)
|
| 346 |
+
X[receivers_idx, col] = value
|
| 347 |
+
|
| 348 |
+
# process in fixed-memory chunks
|
| 349 |
+
gen = pairwise_distances_chunked(
|
| 350 |
+
X[row_missing_idx, :],
|
| 351 |
+
self._fit_X,
|
| 352 |
+
metric=self.metric,
|
| 353 |
+
missing_values=self.missing_values,
|
| 354 |
+
force_all_finite=force_all_finite,
|
| 355 |
+
reduce_func=process_chunk,
|
| 356 |
+
)
|
| 357 |
+
for chunk in gen:
|
| 358 |
+
# process_chunk modifies X in place. No return value.
|
| 359 |
+
pass
|
| 360 |
+
|
| 361 |
+
if self.keep_empty_features:
|
| 362 |
+
Xc = X
|
| 363 |
+
Xc[:, ~valid_mask] = 0
|
| 364 |
+
else:
|
| 365 |
+
Xc = X[:, valid_mask]
|
| 366 |
+
|
| 367 |
+
return super()._concatenate_indicator(Xc, X_indicator)
|
| 368 |
+
|
| 369 |
+
def get_feature_names_out(self, input_features=None):
|
| 370 |
+
"""Get output feature names for transformation.
|
| 371 |
+
|
| 372 |
+
Parameters
|
| 373 |
+
----------
|
| 374 |
+
input_features : array-like of str or None, default=None
|
| 375 |
+
Input features.
|
| 376 |
+
|
| 377 |
+
- If `input_features` is `None`, then `feature_names_in_` is
|
| 378 |
+
used as feature names in. If `feature_names_in_` is not defined,
|
| 379 |
+
then the following input feature names are generated:
|
| 380 |
+
`["x0", "x1", ..., "x(n_features_in_ - 1)"]`.
|
| 381 |
+
- If `input_features` is an array-like, then `input_features` must
|
| 382 |
+
match `feature_names_in_` if `feature_names_in_` is defined.
|
| 383 |
+
|
| 384 |
+
Returns
|
| 385 |
+
-------
|
| 386 |
+
feature_names_out : ndarray of str objects
|
| 387 |
+
Transformed feature names.
|
| 388 |
+
"""
|
| 389 |
+
input_features = _check_feature_names_in(self, input_features)
|
| 390 |
+
names = input_features[self._valid_mask]
|
| 391 |
+
return self._concatenate_indicator_feature_names_out(names, input_features)
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/__init__.py
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
The :mod:`sklearn.semi_supervised` module implements semi-supervised learning
|
| 3 |
+
algorithms. These algorithms utilize small amounts of labeled data and large
|
| 4 |
+
amounts of unlabeled data for classification tasks. This module includes Label
|
| 5 |
+
Propagation.
|
| 6 |
+
"""
|
| 7 |
+
|
| 8 |
+
from ._label_propagation import LabelPropagation, LabelSpreading
|
| 9 |
+
from ._self_training import SelfTrainingClassifier
|
| 10 |
+
|
| 11 |
+
__all__ = ["SelfTrainingClassifier", "LabelPropagation", "LabelSpreading"]
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/__pycache__/__init__.cpython-310.pyc
ADDED
|
Binary file (626 Bytes). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/__pycache__/_label_propagation.cpython-310.pyc
ADDED
|
Binary file (18.8 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/_label_propagation.py
ADDED
|
@@ -0,0 +1,621 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# coding=utf8
|
| 2 |
+
"""
|
| 3 |
+
Label propagation in the context of this module refers to a set of
|
| 4 |
+
semi-supervised classification algorithms. At a high level, these algorithms
|
| 5 |
+
work by forming a fully-connected graph between all points given and solving
|
| 6 |
+
for the steady-state distribution of labels at each point.
|
| 7 |
+
|
| 8 |
+
These algorithms perform very well in practice. The cost of running can be very
|
| 9 |
+
expensive, at approximately O(N^3) where N is the number of (labeled and
|
| 10 |
+
unlabeled) points. The theory (why they perform so well) is motivated by
|
| 11 |
+
intuitions from random walk algorithms and geometric relationships in the data.
|
| 12 |
+
For more information see the references below.
|
| 13 |
+
|
| 14 |
+
Model Features
|
| 15 |
+
--------------
|
| 16 |
+
Label clamping:
|
| 17 |
+
The algorithm tries to learn distributions of labels over the dataset given
|
| 18 |
+
label assignments over an initial subset. In one variant, the algorithm does
|
| 19 |
+
not allow for any errors in the initial assignment (hard-clamping) while
|
| 20 |
+
in another variant, the algorithm allows for some wiggle room for the initial
|
| 21 |
+
assignments, allowing them to change by a fraction alpha in each iteration
|
| 22 |
+
(soft-clamping).
|
| 23 |
+
|
| 24 |
+
Kernel:
|
| 25 |
+
A function which projects a vector into some higher dimensional space. This
|
| 26 |
+
implementation supports RBF and KNN kernels. Using the RBF kernel generates
|
| 27 |
+
a dense matrix of size O(N^2). KNN kernel will generate a sparse matrix of
|
| 28 |
+
size O(k*N) which will run much faster. See the documentation for SVMs for
|
| 29 |
+
more info on kernels.
|
| 30 |
+
|
| 31 |
+
Examples
|
| 32 |
+
--------
|
| 33 |
+
>>> import numpy as np
|
| 34 |
+
>>> from sklearn import datasets
|
| 35 |
+
>>> from sklearn.semi_supervised import LabelPropagation
|
| 36 |
+
>>> label_prop_model = LabelPropagation()
|
| 37 |
+
>>> iris = datasets.load_iris()
|
| 38 |
+
>>> rng = np.random.RandomState(42)
|
| 39 |
+
>>> random_unlabeled_points = rng.rand(len(iris.target)) < 0.3
|
| 40 |
+
>>> labels = np.copy(iris.target)
|
| 41 |
+
>>> labels[random_unlabeled_points] = -1
|
| 42 |
+
>>> label_prop_model.fit(iris.data, labels)
|
| 43 |
+
LabelPropagation(...)
|
| 44 |
+
|
| 45 |
+
Notes
|
| 46 |
+
-----
|
| 47 |
+
References:
|
| 48 |
+
[1] Yoshua Bengio, Olivier Delalleau, Nicolas Le Roux. In Semi-Supervised
|
| 49 |
+
Learning (2006), pp. 193-216
|
| 50 |
+
|
| 51 |
+
[2] Olivier Delalleau, Yoshua Bengio, Nicolas Le Roux. Efficient
|
| 52 |
+
Non-Parametric Function Induction in Semi-Supervised Learning. AISTAT 2005
|
| 53 |
+
"""
|
| 54 |
+
|
| 55 |
+
# Authors: Clay Woolam <clay@woolam.org>
|
| 56 |
+
# Utkarsh Upadhyay <mail@musicallyut.in>
|
| 57 |
+
# License: BSD
|
| 58 |
+
from abc import ABCMeta, abstractmethod
|
| 59 |
+
from numbers import Integral, Real
|
| 60 |
+
|
| 61 |
+
import warnings
|
| 62 |
+
import numpy as np
|
| 63 |
+
from scipy import sparse
|
| 64 |
+
from scipy.sparse import csgraph
|
| 65 |
+
|
| 66 |
+
from ..base import BaseEstimator, ClassifierMixin
|
| 67 |
+
from ..metrics.pairwise import rbf_kernel
|
| 68 |
+
from ..neighbors import NearestNeighbors
|
| 69 |
+
from ..utils.extmath import safe_sparse_dot
|
| 70 |
+
from ..utils.multiclass import check_classification_targets
|
| 71 |
+
from ..utils.validation import check_is_fitted
|
| 72 |
+
from ..utils._param_validation import Interval, StrOptions
|
| 73 |
+
from ..exceptions import ConvergenceWarning
|
| 74 |
+
|
| 75 |
+
|
| 76 |
+
class BaseLabelPropagation(ClassifierMixin, BaseEstimator, metaclass=ABCMeta):
|
| 77 |
+
"""Base class for label propagation module.
|
| 78 |
+
|
| 79 |
+
Parameters
|
| 80 |
+
----------
|
| 81 |
+
kernel : {'knn', 'rbf'} or callable, default='rbf'
|
| 82 |
+
String identifier for kernel function to use or the kernel function
|
| 83 |
+
itself. Only 'rbf' and 'knn' strings are valid inputs. The function
|
| 84 |
+
passed should take two inputs, each of shape (n_samples, n_features),
|
| 85 |
+
and return a (n_samples, n_samples) shaped weight matrix.
|
| 86 |
+
|
| 87 |
+
gamma : float, default=20
|
| 88 |
+
Parameter for rbf kernel.
|
| 89 |
+
|
| 90 |
+
n_neighbors : int, default=7
|
| 91 |
+
Parameter for knn kernel. Need to be strictly positive.
|
| 92 |
+
|
| 93 |
+
alpha : float, default=1.0
|
| 94 |
+
Clamping factor.
|
| 95 |
+
|
| 96 |
+
max_iter : int, default=30
|
| 97 |
+
Change maximum number of iterations allowed.
|
| 98 |
+
|
| 99 |
+
tol : float, default=1e-3
|
| 100 |
+
Convergence tolerance: threshold to consider the system at steady
|
| 101 |
+
state.
|
| 102 |
+
|
| 103 |
+
n_jobs : int, default=None
|
| 104 |
+
The number of parallel jobs to run.
|
| 105 |
+
``None`` means 1 unless in a :obj:`joblib.parallel_backend` context.
|
| 106 |
+
``-1`` means using all processors. See :term:`Glossary <n_jobs>`
|
| 107 |
+
for more details.
|
| 108 |
+
"""
|
| 109 |
+
|
| 110 |
+
_parameter_constraints: dict = {
|
| 111 |
+
"kernel": [StrOptions({"knn", "rbf"}), callable],
|
| 112 |
+
"gamma": [Interval(Real, 0, None, closed="left")],
|
| 113 |
+
"n_neighbors": [Interval(Integral, 0, None, closed="neither")],
|
| 114 |
+
"alpha": [None, Interval(Real, 0, 1, closed="neither")],
|
| 115 |
+
"max_iter": [Interval(Integral, 0, None, closed="neither")],
|
| 116 |
+
"tol": [Interval(Real, 0, None, closed="left")],
|
| 117 |
+
"n_jobs": [None, Integral],
|
| 118 |
+
}
|
| 119 |
+
|
| 120 |
+
def __init__(
|
| 121 |
+
self,
|
| 122 |
+
kernel="rbf",
|
| 123 |
+
*,
|
| 124 |
+
gamma=20,
|
| 125 |
+
n_neighbors=7,
|
| 126 |
+
alpha=1,
|
| 127 |
+
max_iter=30,
|
| 128 |
+
tol=1e-3,
|
| 129 |
+
n_jobs=None,
|
| 130 |
+
):
|
| 131 |
+
|
| 132 |
+
self.max_iter = max_iter
|
| 133 |
+
self.tol = tol
|
| 134 |
+
|
| 135 |
+
# kernel parameters
|
| 136 |
+
self.kernel = kernel
|
| 137 |
+
self.gamma = gamma
|
| 138 |
+
self.n_neighbors = n_neighbors
|
| 139 |
+
|
| 140 |
+
# clamping factor
|
| 141 |
+
self.alpha = alpha
|
| 142 |
+
|
| 143 |
+
self.n_jobs = n_jobs
|
| 144 |
+
|
| 145 |
+
def _get_kernel(self, X, y=None):
|
| 146 |
+
if self.kernel == "rbf":
|
| 147 |
+
if y is None:
|
| 148 |
+
return rbf_kernel(X, X, gamma=self.gamma)
|
| 149 |
+
else:
|
| 150 |
+
return rbf_kernel(X, y, gamma=self.gamma)
|
| 151 |
+
elif self.kernel == "knn":
|
| 152 |
+
if self.nn_fit is None:
|
| 153 |
+
self.nn_fit = NearestNeighbors(
|
| 154 |
+
n_neighbors=self.n_neighbors, n_jobs=self.n_jobs
|
| 155 |
+
).fit(X)
|
| 156 |
+
if y is None:
|
| 157 |
+
return self.nn_fit.kneighbors_graph(
|
| 158 |
+
self.nn_fit._fit_X, self.n_neighbors, mode="connectivity"
|
| 159 |
+
)
|
| 160 |
+
else:
|
| 161 |
+
return self.nn_fit.kneighbors(y, return_distance=False)
|
| 162 |
+
elif callable(self.kernel):
|
| 163 |
+
if y is None:
|
| 164 |
+
return self.kernel(X, X)
|
| 165 |
+
else:
|
| 166 |
+
return self.kernel(X, y)
|
| 167 |
+
|
| 168 |
+
@abstractmethod
|
| 169 |
+
def _build_graph(self):
|
| 170 |
+
raise NotImplementedError(
|
| 171 |
+
"Graph construction must be implemented to fit a label propagation model."
|
| 172 |
+
)
|
| 173 |
+
|
| 174 |
+
def predict(self, X):
|
| 175 |
+
"""Perform inductive inference across the model.
|
| 176 |
+
|
| 177 |
+
Parameters
|
| 178 |
+
----------
|
| 179 |
+
X : array-like of shape (n_samples, n_features)
|
| 180 |
+
The data matrix.
|
| 181 |
+
|
| 182 |
+
Returns
|
| 183 |
+
-------
|
| 184 |
+
y : ndarray of shape (n_samples,)
|
| 185 |
+
Predictions for input data.
|
| 186 |
+
"""
|
| 187 |
+
# Note: since `predict` does not accept semi-supervised labels as input,
|
| 188 |
+
# `fit(X, y).predict(X) != fit(X, y).transduction_`.
|
| 189 |
+
# Hence, `fit_predict` is not implemented.
|
| 190 |
+
# See https://github.com/scikit-learn/scikit-learn/pull/24898
|
| 191 |
+
probas = self.predict_proba(X)
|
| 192 |
+
return self.classes_[np.argmax(probas, axis=1)].ravel()
|
| 193 |
+
|
| 194 |
+
def predict_proba(self, X):
|
| 195 |
+
"""Predict probability for each possible outcome.
|
| 196 |
+
|
| 197 |
+
Compute the probability estimates for each single sample in X
|
| 198 |
+
and each possible outcome seen during training (categorical
|
| 199 |
+
distribution).
|
| 200 |
+
|
| 201 |
+
Parameters
|
| 202 |
+
----------
|
| 203 |
+
X : array-like of shape (n_samples, n_features)
|
| 204 |
+
The data matrix.
|
| 205 |
+
|
| 206 |
+
Returns
|
| 207 |
+
-------
|
| 208 |
+
probabilities : ndarray of shape (n_samples, n_classes)
|
| 209 |
+
Normalized probability distributions across
|
| 210 |
+
class labels.
|
| 211 |
+
"""
|
| 212 |
+
check_is_fitted(self)
|
| 213 |
+
|
| 214 |
+
X_2d = self._validate_data(
|
| 215 |
+
X,
|
| 216 |
+
accept_sparse=["csc", "csr", "coo", "dok", "bsr", "lil", "dia"],
|
| 217 |
+
reset=False,
|
| 218 |
+
)
|
| 219 |
+
weight_matrices = self._get_kernel(self.X_, X_2d)
|
| 220 |
+
if self.kernel == "knn":
|
| 221 |
+
probabilities = np.array(
|
| 222 |
+
[
|
| 223 |
+
np.sum(self.label_distributions_[weight_matrix], axis=0)
|
| 224 |
+
for weight_matrix in weight_matrices
|
| 225 |
+
]
|
| 226 |
+
)
|
| 227 |
+
else:
|
| 228 |
+
weight_matrices = weight_matrices.T
|
| 229 |
+
probabilities = safe_sparse_dot(weight_matrices, self.label_distributions_)
|
| 230 |
+
normalizer = np.atleast_2d(np.sum(probabilities, axis=1)).T
|
| 231 |
+
probabilities /= normalizer
|
| 232 |
+
return probabilities
|
| 233 |
+
|
| 234 |
+
def fit(self, X, y):
|
| 235 |
+
"""Fit a semi-supervised label propagation model to X.
|
| 236 |
+
|
| 237 |
+
The input samples (labeled and unlabeled) are provided by matrix X,
|
| 238 |
+
and target labels are provided by matrix y. We conventionally apply the
|
| 239 |
+
label -1 to unlabeled samples in matrix y in a semi-supervised
|
| 240 |
+
classification.
|
| 241 |
+
|
| 242 |
+
Parameters
|
| 243 |
+
----------
|
| 244 |
+
X : array-like of shape (n_samples, n_features)
|
| 245 |
+
Training data, where `n_samples` is the number of samples
|
| 246 |
+
and `n_features` is the number of features.
|
| 247 |
+
|
| 248 |
+
y : array-like of shape (n_samples,)
|
| 249 |
+
Target class values with unlabeled points marked as -1.
|
| 250 |
+
All unlabeled samples will be transductively assigned labels
|
| 251 |
+
internally, which are stored in `transduction_`.
|
| 252 |
+
|
| 253 |
+
Returns
|
| 254 |
+
-------
|
| 255 |
+
self : object
|
| 256 |
+
Returns the instance itself.
|
| 257 |
+
"""
|
| 258 |
+
self._validate_params()
|
| 259 |
+
X, y = self._validate_data(X, y)
|
| 260 |
+
self.X_ = X
|
| 261 |
+
check_classification_targets(y)
|
| 262 |
+
|
| 263 |
+
# actual graph construction (implementations should override this)
|
| 264 |
+
graph_matrix = self._build_graph()
|
| 265 |
+
|
| 266 |
+
# label construction
|
| 267 |
+
# construct a categorical distribution for classification only
|
| 268 |
+
classes = np.unique(y)
|
| 269 |
+
classes = classes[classes != -1]
|
| 270 |
+
self.classes_ = classes
|
| 271 |
+
|
| 272 |
+
n_samples, n_classes = len(y), len(classes)
|
| 273 |
+
|
| 274 |
+
y = np.asarray(y)
|
| 275 |
+
unlabeled = y == -1
|
| 276 |
+
|
| 277 |
+
# initialize distributions
|
| 278 |
+
self.label_distributions_ = np.zeros((n_samples, n_classes))
|
| 279 |
+
for label in classes:
|
| 280 |
+
self.label_distributions_[y == label, classes == label] = 1
|
| 281 |
+
|
| 282 |
+
y_static = np.copy(self.label_distributions_)
|
| 283 |
+
if self._variant == "propagation":
|
| 284 |
+
# LabelPropagation
|
| 285 |
+
y_static[unlabeled] = 0
|
| 286 |
+
else:
|
| 287 |
+
# LabelSpreading
|
| 288 |
+
y_static *= 1 - self.alpha
|
| 289 |
+
|
| 290 |
+
l_previous = np.zeros((self.X_.shape[0], n_classes))
|
| 291 |
+
|
| 292 |
+
unlabeled = unlabeled[:, np.newaxis]
|
| 293 |
+
if sparse.isspmatrix(graph_matrix):
|
| 294 |
+
graph_matrix = graph_matrix.tocsr()
|
| 295 |
+
|
| 296 |
+
for self.n_iter_ in range(self.max_iter):
|
| 297 |
+
if np.abs(self.label_distributions_ - l_previous).sum() < self.tol:
|
| 298 |
+
break
|
| 299 |
+
|
| 300 |
+
l_previous = self.label_distributions_
|
| 301 |
+
self.label_distributions_ = safe_sparse_dot(
|
| 302 |
+
graph_matrix, self.label_distributions_
|
| 303 |
+
)
|
| 304 |
+
|
| 305 |
+
if self._variant == "propagation":
|
| 306 |
+
normalizer = np.sum(self.label_distributions_, axis=1)[:, np.newaxis]
|
| 307 |
+
normalizer[normalizer == 0] = 1
|
| 308 |
+
self.label_distributions_ /= normalizer
|
| 309 |
+
self.label_distributions_ = np.where(
|
| 310 |
+
unlabeled, self.label_distributions_, y_static
|
| 311 |
+
)
|
| 312 |
+
else:
|
| 313 |
+
# clamp
|
| 314 |
+
self.label_distributions_ = (
|
| 315 |
+
np.multiply(self.alpha, self.label_distributions_) + y_static
|
| 316 |
+
)
|
| 317 |
+
else:
|
| 318 |
+
warnings.warn(
|
| 319 |
+
"max_iter=%d was reached without convergence." % self.max_iter,
|
| 320 |
+
category=ConvergenceWarning,
|
| 321 |
+
)
|
| 322 |
+
self.n_iter_ += 1
|
| 323 |
+
|
| 324 |
+
normalizer = np.sum(self.label_distributions_, axis=1)[:, np.newaxis]
|
| 325 |
+
normalizer[normalizer == 0] = 1
|
| 326 |
+
self.label_distributions_ /= normalizer
|
| 327 |
+
|
| 328 |
+
# set the transduction item
|
| 329 |
+
transduction = self.classes_[np.argmax(self.label_distributions_, axis=1)]
|
| 330 |
+
self.transduction_ = transduction.ravel()
|
| 331 |
+
return self
|
| 332 |
+
|
| 333 |
+
|
| 334 |
+
class LabelPropagation(BaseLabelPropagation):
|
| 335 |
+
"""Label Propagation classifier.
|
| 336 |
+
|
| 337 |
+
Read more in the :ref:`User Guide <label_propagation>`.
|
| 338 |
+
|
| 339 |
+
Parameters
|
| 340 |
+
----------
|
| 341 |
+
kernel : {'knn', 'rbf'} or callable, default='rbf'
|
| 342 |
+
String identifier for kernel function to use or the kernel function
|
| 343 |
+
itself. Only 'rbf' and 'knn' strings are valid inputs. The function
|
| 344 |
+
passed should take two inputs, each of shape (n_samples, n_features),
|
| 345 |
+
and return a (n_samples, n_samples) shaped weight matrix.
|
| 346 |
+
|
| 347 |
+
gamma : float, default=20
|
| 348 |
+
Parameter for rbf kernel.
|
| 349 |
+
|
| 350 |
+
n_neighbors : int, default=7
|
| 351 |
+
Parameter for knn kernel which need to be strictly positive.
|
| 352 |
+
|
| 353 |
+
max_iter : int, default=1000
|
| 354 |
+
Change maximum number of iterations allowed.
|
| 355 |
+
|
| 356 |
+
tol : float, 1e-3
|
| 357 |
+
Convergence tolerance: threshold to consider the system at steady
|
| 358 |
+
state.
|
| 359 |
+
|
| 360 |
+
n_jobs : int, default=None
|
| 361 |
+
The number of parallel jobs to run.
|
| 362 |
+
``None`` means 1 unless in a :obj:`joblib.parallel_backend` context.
|
| 363 |
+
``-1`` means using all processors. See :term:`Glossary <n_jobs>`
|
| 364 |
+
for more details.
|
| 365 |
+
|
| 366 |
+
Attributes
|
| 367 |
+
----------
|
| 368 |
+
X_ : ndarray of shape (n_samples, n_features)
|
| 369 |
+
Input array.
|
| 370 |
+
|
| 371 |
+
classes_ : ndarray of shape (n_classes,)
|
| 372 |
+
The distinct labels used in classifying instances.
|
| 373 |
+
|
| 374 |
+
label_distributions_ : ndarray of shape (n_samples, n_classes)
|
| 375 |
+
Categorical distribution for each item.
|
| 376 |
+
|
| 377 |
+
transduction_ : ndarray of shape (n_samples)
|
| 378 |
+
Label assigned to each item during :term:`fit`.
|
| 379 |
+
|
| 380 |
+
n_features_in_ : int
|
| 381 |
+
Number of features seen during :term:`fit`.
|
| 382 |
+
|
| 383 |
+
.. versionadded:: 0.24
|
| 384 |
+
|
| 385 |
+
feature_names_in_ : ndarray of shape (`n_features_in_`,)
|
| 386 |
+
Names of features seen during :term:`fit`. Defined only when `X`
|
| 387 |
+
has feature names that are all strings.
|
| 388 |
+
|
| 389 |
+
.. versionadded:: 1.0
|
| 390 |
+
|
| 391 |
+
n_iter_ : int
|
| 392 |
+
Number of iterations run.
|
| 393 |
+
|
| 394 |
+
See Also
|
| 395 |
+
--------
|
| 396 |
+
BaseLabelPropagation : Base class for label propagation module.
|
| 397 |
+
LabelSpreading : Alternate label propagation strategy more robust to noise.
|
| 398 |
+
|
| 399 |
+
References
|
| 400 |
+
----------
|
| 401 |
+
Xiaojin Zhu and Zoubin Ghahramani. Learning from labeled and unlabeled data
|
| 402 |
+
with label propagation. Technical Report CMU-CALD-02-107, Carnegie Mellon
|
| 403 |
+
University, 2002 http://pages.cs.wisc.edu/~jerryzhu/pub/CMU-CALD-02-107.pdf
|
| 404 |
+
|
| 405 |
+
Examples
|
| 406 |
+
--------
|
| 407 |
+
>>> import numpy as np
|
| 408 |
+
>>> from sklearn import datasets
|
| 409 |
+
>>> from sklearn.semi_supervised import LabelPropagation
|
| 410 |
+
>>> label_prop_model = LabelPropagation()
|
| 411 |
+
>>> iris = datasets.load_iris()
|
| 412 |
+
>>> rng = np.random.RandomState(42)
|
| 413 |
+
>>> random_unlabeled_points = rng.rand(len(iris.target)) < 0.3
|
| 414 |
+
>>> labels = np.copy(iris.target)
|
| 415 |
+
>>> labels[random_unlabeled_points] = -1
|
| 416 |
+
>>> label_prop_model.fit(iris.data, labels)
|
| 417 |
+
LabelPropagation(...)
|
| 418 |
+
"""
|
| 419 |
+
|
| 420 |
+
_variant = "propagation"
|
| 421 |
+
|
| 422 |
+
_parameter_constraints: dict = {**BaseLabelPropagation._parameter_constraints}
|
| 423 |
+
_parameter_constraints.pop("alpha")
|
| 424 |
+
|
| 425 |
+
def __init__(
|
| 426 |
+
self,
|
| 427 |
+
kernel="rbf",
|
| 428 |
+
*,
|
| 429 |
+
gamma=20,
|
| 430 |
+
n_neighbors=7,
|
| 431 |
+
max_iter=1000,
|
| 432 |
+
tol=1e-3,
|
| 433 |
+
n_jobs=None,
|
| 434 |
+
):
|
| 435 |
+
super().__init__(
|
| 436 |
+
kernel=kernel,
|
| 437 |
+
gamma=gamma,
|
| 438 |
+
n_neighbors=n_neighbors,
|
| 439 |
+
max_iter=max_iter,
|
| 440 |
+
tol=tol,
|
| 441 |
+
n_jobs=n_jobs,
|
| 442 |
+
alpha=None,
|
| 443 |
+
)
|
| 444 |
+
|
| 445 |
+
def _build_graph(self):
|
| 446 |
+
"""Matrix representing a fully connected graph between each sample
|
| 447 |
+
|
| 448 |
+
This basic implementation creates a non-stochastic affinity matrix, so
|
| 449 |
+
class distributions will exceed 1 (normalization may be desired).
|
| 450 |
+
"""
|
| 451 |
+
if self.kernel == "knn":
|
| 452 |
+
self.nn_fit = None
|
| 453 |
+
affinity_matrix = self._get_kernel(self.X_)
|
| 454 |
+
normalizer = affinity_matrix.sum(axis=0)
|
| 455 |
+
if sparse.isspmatrix(affinity_matrix):
|
| 456 |
+
affinity_matrix.data /= np.diag(np.array(normalizer))
|
| 457 |
+
else:
|
| 458 |
+
affinity_matrix /= normalizer[:, np.newaxis]
|
| 459 |
+
return affinity_matrix
|
| 460 |
+
|
| 461 |
+
def fit(self, X, y):
|
| 462 |
+
"""Fit a semi-supervised label propagation model to X.
|
| 463 |
+
|
| 464 |
+
Parameters
|
| 465 |
+
----------
|
| 466 |
+
X : array-like of shape (n_samples, n_features)
|
| 467 |
+
Training data, where `n_samples` is the number of samples
|
| 468 |
+
and `n_features` is the number of features.
|
| 469 |
+
|
| 470 |
+
y : array-like of shape (n_samples,)
|
| 471 |
+
Target class values with unlabeled points marked as -1.
|
| 472 |
+
All unlabeled samples will be transductively assigned labels
|
| 473 |
+
internally, which are stored in `transduction_`.
|
| 474 |
+
|
| 475 |
+
Returns
|
| 476 |
+
-------
|
| 477 |
+
self : object
|
| 478 |
+
Returns the instance itself.
|
| 479 |
+
"""
|
| 480 |
+
return super().fit(X, y)
|
| 481 |
+
|
| 482 |
+
|
| 483 |
+
class LabelSpreading(BaseLabelPropagation):
|
| 484 |
+
"""LabelSpreading model for semi-supervised learning.
|
| 485 |
+
|
| 486 |
+
This model is similar to the basic Label Propagation algorithm,
|
| 487 |
+
but uses affinity matrix based on the normalized graph Laplacian
|
| 488 |
+
and soft clamping across the labels.
|
| 489 |
+
|
| 490 |
+
Read more in the :ref:`User Guide <label_propagation>`.
|
| 491 |
+
|
| 492 |
+
Parameters
|
| 493 |
+
----------
|
| 494 |
+
kernel : {'knn', 'rbf'} or callable, default='rbf'
|
| 495 |
+
String identifier for kernel function to use or the kernel function
|
| 496 |
+
itself. Only 'rbf' and 'knn' strings are valid inputs. The function
|
| 497 |
+
passed should take two inputs, each of shape (n_samples, n_features),
|
| 498 |
+
and return a (n_samples, n_samples) shaped weight matrix.
|
| 499 |
+
|
| 500 |
+
gamma : float, default=20
|
| 501 |
+
Parameter for rbf kernel.
|
| 502 |
+
|
| 503 |
+
n_neighbors : int, default=7
|
| 504 |
+
Parameter for knn kernel which is a strictly positive integer.
|
| 505 |
+
|
| 506 |
+
alpha : float, default=0.2
|
| 507 |
+
Clamping factor. A value in (0, 1) that specifies the relative amount
|
| 508 |
+
that an instance should adopt the information from its neighbors as
|
| 509 |
+
opposed to its initial label.
|
| 510 |
+
alpha=0 means keeping the initial label information; alpha=1 means
|
| 511 |
+
replacing all initial information.
|
| 512 |
+
|
| 513 |
+
max_iter : int, default=30
|
| 514 |
+
Maximum number of iterations allowed.
|
| 515 |
+
|
| 516 |
+
tol : float, default=1e-3
|
| 517 |
+
Convergence tolerance: threshold to consider the system at steady
|
| 518 |
+
state.
|
| 519 |
+
|
| 520 |
+
n_jobs : int, default=None
|
| 521 |
+
The number of parallel jobs to run.
|
| 522 |
+
``None`` means 1 unless in a :obj:`joblib.parallel_backend` context.
|
| 523 |
+
``-1`` means using all processors. See :term:`Glossary <n_jobs>`
|
| 524 |
+
for more details.
|
| 525 |
+
|
| 526 |
+
Attributes
|
| 527 |
+
----------
|
| 528 |
+
X_ : ndarray of shape (n_samples, n_features)
|
| 529 |
+
Input array.
|
| 530 |
+
|
| 531 |
+
classes_ : ndarray of shape (n_classes,)
|
| 532 |
+
The distinct labels used in classifying instances.
|
| 533 |
+
|
| 534 |
+
label_distributions_ : ndarray of shape (n_samples, n_classes)
|
| 535 |
+
Categorical distribution for each item.
|
| 536 |
+
|
| 537 |
+
transduction_ : ndarray of shape (n_samples,)
|
| 538 |
+
Label assigned to each item during :term:`fit`.
|
| 539 |
+
|
| 540 |
+
n_features_in_ : int
|
| 541 |
+
Number of features seen during :term:`fit`.
|
| 542 |
+
|
| 543 |
+
.. versionadded:: 0.24
|
| 544 |
+
|
| 545 |
+
feature_names_in_ : ndarray of shape (`n_features_in_`,)
|
| 546 |
+
Names of features seen during :term:`fit`. Defined only when `X`
|
| 547 |
+
has feature names that are all strings.
|
| 548 |
+
|
| 549 |
+
.. versionadded:: 1.0
|
| 550 |
+
|
| 551 |
+
n_iter_ : int
|
| 552 |
+
Number of iterations run.
|
| 553 |
+
|
| 554 |
+
See Also
|
| 555 |
+
--------
|
| 556 |
+
LabelPropagation : Unregularized graph based semi-supervised learning.
|
| 557 |
+
|
| 558 |
+
References
|
| 559 |
+
----------
|
| 560 |
+
`Dengyong Zhou, Olivier Bousquet, Thomas Navin Lal, Jason Weston,
|
| 561 |
+
Bernhard Schoelkopf. Learning with local and global consistency (2004)
|
| 562 |
+
<https://citeseerx.ist.psu.edu/doc_view/pid/d74c37aabf2d5cae663007cbd8718175466aea8c>`_
|
| 563 |
+
|
| 564 |
+
Examples
|
| 565 |
+
--------
|
| 566 |
+
>>> import numpy as np
|
| 567 |
+
>>> from sklearn import datasets
|
| 568 |
+
>>> from sklearn.semi_supervised import LabelSpreading
|
| 569 |
+
>>> label_prop_model = LabelSpreading()
|
| 570 |
+
>>> iris = datasets.load_iris()
|
| 571 |
+
>>> rng = np.random.RandomState(42)
|
| 572 |
+
>>> random_unlabeled_points = rng.rand(len(iris.target)) < 0.3
|
| 573 |
+
>>> labels = np.copy(iris.target)
|
| 574 |
+
>>> labels[random_unlabeled_points] = -1
|
| 575 |
+
>>> label_prop_model.fit(iris.data, labels)
|
| 576 |
+
LabelSpreading(...)
|
| 577 |
+
"""
|
| 578 |
+
|
| 579 |
+
_variant = "spreading"
|
| 580 |
+
|
| 581 |
+
_parameter_constraints: dict = {**BaseLabelPropagation._parameter_constraints}
|
| 582 |
+
_parameter_constraints["alpha"] = [Interval(Real, 0, 1, closed="neither")]
|
| 583 |
+
|
| 584 |
+
def __init__(
|
| 585 |
+
self,
|
| 586 |
+
kernel="rbf",
|
| 587 |
+
*,
|
| 588 |
+
gamma=20,
|
| 589 |
+
n_neighbors=7,
|
| 590 |
+
alpha=0.2,
|
| 591 |
+
max_iter=30,
|
| 592 |
+
tol=1e-3,
|
| 593 |
+
n_jobs=None,
|
| 594 |
+
):
|
| 595 |
+
|
| 596 |
+
# this one has different base parameters
|
| 597 |
+
super().__init__(
|
| 598 |
+
kernel=kernel,
|
| 599 |
+
gamma=gamma,
|
| 600 |
+
n_neighbors=n_neighbors,
|
| 601 |
+
alpha=alpha,
|
| 602 |
+
max_iter=max_iter,
|
| 603 |
+
tol=tol,
|
| 604 |
+
n_jobs=n_jobs,
|
| 605 |
+
)
|
| 606 |
+
|
| 607 |
+
def _build_graph(self):
|
| 608 |
+
"""Graph matrix for Label Spreading computes the graph laplacian"""
|
| 609 |
+
# compute affinity matrix (or gram matrix)
|
| 610 |
+
if self.kernel == "knn":
|
| 611 |
+
self.nn_fit = None
|
| 612 |
+
n_samples = self.X_.shape[0]
|
| 613 |
+
affinity_matrix = self._get_kernel(self.X_)
|
| 614 |
+
laplacian = csgraph.laplacian(affinity_matrix, normed=True)
|
| 615 |
+
laplacian = -laplacian
|
| 616 |
+
if sparse.isspmatrix(laplacian):
|
| 617 |
+
diag_mask = laplacian.row == laplacian.col
|
| 618 |
+
laplacian.data[diag_mask] = 0.0
|
| 619 |
+
else:
|
| 620 |
+
laplacian.flat[:: n_samples + 1] = 0.0 # set diag to 0.0
|
| 621 |
+
return laplacian
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/_self_training.py
ADDED
|
@@ -0,0 +1,400 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import warnings
|
| 2 |
+
from numbers import Integral, Real
|
| 3 |
+
|
| 4 |
+
import numpy as np
|
| 5 |
+
|
| 6 |
+
from ..base import MetaEstimatorMixin, clone, BaseEstimator
|
| 7 |
+
from ..utils._param_validation import HasMethods, Interval, StrOptions
|
| 8 |
+
from ..utils.validation import check_is_fitted
|
| 9 |
+
from ..utils.metaestimators import available_if
|
| 10 |
+
from ..utils import safe_mask
|
| 11 |
+
|
| 12 |
+
__all__ = ["SelfTrainingClassifier"]
|
| 13 |
+
|
| 14 |
+
# Authors: Oliver Rausch <rauscho@ethz.ch>
|
| 15 |
+
# Patrice Becker <beckerp@ethz.ch>
|
| 16 |
+
# License: BSD 3 clause
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
def _estimator_has(attr):
|
| 20 |
+
"""Check if `self.base_estimator_ `or `self.base_estimator_` has `attr`."""
|
| 21 |
+
return lambda self: (
|
| 22 |
+
hasattr(self.base_estimator_, attr)
|
| 23 |
+
if hasattr(self, "base_estimator_")
|
| 24 |
+
else hasattr(self.base_estimator, attr)
|
| 25 |
+
)
|
| 26 |
+
|
| 27 |
+
|
| 28 |
+
class SelfTrainingClassifier(MetaEstimatorMixin, BaseEstimator):
|
| 29 |
+
"""Self-training classifier.
|
| 30 |
+
|
| 31 |
+
This :term:`metaestimator` allows a given supervised classifier to function as a
|
| 32 |
+
semi-supervised classifier, allowing it to learn from unlabeled data. It
|
| 33 |
+
does this by iteratively predicting pseudo-labels for the unlabeled data
|
| 34 |
+
and adding them to the training set.
|
| 35 |
+
|
| 36 |
+
The classifier will continue iterating until either max_iter is reached, or
|
| 37 |
+
no pseudo-labels were added to the training set in the previous iteration.
|
| 38 |
+
|
| 39 |
+
Read more in the :ref:`User Guide <self_training>`.
|
| 40 |
+
|
| 41 |
+
Parameters
|
| 42 |
+
----------
|
| 43 |
+
base_estimator : estimator object
|
| 44 |
+
An estimator object implementing `fit` and `predict_proba`.
|
| 45 |
+
Invoking the `fit` method will fit a clone of the passed estimator,
|
| 46 |
+
which will be stored in the `base_estimator_` attribute.
|
| 47 |
+
|
| 48 |
+
threshold : float, default=0.75
|
| 49 |
+
The decision threshold for use with `criterion='threshold'`.
|
| 50 |
+
Should be in [0, 1). When using the `'threshold'` criterion, a
|
| 51 |
+
:ref:`well calibrated classifier <calibration>` should be used.
|
| 52 |
+
|
| 53 |
+
criterion : {'threshold', 'k_best'}, default='threshold'
|
| 54 |
+
The selection criterion used to select which labels to add to the
|
| 55 |
+
training set. If `'threshold'`, pseudo-labels with prediction
|
| 56 |
+
probabilities above `threshold` are added to the dataset. If `'k_best'`,
|
| 57 |
+
the `k_best` pseudo-labels with highest prediction probabilities are
|
| 58 |
+
added to the dataset. When using the 'threshold' criterion, a
|
| 59 |
+
:ref:`well calibrated classifier <calibration>` should be used.
|
| 60 |
+
|
| 61 |
+
k_best : int, default=10
|
| 62 |
+
The amount of samples to add in each iteration. Only used when
|
| 63 |
+
`criterion='k_best'`.
|
| 64 |
+
|
| 65 |
+
max_iter : int or None, default=10
|
| 66 |
+
Maximum number of iterations allowed. Should be greater than or equal
|
| 67 |
+
to 0. If it is `None`, the classifier will continue to predict labels
|
| 68 |
+
until no new pseudo-labels are added, or all unlabeled samples have
|
| 69 |
+
been labeled.
|
| 70 |
+
|
| 71 |
+
verbose : bool, default=False
|
| 72 |
+
Enable verbose output.
|
| 73 |
+
|
| 74 |
+
Attributes
|
| 75 |
+
----------
|
| 76 |
+
base_estimator_ : estimator object
|
| 77 |
+
The fitted estimator.
|
| 78 |
+
|
| 79 |
+
classes_ : ndarray or list of ndarray of shape (n_classes,)
|
| 80 |
+
Class labels for each output. (Taken from the trained
|
| 81 |
+
`base_estimator_`).
|
| 82 |
+
|
| 83 |
+
transduction_ : ndarray of shape (n_samples,)
|
| 84 |
+
The labels used for the final fit of the classifier, including
|
| 85 |
+
pseudo-labels added during fit.
|
| 86 |
+
|
| 87 |
+
labeled_iter_ : ndarray of shape (n_samples,)
|
| 88 |
+
The iteration in which each sample was labeled. When a sample has
|
| 89 |
+
iteration 0, the sample was already labeled in the original dataset.
|
| 90 |
+
When a sample has iteration -1, the sample was not labeled in any
|
| 91 |
+
iteration.
|
| 92 |
+
|
| 93 |
+
n_features_in_ : int
|
| 94 |
+
Number of features seen during :term:`fit`.
|
| 95 |
+
|
| 96 |
+
.. versionadded:: 0.24
|
| 97 |
+
|
| 98 |
+
feature_names_in_ : ndarray of shape (`n_features_in_`,)
|
| 99 |
+
Names of features seen during :term:`fit`. Defined only when `X`
|
| 100 |
+
has feature names that are all strings.
|
| 101 |
+
|
| 102 |
+
.. versionadded:: 1.0
|
| 103 |
+
|
| 104 |
+
n_iter_ : int
|
| 105 |
+
The number of rounds of self-training, that is the number of times the
|
| 106 |
+
base estimator is fitted on relabeled variants of the training set.
|
| 107 |
+
|
| 108 |
+
termination_condition_ : {'max_iter', 'no_change', 'all_labeled'}
|
| 109 |
+
The reason that fitting was stopped.
|
| 110 |
+
|
| 111 |
+
- `'max_iter'`: `n_iter_` reached `max_iter`.
|
| 112 |
+
- `'no_change'`: no new labels were predicted.
|
| 113 |
+
- `'all_labeled'`: all unlabeled samples were labeled before `max_iter`
|
| 114 |
+
was reached.
|
| 115 |
+
|
| 116 |
+
See Also
|
| 117 |
+
--------
|
| 118 |
+
LabelPropagation : Label propagation classifier.
|
| 119 |
+
LabelSpreading : Label spreading model for semi-supervised learning.
|
| 120 |
+
|
| 121 |
+
References
|
| 122 |
+
----------
|
| 123 |
+
:doi:`David Yarowsky. 1995. Unsupervised word sense disambiguation rivaling
|
| 124 |
+
supervised methods. In Proceedings of the 33rd annual meeting on
|
| 125 |
+
Association for Computational Linguistics (ACL '95). Association for
|
| 126 |
+
Computational Linguistics, Stroudsburg, PA, USA, 189-196.
|
| 127 |
+
<10.3115/981658.981684>`
|
| 128 |
+
|
| 129 |
+
Examples
|
| 130 |
+
--------
|
| 131 |
+
>>> import numpy as np
|
| 132 |
+
>>> from sklearn import datasets
|
| 133 |
+
>>> from sklearn.semi_supervised import SelfTrainingClassifier
|
| 134 |
+
>>> from sklearn.svm import SVC
|
| 135 |
+
>>> rng = np.random.RandomState(42)
|
| 136 |
+
>>> iris = datasets.load_iris()
|
| 137 |
+
>>> random_unlabeled_points = rng.rand(iris.target.shape[0]) < 0.3
|
| 138 |
+
>>> iris.target[random_unlabeled_points] = -1
|
| 139 |
+
>>> svc = SVC(probability=True, gamma="auto")
|
| 140 |
+
>>> self_training_model = SelfTrainingClassifier(svc)
|
| 141 |
+
>>> self_training_model.fit(iris.data, iris.target)
|
| 142 |
+
SelfTrainingClassifier(...)
|
| 143 |
+
"""
|
| 144 |
+
|
| 145 |
+
_estimator_type = "classifier"
|
| 146 |
+
|
| 147 |
+
_parameter_constraints: dict = {
|
| 148 |
+
# We don't require `predic_proba` here to allow passing a meta-estimator
|
| 149 |
+
# that only exposes `predict_proba` after fitting.
|
| 150 |
+
"base_estimator": [HasMethods(["fit"])],
|
| 151 |
+
"threshold": [Interval(Real, 0.0, 1.0, closed="left")],
|
| 152 |
+
"criterion": [StrOptions({"threshold", "k_best"})],
|
| 153 |
+
"k_best": [Interval(Integral, 1, None, closed="left")],
|
| 154 |
+
"max_iter": [Interval(Integral, 0, None, closed="left"), None],
|
| 155 |
+
"verbose": ["verbose"],
|
| 156 |
+
}
|
| 157 |
+
|
| 158 |
+
def __init__(
|
| 159 |
+
self,
|
| 160 |
+
base_estimator,
|
| 161 |
+
threshold=0.75,
|
| 162 |
+
criterion="threshold",
|
| 163 |
+
k_best=10,
|
| 164 |
+
max_iter=10,
|
| 165 |
+
verbose=False,
|
| 166 |
+
):
|
| 167 |
+
self.base_estimator = base_estimator
|
| 168 |
+
self.threshold = threshold
|
| 169 |
+
self.criterion = criterion
|
| 170 |
+
self.k_best = k_best
|
| 171 |
+
self.max_iter = max_iter
|
| 172 |
+
self.verbose = verbose
|
| 173 |
+
|
| 174 |
+
def fit(self, X, y):
|
| 175 |
+
"""
|
| 176 |
+
Fit self-training classifier using `X`, `y` as training data.
|
| 177 |
+
|
| 178 |
+
Parameters
|
| 179 |
+
----------
|
| 180 |
+
X : {array-like, sparse matrix} of shape (n_samples, n_features)
|
| 181 |
+
Array representing the data.
|
| 182 |
+
|
| 183 |
+
y : {array-like, sparse matrix} of shape (n_samples,)
|
| 184 |
+
Array representing the labels. Unlabeled samples should have the
|
| 185 |
+
label -1.
|
| 186 |
+
|
| 187 |
+
Returns
|
| 188 |
+
-------
|
| 189 |
+
self : object
|
| 190 |
+
Fitted estimator.
|
| 191 |
+
"""
|
| 192 |
+
self._validate_params()
|
| 193 |
+
|
| 194 |
+
# we need row slicing support for sparce matrices, but costly finiteness check
|
| 195 |
+
# can be delegated to the base estimator.
|
| 196 |
+
X, y = self._validate_data(
|
| 197 |
+
X, y, accept_sparse=["csr", "csc", "lil", "dok"], force_all_finite=False
|
| 198 |
+
)
|
| 199 |
+
|
| 200 |
+
self.base_estimator_ = clone(self.base_estimator)
|
| 201 |
+
|
| 202 |
+
if y.dtype.kind in ["U", "S"]:
|
| 203 |
+
raise ValueError(
|
| 204 |
+
"y has dtype string. If you wish to predict on "
|
| 205 |
+
"string targets, use dtype object, and use -1"
|
| 206 |
+
" as the label for unlabeled samples."
|
| 207 |
+
)
|
| 208 |
+
|
| 209 |
+
has_label = y != -1
|
| 210 |
+
|
| 211 |
+
if np.all(has_label):
|
| 212 |
+
warnings.warn("y contains no unlabeled samples", UserWarning)
|
| 213 |
+
|
| 214 |
+
if self.criterion == "k_best" and (
|
| 215 |
+
self.k_best > X.shape[0] - np.sum(has_label)
|
| 216 |
+
):
|
| 217 |
+
warnings.warn(
|
| 218 |
+
"k_best is larger than the amount of unlabeled "
|
| 219 |
+
"samples. All unlabeled samples will be labeled in "
|
| 220 |
+
"the first iteration",
|
| 221 |
+
UserWarning,
|
| 222 |
+
)
|
| 223 |
+
|
| 224 |
+
self.transduction_ = np.copy(y)
|
| 225 |
+
self.labeled_iter_ = np.full_like(y, -1)
|
| 226 |
+
self.labeled_iter_[has_label] = 0
|
| 227 |
+
|
| 228 |
+
self.n_iter_ = 0
|
| 229 |
+
|
| 230 |
+
while not np.all(has_label) and (
|
| 231 |
+
self.max_iter is None or self.n_iter_ < self.max_iter
|
| 232 |
+
):
|
| 233 |
+
self.n_iter_ += 1
|
| 234 |
+
self.base_estimator_.fit(
|
| 235 |
+
X[safe_mask(X, has_label)], self.transduction_[has_label]
|
| 236 |
+
)
|
| 237 |
+
|
| 238 |
+
# Predict on the unlabeled samples
|
| 239 |
+
prob = self.base_estimator_.predict_proba(X[safe_mask(X, ~has_label)])
|
| 240 |
+
pred = self.base_estimator_.classes_[np.argmax(prob, axis=1)]
|
| 241 |
+
max_proba = np.max(prob, axis=1)
|
| 242 |
+
|
| 243 |
+
# Select new labeled samples
|
| 244 |
+
if self.criterion == "threshold":
|
| 245 |
+
selected = max_proba > self.threshold
|
| 246 |
+
else:
|
| 247 |
+
n_to_select = min(self.k_best, max_proba.shape[0])
|
| 248 |
+
if n_to_select == max_proba.shape[0]:
|
| 249 |
+
selected = np.ones_like(max_proba, dtype=bool)
|
| 250 |
+
else:
|
| 251 |
+
# NB these are indices, not a mask
|
| 252 |
+
selected = np.argpartition(-max_proba, n_to_select)[:n_to_select]
|
| 253 |
+
|
| 254 |
+
# Map selected indices into original array
|
| 255 |
+
selected_full = np.nonzero(~has_label)[0][selected]
|
| 256 |
+
|
| 257 |
+
# Add newly labeled confident predictions to the dataset
|
| 258 |
+
self.transduction_[selected_full] = pred[selected]
|
| 259 |
+
has_label[selected_full] = True
|
| 260 |
+
self.labeled_iter_[selected_full] = self.n_iter_
|
| 261 |
+
|
| 262 |
+
if selected_full.shape[0] == 0:
|
| 263 |
+
# no changed labels
|
| 264 |
+
self.termination_condition_ = "no_change"
|
| 265 |
+
break
|
| 266 |
+
|
| 267 |
+
if self.verbose:
|
| 268 |
+
print(
|
| 269 |
+
f"End of iteration {self.n_iter_},"
|
| 270 |
+
f" added {selected_full.shape[0]} new labels."
|
| 271 |
+
)
|
| 272 |
+
|
| 273 |
+
if self.n_iter_ == self.max_iter:
|
| 274 |
+
self.termination_condition_ = "max_iter"
|
| 275 |
+
if np.all(has_label):
|
| 276 |
+
self.termination_condition_ = "all_labeled"
|
| 277 |
+
|
| 278 |
+
self.base_estimator_.fit(
|
| 279 |
+
X[safe_mask(X, has_label)], self.transduction_[has_label]
|
| 280 |
+
)
|
| 281 |
+
self.classes_ = self.base_estimator_.classes_
|
| 282 |
+
return self
|
| 283 |
+
|
| 284 |
+
@available_if(_estimator_has("predict"))
|
| 285 |
+
def predict(self, X):
|
| 286 |
+
"""Predict the classes of `X`.
|
| 287 |
+
|
| 288 |
+
Parameters
|
| 289 |
+
----------
|
| 290 |
+
X : {array-like, sparse matrix} of shape (n_samples, n_features)
|
| 291 |
+
Array representing the data.
|
| 292 |
+
|
| 293 |
+
Returns
|
| 294 |
+
-------
|
| 295 |
+
y : ndarray of shape (n_samples,)
|
| 296 |
+
Array with predicted labels.
|
| 297 |
+
"""
|
| 298 |
+
check_is_fitted(self)
|
| 299 |
+
X = self._validate_data(
|
| 300 |
+
X,
|
| 301 |
+
accept_sparse=True,
|
| 302 |
+
force_all_finite=False,
|
| 303 |
+
reset=False,
|
| 304 |
+
)
|
| 305 |
+
return self.base_estimator_.predict(X)
|
| 306 |
+
|
| 307 |
+
@available_if(_estimator_has("predict_proba"))
|
| 308 |
+
def predict_proba(self, X):
|
| 309 |
+
"""Predict probability for each possible outcome.
|
| 310 |
+
|
| 311 |
+
Parameters
|
| 312 |
+
----------
|
| 313 |
+
X : {array-like, sparse matrix} of shape (n_samples, n_features)
|
| 314 |
+
Array representing the data.
|
| 315 |
+
|
| 316 |
+
Returns
|
| 317 |
+
-------
|
| 318 |
+
y : ndarray of shape (n_samples, n_features)
|
| 319 |
+
Array with prediction probabilities.
|
| 320 |
+
"""
|
| 321 |
+
check_is_fitted(self)
|
| 322 |
+
X = self._validate_data(
|
| 323 |
+
X,
|
| 324 |
+
accept_sparse=True,
|
| 325 |
+
force_all_finite=False,
|
| 326 |
+
reset=False,
|
| 327 |
+
)
|
| 328 |
+
return self.base_estimator_.predict_proba(X)
|
| 329 |
+
|
| 330 |
+
@available_if(_estimator_has("decision_function"))
|
| 331 |
+
def decision_function(self, X):
|
| 332 |
+
"""Call decision function of the `base_estimator`.
|
| 333 |
+
|
| 334 |
+
Parameters
|
| 335 |
+
----------
|
| 336 |
+
X : {array-like, sparse matrix} of shape (n_samples, n_features)
|
| 337 |
+
Array representing the data.
|
| 338 |
+
|
| 339 |
+
Returns
|
| 340 |
+
-------
|
| 341 |
+
y : ndarray of shape (n_samples, n_features)
|
| 342 |
+
Result of the decision function of the `base_estimator`.
|
| 343 |
+
"""
|
| 344 |
+
check_is_fitted(self)
|
| 345 |
+
X = self._validate_data(
|
| 346 |
+
X,
|
| 347 |
+
accept_sparse=True,
|
| 348 |
+
force_all_finite=False,
|
| 349 |
+
reset=False,
|
| 350 |
+
)
|
| 351 |
+
return self.base_estimator_.decision_function(X)
|
| 352 |
+
|
| 353 |
+
@available_if(_estimator_has("predict_log_proba"))
|
| 354 |
+
def predict_log_proba(self, X):
|
| 355 |
+
"""Predict log probability for each possible outcome.
|
| 356 |
+
|
| 357 |
+
Parameters
|
| 358 |
+
----------
|
| 359 |
+
X : {array-like, sparse matrix} of shape (n_samples, n_features)
|
| 360 |
+
Array representing the data.
|
| 361 |
+
|
| 362 |
+
Returns
|
| 363 |
+
-------
|
| 364 |
+
y : ndarray of shape (n_samples, n_features)
|
| 365 |
+
Array with log prediction probabilities.
|
| 366 |
+
"""
|
| 367 |
+
check_is_fitted(self)
|
| 368 |
+
X = self._validate_data(
|
| 369 |
+
X,
|
| 370 |
+
accept_sparse=True,
|
| 371 |
+
force_all_finite=False,
|
| 372 |
+
reset=False,
|
| 373 |
+
)
|
| 374 |
+
return self.base_estimator_.predict_log_proba(X)
|
| 375 |
+
|
| 376 |
+
@available_if(_estimator_has("score"))
|
| 377 |
+
def score(self, X, y):
|
| 378 |
+
"""Call score on the `base_estimator`.
|
| 379 |
+
|
| 380 |
+
Parameters
|
| 381 |
+
----------
|
| 382 |
+
X : {array-like, sparse matrix} of shape (n_samples, n_features)
|
| 383 |
+
Array representing the data.
|
| 384 |
+
|
| 385 |
+
y : array-like of shape (n_samples,)
|
| 386 |
+
Array representing the labels.
|
| 387 |
+
|
| 388 |
+
Returns
|
| 389 |
+
-------
|
| 390 |
+
score : float
|
| 391 |
+
Result of calling score on the `base_estimator`.
|
| 392 |
+
"""
|
| 393 |
+
check_is_fitted(self)
|
| 394 |
+
X = self._validate_data(
|
| 395 |
+
X,
|
| 396 |
+
accept_sparse=True,
|
| 397 |
+
force_all_finite=False,
|
| 398 |
+
reset=False,
|
| 399 |
+
)
|
| 400 |
+
return self.base_estimator_.score(X, y)
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__init__.py
ADDED
|
File without changes
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__pycache__/__init__.cpython-310.pyc
ADDED
|
Binary file (186 Bytes). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__pycache__/test_label_propagation.cpython-310.pyc
ADDED
|
Binary file (6.95 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/__pycache__/test_self_training.cpython-310.pyc
ADDED
|
Binary file (7.97 kB). View file
|
|
|
videollama2/lib/python3.10/site-packages/sklearn/semi_supervised/tests/test_label_propagation.py
ADDED
|
@@ -0,0 +1,216 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
""" test the label propagation module """
|
| 2 |
+
|
| 3 |
+
import numpy as np
|
| 4 |
+
import pytest
|
| 5 |
+
import warnings
|
| 6 |
+
|
| 7 |
+
from scipy.sparse import issparse
|
| 8 |
+
from sklearn.semi_supervised import _label_propagation as label_propagation
|
| 9 |
+
from sklearn.metrics.pairwise import rbf_kernel
|
| 10 |
+
from sklearn.model_selection import train_test_split
|
| 11 |
+
from sklearn.neighbors import NearestNeighbors
|
| 12 |
+
from sklearn.datasets import make_classification
|
| 13 |
+
from sklearn.exceptions import ConvergenceWarning
|
| 14 |
+
from sklearn.utils._testing import (
|
| 15 |
+
assert_allclose,
|
| 16 |
+
assert_array_equal,
|
| 17 |
+
)
|
| 18 |
+
|
| 19 |
+
ESTIMATORS = [
|
| 20 |
+
(label_propagation.LabelPropagation, {"kernel": "rbf"}),
|
| 21 |
+
(label_propagation.LabelPropagation, {"kernel": "knn", "n_neighbors": 2}),
|
| 22 |
+
(
|
| 23 |
+
label_propagation.LabelPropagation,
|
| 24 |
+
{"kernel": lambda x, y: rbf_kernel(x, y, gamma=20)},
|
| 25 |
+
),
|
| 26 |
+
(label_propagation.LabelSpreading, {"kernel": "rbf"}),
|
| 27 |
+
(label_propagation.LabelSpreading, {"kernel": "knn", "n_neighbors": 2}),
|
| 28 |
+
(
|
| 29 |
+
label_propagation.LabelSpreading,
|
| 30 |
+
{"kernel": lambda x, y: rbf_kernel(x, y, gamma=20)},
|
| 31 |
+
),
|
| 32 |
+
]
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
@pytest.mark.parametrize("Estimator, parameters", ESTIMATORS)
|
| 36 |
+
def test_fit_transduction(global_dtype, Estimator, parameters):
|
| 37 |
+
samples = np.asarray([[1.0, 0.0], [0.0, 2.0], [1.0, 3.0]], dtype=global_dtype)
|
| 38 |
+
labels = [0, 1, -1]
|
| 39 |
+
clf = Estimator(**parameters).fit(samples, labels)
|
| 40 |
+
assert clf.transduction_[2] == 1
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
@pytest.mark.parametrize("Estimator, parameters", ESTIMATORS)
|
| 44 |
+
def test_distribution(global_dtype, Estimator, parameters):
|
| 45 |
+
if parameters["kernel"] == "knn":
|
| 46 |
+
pytest.skip(
|
| 47 |
+
"Unstable test for this configuration: changes in k-NN ordering break it."
|
| 48 |
+
)
|
| 49 |
+
samples = np.asarray([[1.0, 0.0], [0.0, 1.0], [1.0, 1.0]], dtype=global_dtype)
|
| 50 |
+
labels = [0, 1, -1]
|
| 51 |
+
clf = Estimator(**parameters).fit(samples, labels)
|
| 52 |
+
assert_allclose(clf.label_distributions_[2], [0.5, 0.5], atol=1e-2)
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
@pytest.mark.parametrize("Estimator, parameters", ESTIMATORS)
|
| 56 |
+
def test_predict(global_dtype, Estimator, parameters):
|
| 57 |
+
samples = np.asarray([[1.0, 0.0], [0.0, 2.0], [1.0, 3.0]], dtype=global_dtype)
|
| 58 |
+
labels = [0, 1, -1]
|
| 59 |
+
clf = Estimator(**parameters).fit(samples, labels)
|
| 60 |
+
assert_array_equal(clf.predict([[0.5, 2.5]]), np.array([1]))
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
@pytest.mark.parametrize("Estimator, parameters", ESTIMATORS)
|
| 64 |
+
def test_predict_proba(global_dtype, Estimator, parameters):
|
| 65 |
+
samples = np.asarray([[1.0, 0.0], [0.0, 1.0], [1.0, 2.5]], dtype=global_dtype)
|
| 66 |
+
labels = [0, 1, -1]
|
| 67 |
+
clf = Estimator(**parameters).fit(samples, labels)
|
| 68 |
+
assert_allclose(clf.predict_proba([[1.0, 1.0]]), np.array([[0.5, 0.5]]))
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
@pytest.mark.parametrize("alpha", [0.1, 0.3, 0.5, 0.7, 0.9])
|
| 72 |
+
@pytest.mark.parametrize("Estimator, parameters", ESTIMATORS)
|
| 73 |
+
def test_label_spreading_closed_form(global_dtype, Estimator, parameters, alpha):
|
| 74 |
+
n_classes = 2
|
| 75 |
+
X, y = make_classification(n_classes=n_classes, n_samples=200, random_state=0)
|
| 76 |
+
X = X.astype(global_dtype, copy=False)
|
| 77 |
+
y[::3] = -1
|
| 78 |
+
|
| 79 |
+
gamma = 0.1
|
| 80 |
+
clf = label_propagation.LabelSpreading(gamma=gamma).fit(X, y)
|
| 81 |
+
# adopting notation from Zhou et al (2004):
|
| 82 |
+
S = clf._build_graph()
|
| 83 |
+
Y = np.zeros((len(y), n_classes + 1), dtype=X.dtype)
|
| 84 |
+
Y[np.arange(len(y)), y] = 1
|
| 85 |
+
Y = Y[:, :-1]
|
| 86 |
+
|
| 87 |
+
expected = np.dot(np.linalg.inv(np.eye(len(S), dtype=S.dtype) - alpha * S), Y)
|
| 88 |
+
expected /= expected.sum(axis=1)[:, np.newaxis]
|
| 89 |
+
|
| 90 |
+
clf = label_propagation.LabelSpreading(
|
| 91 |
+
max_iter=100, alpha=alpha, tol=1e-10, gamma=gamma
|
| 92 |
+
)
|
| 93 |
+
clf.fit(X, y)
|
| 94 |
+
|
| 95 |
+
assert_allclose(expected, clf.label_distributions_)
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
def test_label_propagation_closed_form(global_dtype):
|
| 99 |
+
n_classes = 2
|
| 100 |
+
X, y = make_classification(n_classes=n_classes, n_samples=200, random_state=0)
|
| 101 |
+
X = X.astype(global_dtype, copy=False)
|
| 102 |
+
y[::3] = -1
|
| 103 |
+
Y = np.zeros((len(y), n_classes + 1))
|
| 104 |
+
Y[np.arange(len(y)), y] = 1
|
| 105 |
+
unlabelled_idx = Y[:, (-1,)].nonzero()[0]
|
| 106 |
+
labelled_idx = (Y[:, (-1,)] == 0).nonzero()[0]
|
| 107 |
+
|
| 108 |
+
clf = label_propagation.LabelPropagation(max_iter=100, tol=1e-10, gamma=0.1)
|
| 109 |
+
clf.fit(X, y)
|
| 110 |
+
# adopting notation from Zhu et al 2002
|
| 111 |
+
T_bar = clf._build_graph()
|
| 112 |
+
Tuu = T_bar[tuple(np.meshgrid(unlabelled_idx, unlabelled_idx, indexing="ij"))]
|
| 113 |
+
Tul = T_bar[tuple(np.meshgrid(unlabelled_idx, labelled_idx, indexing="ij"))]
|
| 114 |
+
Y = Y[:, :-1]
|
| 115 |
+
Y_l = Y[labelled_idx, :]
|
| 116 |
+
Y_u = np.dot(np.dot(np.linalg.inv(np.eye(Tuu.shape[0]) - Tuu), Tul), Y_l)
|
| 117 |
+
|
| 118 |
+
expected = Y.copy()
|
| 119 |
+
expected[unlabelled_idx, :] = Y_u
|
| 120 |
+
expected /= expected.sum(axis=1)[:, np.newaxis]
|
| 121 |
+
|
| 122 |
+
assert_allclose(expected, clf.label_distributions_, atol=1e-4)
|
| 123 |
+
|
| 124 |
+
|
| 125 |
+
def test_convergence_speed():
|
| 126 |
+
# This is a non-regression test for #5774
|
| 127 |
+
X = np.array([[1.0, 0.0], [0.0, 1.0], [1.0, 2.5]])
|
| 128 |
+
y = np.array([0, 1, -1])
|
| 129 |
+
mdl = label_propagation.LabelSpreading(kernel="rbf", max_iter=5000)
|
| 130 |
+
mdl.fit(X, y)
|
| 131 |
+
|
| 132 |
+
# this should converge quickly:
|
| 133 |
+
assert mdl.n_iter_ < 10
|
| 134 |
+
assert_array_equal(mdl.predict(X), [0, 1, 1])
|
| 135 |
+
|
| 136 |
+
|
| 137 |
+
def test_convergence_warning():
|
| 138 |
+
# This is a non-regression test for #5774
|
| 139 |
+
X = np.array([[1.0, 0.0], [0.0, 1.0], [1.0, 2.5]])
|
| 140 |
+
y = np.array([0, 1, -1])
|
| 141 |
+
mdl = label_propagation.LabelSpreading(kernel="rbf", max_iter=1)
|
| 142 |
+
warn_msg = "max_iter=1 was reached without convergence."
|
| 143 |
+
with pytest.warns(ConvergenceWarning, match=warn_msg):
|
| 144 |
+
mdl.fit(X, y)
|
| 145 |
+
assert mdl.n_iter_ == mdl.max_iter
|
| 146 |
+
|
| 147 |
+
mdl = label_propagation.LabelPropagation(kernel="rbf", max_iter=1)
|
| 148 |
+
with pytest.warns(ConvergenceWarning, match=warn_msg):
|
| 149 |
+
mdl.fit(X, y)
|
| 150 |
+
assert mdl.n_iter_ == mdl.max_iter
|
| 151 |
+
|
| 152 |
+
mdl = label_propagation.LabelSpreading(kernel="rbf", max_iter=500)
|
| 153 |
+
with warnings.catch_warnings():
|
| 154 |
+
warnings.simplefilter("error", ConvergenceWarning)
|
| 155 |
+
mdl.fit(X, y)
|
| 156 |
+
|
| 157 |
+
mdl = label_propagation.LabelPropagation(kernel="rbf", max_iter=500)
|
| 158 |
+
with warnings.catch_warnings():
|
| 159 |
+
warnings.simplefilter("error", ConvergenceWarning)
|
| 160 |
+
mdl.fit(X, y)
|
| 161 |
+
|
| 162 |
+
|
| 163 |
+
@pytest.mark.parametrize(
|
| 164 |
+
"LabelPropagationCls",
|
| 165 |
+
[label_propagation.LabelSpreading, label_propagation.LabelPropagation],
|
| 166 |
+
)
|
| 167 |
+
def test_label_propagation_non_zero_normalizer(LabelPropagationCls):
|
| 168 |
+
# check that we don't divide by zero in case of null normalizer
|
| 169 |
+
# non-regression test for
|
| 170 |
+
# https://github.com/scikit-learn/scikit-learn/pull/15946
|
| 171 |
+
# https://github.com/scikit-learn/scikit-learn/issues/9292
|
| 172 |
+
X = np.array([[100.0, 100.0], [100.0, 100.0], [0.0, 0.0], [0.0, 0.0]])
|
| 173 |
+
y = np.array([0, 1, -1, -1])
|
| 174 |
+
mdl = LabelPropagationCls(kernel="knn", max_iter=100, n_neighbors=1)
|
| 175 |
+
with warnings.catch_warnings():
|
| 176 |
+
warnings.simplefilter("error", RuntimeWarning)
|
| 177 |
+
mdl.fit(X, y)
|
| 178 |
+
|
| 179 |
+
|
| 180 |
+
def test_predict_sparse_callable_kernel(global_dtype):
|
| 181 |
+
# This is a non-regression test for #15866
|
| 182 |
+
|
| 183 |
+
# Custom sparse kernel (top-K RBF)
|
| 184 |
+
def topk_rbf(X, Y=None, n_neighbors=10, gamma=1e-5):
|
| 185 |
+
nn = NearestNeighbors(n_neighbors=10, metric="euclidean", n_jobs=2)
|
| 186 |
+
nn.fit(X)
|
| 187 |
+
W = -1 * nn.kneighbors_graph(Y, mode="distance").power(2) * gamma
|
| 188 |
+
np.exp(W.data, out=W.data)
|
| 189 |
+
assert issparse(W)
|
| 190 |
+
return W.T
|
| 191 |
+
|
| 192 |
+
n_classes = 4
|
| 193 |
+
n_samples = 500
|
| 194 |
+
n_test = 10
|
| 195 |
+
X, y = make_classification(
|
| 196 |
+
n_classes=n_classes,
|
| 197 |
+
n_samples=n_samples,
|
| 198 |
+
n_features=20,
|
| 199 |
+
n_informative=20,
|
| 200 |
+
n_redundant=0,
|
| 201 |
+
n_repeated=0,
|
| 202 |
+
random_state=0,
|
| 203 |
+
)
|
| 204 |
+
X = X.astype(global_dtype)
|
| 205 |
+
|
| 206 |
+
X_train, X_test, y_train, y_test = train_test_split(
|
| 207 |
+
X, y, test_size=n_test, random_state=0
|
| 208 |
+
)
|
| 209 |
+
|
| 210 |
+
model = label_propagation.LabelSpreading(kernel=topk_rbf)
|
| 211 |
+
model.fit(X_train, y_train)
|
| 212 |
+
assert model.score(X_test, y_test) >= 0.9
|
| 213 |
+
|
| 214 |
+
model = label_propagation.LabelPropagation(kernel=topk_rbf)
|
| 215 |
+
model.fit(X_train, y_train)
|
| 216 |
+
assert model.score(X_test, y_test) >= 0.9
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_compute_linear_combination_native.h
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API at::Tensor _compute_linear_combination(const at::Tensor & input, const at::Tensor & coefficients);
|
| 20 |
+
TORCH_API at::Tensor & _compute_linear_combination_out(const at::Tensor & input, const at::Tensor & coefficients, at::Tensor & out);
|
| 21 |
+
} // namespace native
|
| 22 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_cudnn_rnn_backward_native.h
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API void _cudnn_rnn_backward_out_symint(const at::Tensor & input, at::TensorList weight, int64_t weight_stride0, const at::Tensor & weight_buf, const at::Tensor & hx, const ::std::optional<at::Tensor> & cx, const at::Tensor & output, const ::std::optional<at::Tensor> & grad_output, const ::std::optional<at::Tensor> & grad_hy, const ::std::optional<at::Tensor> & grad_cy, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, double dropout, bool train, bool bidirectional, c10::SymIntArrayRef batch_sizes, const ::std::optional<at::Tensor> & dropout_state, const at::Tensor & reserve, ::std::array<bool,4> output_mask, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::TensorList out3);
|
| 20 |
+
TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,::std::vector<at::Tensor>> _cudnn_rnn_backward(const at::Tensor & input, at::TensorList weight, int64_t weight_stride0, const at::Tensor & weight_buf, const at::Tensor & hx, const ::std::optional<at::Tensor> & cx, const at::Tensor & output, const ::std::optional<at::Tensor> & grad_output, const ::std::optional<at::Tensor> & grad_hy, const ::std::optional<at::Tensor> & grad_cy, int64_t mode, int64_t hidden_size, int64_t proj_size, int64_t num_layers, bool batch_first, double dropout, bool train, bool bidirectional, at::IntArrayRef batch_sizes, const ::std::optional<at::Tensor> & dropout_state, const at::Tensor & reserve, ::std::array<bool,4> output_mask);
|
| 21 |
+
} // namespace native
|
| 22 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_empty_per_channel_affine_quantized_ops.h
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Operator.h
|
| 4 |
+
|
| 5 |
+
#include <tuple>
|
| 6 |
+
#include <vector>
|
| 7 |
+
|
| 8 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 9 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 10 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 11 |
+
#include <ATen/core/ATen_fwd.h>
|
| 12 |
+
|
| 13 |
+
namespace at {
|
| 14 |
+
namespace _ops {
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
struct TORCH_API _empty_per_channel_affine_quantized {
|
| 18 |
+
using schema = at::Tensor (c10::SymIntArrayRef, const at::Tensor &, const at::Tensor &, int64_t, ::std::optional<at::ScalarType>, ::std::optional<at::Layout>, ::std::optional<at::Device>, ::std::optional<bool>, ::std::optional<at::MemoryFormat>);
|
| 19 |
+
using ptr_schema = schema*;
|
| 20 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 21 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_empty_per_channel_affine_quantized")
|
| 22 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
|
| 23 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_empty_per_channel_affine_quantized(SymInt[] size, *, Tensor scales, Tensor zero_points, int axis, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None, MemoryFormat? memory_format=contiguous_format) -> Tensor")
|
| 24 |
+
static at::Tensor call(c10::SymIntArrayRef size, const at::Tensor & scales, const at::Tensor & zero_points, int64_t axis, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory, ::std::optional<at::MemoryFormat> memory_format);
|
| 25 |
+
static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, c10::SymIntArrayRef size, const at::Tensor & scales, const at::Tensor & zero_points, int64_t axis, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory, ::std::optional<at::MemoryFormat> memory_format);
|
| 26 |
+
};
|
| 27 |
+
|
| 28 |
+
struct TORCH_API _empty_per_channel_affine_quantized_out {
|
| 29 |
+
using schema = at::Tensor & (c10::SymIntArrayRef, const at::Tensor &, const at::Tensor &, int64_t, ::std::optional<at::MemoryFormat>, at::Tensor &);
|
| 30 |
+
using ptr_schema = schema*;
|
| 31 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 32 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_empty_per_channel_affine_quantized")
|
| 33 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
|
| 34 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_empty_per_channel_affine_quantized.out(SymInt[] size, *, Tensor scales, Tensor zero_points, int axis, MemoryFormat? memory_format=contiguous_format, Tensor(a!) out) -> Tensor(a!)")
|
| 35 |
+
static at::Tensor & call(c10::SymIntArrayRef size, const at::Tensor & scales, const at::Tensor & zero_points, int64_t axis, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out);
|
| 36 |
+
static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, c10::SymIntArrayRef size, const at::Tensor & scales, const at::Tensor & zero_points, int64_t axis, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out);
|
| 37 |
+
};
|
| 38 |
+
|
| 39 |
+
}} // namespace at::_ops
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_floor_compositeexplicitautograd_dispatch.h
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace compositeexplicitautograd {
|
| 19 |
+
|
| 20 |
+
TORCH_API ::std::vector<at::Tensor> _foreach_floor(at::TensorList self);
|
| 21 |
+
TORCH_API void _foreach_floor_out(at::TensorList out, at::TensorList self);
|
| 22 |
+
TORCH_API void _foreach_floor_outf(at::TensorList self, at::TensorList out);
|
| 23 |
+
TORCH_API void _foreach_floor_(at::TensorList self);
|
| 24 |
+
|
| 25 |
+
} // namespace compositeexplicitautograd
|
| 26 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_lerp_native.h
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API ::std::vector<at::Tensor> foreach_tensor_ternary_lerp_slow(at::TensorList self, at::TensorList tensors1, at::TensorList weights);
|
| 20 |
+
TORCH_API void _foreach_lerp_List_out(at::TensorList self, at::TensorList tensors1, at::TensorList weights, at::TensorList out);
|
| 21 |
+
TORCH_API void foreach_tensor_ternary_lerp_slow_(at::TensorList self, at::TensorList tensors1, at::TensorList weights);
|
| 22 |
+
TORCH_API ::std::vector<at::Tensor> foreach_tensor_lerp_ternary_cuda(at::TensorList self, at::TensorList tensors1, at::TensorList weights);
|
| 23 |
+
TORCH_API void foreach_tensor_lerp_ternary_cuda_(at::TensorList self, at::TensorList tensors1, at::TensorList weights);
|
| 24 |
+
TORCH_API ::std::vector<at::Tensor> foreach_tensor_lerp_list_kernel_slow(at::TensorList self, at::TensorList tensors1, const at::Scalar & weight);
|
| 25 |
+
TORCH_API void _foreach_lerp_Scalar_out(at::TensorList self, at::TensorList tensors1, const at::Scalar & weight, at::TensorList out);
|
| 26 |
+
TORCH_API void foreach_tensor_lerp_list_kernel_slow_(at::TensorList self, at::TensorList tensors1, const at::Scalar & weight);
|
| 27 |
+
TORCH_API ::std::vector<at::Tensor> foreach_tensor_lerp_list_cuda(at::TensorList self, at::TensorList tensors1, const at::Scalar & weight);
|
| 28 |
+
TORCH_API void foreach_tensor_lerp_list_cuda_(at::TensorList self, at::TensorList tensors1, const at::Scalar & weight);
|
| 29 |
+
} // namespace native
|
| 30 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_reciprocal_cuda_dispatch.h
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace cuda {
|
| 19 |
+
|
| 20 |
+
TORCH_API ::std::vector<at::Tensor> _foreach_reciprocal(at::TensorList self);
|
| 21 |
+
TORCH_API void _foreach_reciprocal_(at::TensorList self);
|
| 22 |
+
|
| 23 |
+
} // namespace cuda
|
| 24 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_grid_sampler_2d_cpu_fallback.h
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Function.h
|
| 4 |
+
|
| 5 |
+
#include <ATen/Context.h>
|
| 6 |
+
#include <ATen/DeviceGuard.h>
|
| 7 |
+
#include <ATen/TensorUtils.h>
|
| 8 |
+
#include <ATen/TracerMode.h>
|
| 9 |
+
#include <ATen/core/Generator.h>
|
| 10 |
+
#include <ATen/core/Reduction.h>
|
| 11 |
+
#include <ATen/core/Tensor.h>
|
| 12 |
+
#include <c10/core/Scalar.h>
|
| 13 |
+
#include <c10/core/Storage.h>
|
| 14 |
+
#include <c10/core/TensorOptions.h>
|
| 15 |
+
#include <c10/util/Deprecated.h>
|
| 16 |
+
#include <optional>
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
#include <ATen/ops/_grid_sampler_2d_cpu_fallback_ops.h>
|
| 21 |
+
|
| 22 |
+
namespace at {
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
// aten::_grid_sampler_2d_cpu_fallback(Tensor input, Tensor grid, int interpolation_mode, int padding_mode, bool align_corners) -> Tensor
|
| 26 |
+
inline at::Tensor _grid_sampler_2d_cpu_fallback(const at::Tensor & input, const at::Tensor & grid, int64_t interpolation_mode, int64_t padding_mode, bool align_corners) {
|
| 27 |
+
return at::_ops::_grid_sampler_2d_cpu_fallback::call(input, grid, interpolation_mode, padding_mode, align_corners);
|
| 28 |
+
}
|
| 29 |
+
|
| 30 |
+
// aten::_grid_sampler_2d_cpu_fallback.out(Tensor input, Tensor grid, int interpolation_mode, int padding_mode, bool align_corners, *, Tensor(a!) out) -> Tensor(a!)
|
| 31 |
+
inline at::Tensor & _grid_sampler_2d_cpu_fallback_out(at::Tensor & out, const at::Tensor & input, const at::Tensor & grid, int64_t interpolation_mode, int64_t padding_mode, bool align_corners) {
|
| 32 |
+
return at::_ops::_grid_sampler_2d_cpu_fallback_out::call(input, grid, interpolation_mode, padding_mode, align_corners, out);
|
| 33 |
+
}
|
| 34 |
+
// aten::_grid_sampler_2d_cpu_fallback.out(Tensor input, Tensor grid, int interpolation_mode, int padding_mode, bool align_corners, *, Tensor(a!) out) -> Tensor(a!)
|
| 35 |
+
inline at::Tensor & _grid_sampler_2d_cpu_fallback_outf(const at::Tensor & input, const at::Tensor & grid, int64_t interpolation_mode, int64_t padding_mode, bool align_corners, at::Tensor & out) {
|
| 36 |
+
return at::_ops::_grid_sampler_2d_cpu_fallback_out::call(input, grid, interpolation_mode, padding_mode, align_corners, out);
|
| 37 |
+
}
|
| 38 |
+
|
| 39 |
+
}
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_lstm_mps_ops.h
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Operator.h
|
| 4 |
+
|
| 5 |
+
#include <tuple>
|
| 6 |
+
#include <vector>
|
| 7 |
+
|
| 8 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 9 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 10 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 11 |
+
#include <ATen/core/ATen_fwd.h>
|
| 12 |
+
|
| 13 |
+
namespace at {
|
| 14 |
+
namespace _ops {
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
struct TORCH_API _lstm_mps {
|
| 18 |
+
using schema = ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> (const at::Tensor &, at::TensorList, at::TensorList, bool, int64_t, double, bool, bool, bool);
|
| 19 |
+
using ptr_schema = schema*;
|
| 20 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 21 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_lstm_mps")
|
| 22 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
|
| 23 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_lstm_mps(Tensor input, Tensor[] hx, Tensor[] params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional, bool batch_first) -> (Tensor, Tensor, Tensor, Tensor, Tensor, Tensor)")
|
| 24 |
+
static ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> call(const at::Tensor & input, at::TensorList hx, at::TensorList params, bool has_biases, int64_t num_layers, double dropout, bool train, bool bidirectional, bool batch_first);
|
| 25 |
+
static ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & input, at::TensorList hx, at::TensorList params, bool has_biases, int64_t num_layers, double dropout, bool train, bool bidirectional, bool batch_first);
|
| 26 |
+
};
|
| 27 |
+
|
| 28 |
+
struct TORCH_API _lstm_mps_out {
|
| 29 |
+
using schema = ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> (const at::Tensor &, at::TensorList, at::TensorList, bool, int64_t, double, bool, bool, bool, at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &);
|
| 30 |
+
using ptr_schema = schema*;
|
| 31 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 32 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_lstm_mps")
|
| 33 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
|
| 34 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_lstm_mps.out(Tensor input, Tensor[] hx, Tensor[] params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional, bool batch_first, *, Tensor(a!) out0, Tensor(b!) out1, Tensor(c!) out2, Tensor(d!) out3, Tensor(e!) out4, Tensor(f!) out5) -> (Tensor(a!), Tensor(b!), Tensor(c!), Tensor(d!), Tensor(e!), Tensor(f!))")
|
| 35 |
+
static ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> call(const at::Tensor & input, at::TensorList hx, at::TensorList params, bool has_biases, int64_t num_layers, double dropout, bool train, bool bidirectional, bool batch_first, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3, at::Tensor & out4, at::Tensor & out5);
|
| 36 |
+
static ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & input, at::TensorList hx, at::TensorList params, bool has_biases, int64_t num_layers, double dropout, bool train, bool bidirectional, bool batch_first, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3, at::Tensor & out4, at::Tensor & out5);
|
| 37 |
+
};
|
| 38 |
+
|
| 39 |
+
}} // namespace at::_ops
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_make_per_channel_quantized_tensor_cpu_dispatch.h
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace cpu {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor _make_per_channel_quantized_tensor(const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, int64_t axis);
|
| 21 |
+
|
| 22 |
+
} // namespace cpu
|
| 23 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_compute_contiguous_strides_offsets_native.h
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API ::std::tuple<at::Tensor,at::Tensor> _nested_compute_contiguous_strides_offsets(const at::Tensor & nested_size);
|
| 20 |
+
} // namespace native
|
| 21 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_get_values_copy_compositeexplicitautogradnonfunctional_dispatch.h
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace compositeexplicitautogradnonfunctional {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor _nested_get_values_copy(const at::Tensor & self);
|
| 21 |
+
|
| 22 |
+
} // namespace compositeexplicitautogradnonfunctional
|
| 23 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_tensor_from_mask_native.h
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API at::Tensor & _nested_tensor_from_mask_out(const at::Tensor & t, const at::Tensor & mask, bool mask_check, at::Tensor & out);
|
| 20 |
+
TORCH_API at::Tensor NestedTensor_nested_tensor_from_mask(const at::Tensor & t, const at::Tensor & mask, bool mask_check=true);
|
| 21 |
+
} // namespace native
|
| 22 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_nested_tensor_strides_compositeexplicitautograd_dispatch.h
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace compositeexplicitautograd {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor & _nested_tensor_strides_out(at::Tensor & out, const at::Tensor & self);
|
| 21 |
+
TORCH_API at::Tensor & _nested_tensor_strides_outf(const at::Tensor & self, at::Tensor & out);
|
| 22 |
+
|
| 23 |
+
} // namespace compositeexplicitautograd
|
| 24 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_prelu_kernel_ops.h
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Operator.h
|
| 4 |
+
|
| 5 |
+
#include <tuple>
|
| 6 |
+
#include <vector>
|
| 7 |
+
|
| 8 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 9 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 10 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 11 |
+
#include <ATen/core/ATen_fwd.h>
|
| 12 |
+
|
| 13 |
+
namespace at {
|
| 14 |
+
namespace _ops {
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
struct TORCH_API _prelu_kernel {
|
| 18 |
+
using schema = at::Tensor (const at::Tensor &, const at::Tensor &);
|
| 19 |
+
using ptr_schema = schema*;
|
| 20 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 21 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_prelu_kernel")
|
| 22 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
|
| 23 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_prelu_kernel(Tensor self, Tensor weight) -> Tensor")
|
| 24 |
+
static at::Tensor call(const at::Tensor & self, const at::Tensor & weight);
|
| 25 |
+
static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, const at::Tensor & weight);
|
| 26 |
+
};
|
| 27 |
+
|
| 28 |
+
}} // namespace at::_ops
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_scaled_dot_product_flash_attention_backward.h
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Function.h
|
| 4 |
+
|
| 5 |
+
#include <ATen/Context.h>
|
| 6 |
+
#include <ATen/DeviceGuard.h>
|
| 7 |
+
#include <ATen/TensorUtils.h>
|
| 8 |
+
#include <ATen/TracerMode.h>
|
| 9 |
+
#include <ATen/core/Generator.h>
|
| 10 |
+
#include <ATen/core/Reduction.h>
|
| 11 |
+
#include <ATen/core/Tensor.h>
|
| 12 |
+
#include <c10/core/Scalar.h>
|
| 13 |
+
#include <c10/core/Storage.h>
|
| 14 |
+
#include <c10/core/TensorOptions.h>
|
| 15 |
+
#include <c10/util/Deprecated.h>
|
| 16 |
+
#include <optional>
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
#include <ATen/ops/_scaled_dot_product_flash_attention_backward_ops.h>
|
| 21 |
+
|
| 22 |
+
namespace at {
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
// aten::_scaled_dot_product_flash_attention_backward(Tensor grad_out, Tensor query, Tensor key, Tensor value, Tensor out, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, SymInt max_q, SymInt max_k, float dropout_p, bool is_causal, Tensor philox_seed, Tensor philox_offset, *, float? scale=None) -> (Tensor grad_query, Tensor grad_key, Tensor grad_value)
|
| 26 |
+
inline ::std::tuple<at::Tensor,at::Tensor,at::Tensor> _scaled_dot_product_flash_attention_backward(const at::Tensor & grad_out, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const at::Tensor & out, const at::Tensor & logsumexp, const at::Tensor & cum_seq_q, const at::Tensor & cum_seq_k, int64_t max_q, int64_t max_k, double dropout_p, bool is_causal, const at::Tensor & philox_seed, const at::Tensor & philox_offset, ::std::optional<double> scale=::std::nullopt) {
|
| 27 |
+
return at::_ops::_scaled_dot_product_flash_attention_backward::call(grad_out, query, key, value, out, logsumexp, cum_seq_q, cum_seq_k, max_q, max_k, dropout_p, is_causal, philox_seed, philox_offset, scale);
|
| 28 |
+
}
|
| 29 |
+
namespace symint {
|
| 30 |
+
template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
|
| 31 |
+
::std::tuple<at::Tensor,at::Tensor,at::Tensor> _scaled_dot_product_flash_attention_backward(const at::Tensor & grad_out, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const at::Tensor & out, const at::Tensor & logsumexp, const at::Tensor & cum_seq_q, const at::Tensor & cum_seq_k, int64_t max_q, int64_t max_k, double dropout_p, bool is_causal, const at::Tensor & philox_seed, const at::Tensor & philox_offset, ::std::optional<double> scale=::std::nullopt) {
|
| 32 |
+
return at::_ops::_scaled_dot_product_flash_attention_backward::call(grad_out, query, key, value, out, logsumexp, cum_seq_q, cum_seq_k, max_q, max_k, dropout_p, is_causal, philox_seed, philox_offset, scale);
|
| 33 |
+
}
|
| 34 |
+
}
|
| 35 |
+
|
| 36 |
+
// aten::_scaled_dot_product_flash_attention_backward(Tensor grad_out, Tensor query, Tensor key, Tensor value, Tensor out, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, SymInt max_q, SymInt max_k, float dropout_p, bool is_causal, Tensor philox_seed, Tensor philox_offset, *, float? scale=None) -> (Tensor grad_query, Tensor grad_key, Tensor grad_value)
|
| 37 |
+
inline ::std::tuple<at::Tensor,at::Tensor,at::Tensor> _scaled_dot_product_flash_attention_backward_symint(const at::Tensor & grad_out, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const at::Tensor & out, const at::Tensor & logsumexp, const at::Tensor & cum_seq_q, const at::Tensor & cum_seq_k, c10::SymInt max_q, c10::SymInt max_k, double dropout_p, bool is_causal, const at::Tensor & philox_seed, const at::Tensor & philox_offset, ::std::optional<double> scale=::std::nullopt) {
|
| 38 |
+
return at::_ops::_scaled_dot_product_flash_attention_backward::call(grad_out, query, key, value, out, logsumexp, cum_seq_q, cum_seq_k, max_q, max_k, dropout_p, is_causal, philox_seed, philox_offset, scale);
|
| 39 |
+
}
|
| 40 |
+
namespace symint {
|
| 41 |
+
template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
|
| 42 |
+
::std::tuple<at::Tensor,at::Tensor,at::Tensor> _scaled_dot_product_flash_attention_backward(const at::Tensor & grad_out, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const at::Tensor & out, const at::Tensor & logsumexp, const at::Tensor & cum_seq_q, const at::Tensor & cum_seq_k, c10::SymInt max_q, c10::SymInt max_k, double dropout_p, bool is_causal, const at::Tensor & philox_seed, const at::Tensor & philox_offset, ::std::optional<double> scale=::std::nullopt) {
|
| 43 |
+
return at::_ops::_scaled_dot_product_flash_attention_backward::call(grad_out, query, key, value, out, logsumexp, cum_seq_q, cum_seq_k, max_q, max_k, dropout_p, is_causal, philox_seed, philox_offset, scale);
|
| 44 |
+
}
|
| 45 |
+
}
|
| 46 |
+
|
| 47 |
+
}
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_test_warn_in_autograd_native.h
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API at::Tensor _test_warn_in_autograd(const at::Tensor & self);
|
| 20 |
+
TORCH_API at::Tensor & _test_warn_in_autograd_out(const at::Tensor & self, at::Tensor & out);
|
| 21 |
+
} // namespace native
|
| 22 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_to_dense_native.h
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API at::Tensor & _to_dense_out(const at::Tensor & self, ::std::optional<at::ScalarType> dtype, ::std::optional<bool> masked_grad, at::Tensor & out);
|
| 20 |
+
TORCH_API at::Tensor sparse_to_dense(const at::Tensor & self, ::std::optional<at::ScalarType> dtype=::std::nullopt, ::std::optional<bool> masked_grad=::std::nullopt);
|
| 21 |
+
TORCH_API at::Tensor sparse_compressed_to_dense(const at::Tensor & self, ::std::optional<at::ScalarType> dtype=::std::nullopt, ::std::optional<bool> masked_grad=::std::nullopt);
|
| 22 |
+
TORCH_API at::Tensor mkldnn_to_dense(const at::Tensor & self, ::std::optional<at::ScalarType> dtype=::std::nullopt, ::std::optional<bool> masked_grad=::std::nullopt);
|
| 23 |
+
} // namespace native
|
| 24 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_bilinear2d_aa_compositeexplicitautogradnonfunctional_dispatch.h
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace compositeexplicitautogradnonfunctional {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor _upsample_bilinear2d_aa(const at::Tensor & self, at::IntArrayRef output_size, bool align_corners, ::std::optional<double> scales_h=::std::nullopt, ::std::optional<double> scales_w=::std::nullopt);
|
| 21 |
+
TORCH_API at::Tensor _upsample_bilinear2d_aa_symint(const at::Tensor & self, c10::SymIntArrayRef output_size, bool align_corners, ::std::optional<double> scales_h=::std::nullopt, ::std::optional<double> scales_w=::std::nullopt);
|
| 22 |
+
|
| 23 |
+
} // namespace compositeexplicitautogradnonfunctional
|
| 24 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact1d_meta_dispatch.h
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace meta {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor _upsample_nearest_exact1d(const at::Tensor & self, at::IntArrayRef output_size, ::std::optional<double> scales=::std::nullopt);
|
| 21 |
+
TORCH_API at::Tensor _upsample_nearest_exact1d_symint(const at::Tensor & self, c10::SymIntArrayRef output_size, ::std::optional<double> scales=::std::nullopt);
|
| 22 |
+
TORCH_API at::Tensor & _upsample_nearest_exact1d_out(at::Tensor & out, const at::Tensor & self, at::IntArrayRef output_size, ::std::optional<double> scales=::std::nullopt);
|
| 23 |
+
TORCH_API at::Tensor & _upsample_nearest_exact1d_outf(const at::Tensor & self, at::IntArrayRef output_size, ::std::optional<double> scales, at::Tensor & out);
|
| 24 |
+
TORCH_API at::Tensor & _upsample_nearest_exact1d_symint_out(at::Tensor & out, const at::Tensor & self, c10::SymIntArrayRef output_size, ::std::optional<double> scales=::std::nullopt);
|
| 25 |
+
TORCH_API at::Tensor & _upsample_nearest_exact1d_symint_outf(const at::Tensor & self, c10::SymIntArrayRef output_size, ::std::optional<double> scales, at::Tensor & out);
|
| 26 |
+
|
| 27 |
+
} // namespace meta
|
| 28 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/addcmul_cuda_dispatch.h
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace cuda {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor addcmul(const at::Tensor & self, const at::Tensor & tensor1, const at::Tensor & tensor2, const at::Scalar & value=1);
|
| 21 |
+
TORCH_API at::Tensor & addcmul_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & tensor1, const at::Tensor & tensor2, const at::Scalar & value=1);
|
| 22 |
+
TORCH_API at::Tensor & addcmul_outf(const at::Tensor & self, const at::Tensor & tensor1, const at::Tensor & tensor2, const at::Scalar & value, at::Tensor & out);
|
| 23 |
+
TORCH_API at::Tensor & addcmul_(at::Tensor & self, const at::Tensor & tensor1, const at::Tensor & tensor2, const at::Scalar & value=1);
|
| 24 |
+
|
| 25 |
+
} // namespace cuda
|
| 26 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/and_native.h
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API at::Tensor __and__(const at::Tensor & self, const at::Scalar & other);
|
| 20 |
+
TORCH_API at::Tensor & __iand__(at::Tensor & self, const at::Scalar & other);
|
| 21 |
+
TORCH_API at::Tensor __and__(const at::Tensor & self, const at::Tensor & other);
|
| 22 |
+
TORCH_API at::Tensor & __iand__(at::Tensor & self, const at::Tensor & other);
|
| 23 |
+
} // namespace native
|
| 24 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/as_strided_copy_ops.h
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Operator.h
|
| 4 |
+
|
| 5 |
+
#include <tuple>
|
| 6 |
+
#include <vector>
|
| 7 |
+
|
| 8 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 9 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 10 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 11 |
+
#include <ATen/core/ATen_fwd.h>
|
| 12 |
+
|
| 13 |
+
namespace at {
|
| 14 |
+
namespace _ops {
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
struct TORCH_API as_strided_copy {
|
| 18 |
+
using schema = at::Tensor (const at::Tensor &, c10::SymIntArrayRef, c10::SymIntArrayRef, ::std::optional<c10::SymInt>);
|
| 19 |
+
using ptr_schema = schema*;
|
| 20 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 21 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::as_strided_copy")
|
| 22 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
|
| 23 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "as_strided_copy(Tensor self, SymInt[] size, SymInt[] stride, SymInt? storage_offset=None) -> Tensor")
|
| 24 |
+
static at::Tensor call(const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, ::std::optional<c10::SymInt> storage_offset);
|
| 25 |
+
static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, ::std::optional<c10::SymInt> storage_offset);
|
| 26 |
+
};
|
| 27 |
+
|
| 28 |
+
struct TORCH_API as_strided_copy_out {
|
| 29 |
+
using schema = at::Tensor & (const at::Tensor &, c10::SymIntArrayRef, c10::SymIntArrayRef, ::std::optional<c10::SymInt>, at::Tensor &);
|
| 30 |
+
using ptr_schema = schema*;
|
| 31 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 32 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::as_strided_copy")
|
| 33 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
|
| 34 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "as_strided_copy.out(Tensor self, SymInt[] size, SymInt[] stride, SymInt? storage_offset=None, *, Tensor(a!) out) -> Tensor(a!)")
|
| 35 |
+
static at::Tensor & call(const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, ::std::optional<c10::SymInt> storage_offset, at::Tensor & out);
|
| 36 |
+
static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, ::std::optional<c10::SymInt> storage_offset, at::Tensor & out);
|
| 37 |
+
};
|
| 38 |
+
|
| 39 |
+
}} // namespace at::_ops
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/as_strided_native.h
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from NativeFunction.h
|
| 4 |
+
|
| 5 |
+
#include <c10/core/Scalar.h>
|
| 6 |
+
#include <c10/core/Storage.h>
|
| 7 |
+
#include <c10/core/TensorOptions.h>
|
| 8 |
+
#include <c10/util/Deprecated.h>
|
| 9 |
+
#include <optional>
|
| 10 |
+
#include <c10/core/QScheme.h>
|
| 11 |
+
#include <ATen/core/Reduction.h>
|
| 12 |
+
#include <ATen/core/Tensor.h>
|
| 13 |
+
#include <tuple>
|
| 14 |
+
#include <vector>
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
namespace at {
|
| 18 |
+
namespace native {
|
| 19 |
+
TORCH_API at::Tensor as_strided_tensorimpl(const at::Tensor & self, at::IntArrayRef size, at::IntArrayRef stride, ::std::optional<int64_t> storage_offset=::std::nullopt);
|
| 20 |
+
TORCH_API at::Tensor as_strided_tensorimpl_meta_symint(const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, ::std::optional<c10::SymInt> storage_offset=::std::nullopt);
|
| 21 |
+
TORCH_API at::Tensor as_strided_qtensorimpl(const at::Tensor & self, at::IntArrayRef size, at::IntArrayRef stride, ::std::optional<int64_t> storage_offset=::std::nullopt);
|
| 22 |
+
TORCH_API const at::Tensor & as_strided__symint(const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, ::std::optional<c10::SymInt> storage_offset=::std::nullopt);
|
| 23 |
+
} // namespace native
|
| 24 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/avg_pool2d_ops.h
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Operator.h
|
| 4 |
+
|
| 5 |
+
#include <tuple>
|
| 6 |
+
#include <vector>
|
| 7 |
+
|
| 8 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 9 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 10 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 11 |
+
#include <ATen/core/ATen_fwd.h>
|
| 12 |
+
|
| 13 |
+
namespace at {
|
| 14 |
+
namespace _ops {
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
struct TORCH_API avg_pool2d_out {
|
| 18 |
+
using schema = at::Tensor & (const at::Tensor &, at::IntArrayRef, at::IntArrayRef, at::IntArrayRef, bool, bool, ::std::optional<int64_t>, at::Tensor &);
|
| 19 |
+
using ptr_schema = schema*;
|
| 20 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 21 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::avg_pool2d")
|
| 22 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
|
| 23 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "avg_pool2d.out(Tensor self, int[2] kernel_size, int[2] stride=[], int[2] padding=0, bool ceil_mode=False, bool count_include_pad=True, int? divisor_override=None, *, Tensor(a!) out) -> Tensor(a!)")
|
| 24 |
+
static at::Tensor & call(const at::Tensor & self, at::IntArrayRef kernel_size, at::IntArrayRef stride, at::IntArrayRef padding, bool ceil_mode, bool count_include_pad, ::std::optional<int64_t> divisor_override, at::Tensor & out);
|
| 25 |
+
static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, at::IntArrayRef kernel_size, at::IntArrayRef stride, at::IntArrayRef padding, bool ceil_mode, bool count_include_pad, ::std::optional<int64_t> divisor_override, at::Tensor & out);
|
| 26 |
+
};
|
| 27 |
+
|
| 28 |
+
struct TORCH_API avg_pool2d {
|
| 29 |
+
using schema = at::Tensor (const at::Tensor &, at::IntArrayRef, at::IntArrayRef, at::IntArrayRef, bool, bool, ::std::optional<int64_t>);
|
| 30 |
+
using ptr_schema = schema*;
|
| 31 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 32 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::avg_pool2d")
|
| 33 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
|
| 34 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "avg_pool2d(Tensor self, int[2] kernel_size, int[2] stride=[], int[2] padding=0, bool ceil_mode=False, bool count_include_pad=True, int? divisor_override=None) -> Tensor")
|
| 35 |
+
static at::Tensor call(const at::Tensor & self, at::IntArrayRef kernel_size, at::IntArrayRef stride, at::IntArrayRef padding, bool ceil_mode, bool count_include_pad, ::std::optional<int64_t> divisor_override);
|
| 36 |
+
static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, at::IntArrayRef kernel_size, at::IntArrayRef stride, at::IntArrayRef padding, bool ceil_mode, bool count_include_pad, ::std::optional<int64_t> divisor_override);
|
| 37 |
+
};
|
| 38 |
+
|
| 39 |
+
}} // namespace at::_ops
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/bitwise_or_cpu_dispatch.h
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace cpu {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor bitwise_or(const at::Tensor & self, const at::Tensor & other);
|
| 21 |
+
TORCH_API at::Tensor & bitwise_or_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & other);
|
| 22 |
+
TORCH_API at::Tensor & bitwise_or_outf(const at::Tensor & self, const at::Tensor & other, at::Tensor & out);
|
| 23 |
+
TORCH_API at::Tensor & bitwise_or_(at::Tensor & self, const at::Tensor & other);
|
| 24 |
+
|
| 25 |
+
} // namespace cpu
|
| 26 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/bucketize_ops.h
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
// @generated by torchgen/gen.py from Operator.h
|
| 4 |
+
|
| 5 |
+
#include <tuple>
|
| 6 |
+
#include <vector>
|
| 7 |
+
|
| 8 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 9 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 10 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 11 |
+
#include <ATen/core/ATen_fwd.h>
|
| 12 |
+
|
| 13 |
+
namespace at {
|
| 14 |
+
namespace _ops {
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
struct TORCH_API bucketize_Tensor {
|
| 18 |
+
using schema = at::Tensor (const at::Tensor &, const at::Tensor &, bool, bool);
|
| 19 |
+
using ptr_schema = schema*;
|
| 20 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 21 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::bucketize")
|
| 22 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "Tensor")
|
| 23 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "bucketize.Tensor(Tensor self, Tensor boundaries, *, bool out_int32=False, bool right=False) -> Tensor")
|
| 24 |
+
static at::Tensor call(const at::Tensor & self, const at::Tensor & boundaries, bool out_int32, bool right);
|
| 25 |
+
static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, const at::Tensor & boundaries, bool out_int32, bool right);
|
| 26 |
+
};
|
| 27 |
+
|
| 28 |
+
struct TORCH_API bucketize_Tensor_out {
|
| 29 |
+
using schema = at::Tensor & (const at::Tensor &, const at::Tensor &, bool, bool, at::Tensor &);
|
| 30 |
+
using ptr_schema = schema*;
|
| 31 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 32 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::bucketize")
|
| 33 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "Tensor_out")
|
| 34 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "bucketize.Tensor_out(Tensor self, Tensor boundaries, *, bool out_int32=False, bool right=False, Tensor(a!) out) -> Tensor(a!)")
|
| 35 |
+
static at::Tensor & call(const at::Tensor & self, const at::Tensor & boundaries, bool out_int32, bool right, at::Tensor & out);
|
| 36 |
+
static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, const at::Tensor & boundaries, bool out_int32, bool right, at::Tensor & out);
|
| 37 |
+
};
|
| 38 |
+
|
| 39 |
+
struct TORCH_API bucketize_Scalar {
|
| 40 |
+
using schema = at::Tensor (const at::Scalar &, const at::Tensor &, bool, bool);
|
| 41 |
+
using ptr_schema = schema*;
|
| 42 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 43 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::bucketize")
|
| 44 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "Scalar")
|
| 45 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "bucketize.Scalar(Scalar self, Tensor boundaries, *, bool out_int32=False, bool right=False) -> Tensor")
|
| 46 |
+
static at::Tensor call(const at::Scalar & self, const at::Tensor & boundaries, bool out_int32, bool right);
|
| 47 |
+
static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Scalar & self, const at::Tensor & boundaries, bool out_int32, bool right);
|
| 48 |
+
};
|
| 49 |
+
|
| 50 |
+
struct TORCH_API bucketize_Scalar_out {
|
| 51 |
+
using schema = at::Tensor & (const at::Scalar &, const at::Tensor &, bool, bool, at::Tensor &);
|
| 52 |
+
using ptr_schema = schema*;
|
| 53 |
+
// See Note [static constexpr char* members for windows NVCC]
|
| 54 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::bucketize")
|
| 55 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "Scalar_out")
|
| 56 |
+
STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "bucketize.Scalar_out(Scalar self, Tensor boundaries, *, bool out_int32=False, bool right=False, Tensor(a!) out) -> Tensor(a!)")
|
| 57 |
+
static at::Tensor & call(const at::Scalar & self, const at::Tensor & boundaries, bool out_int32, bool right, at::Tensor & out);
|
| 58 |
+
static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Scalar & self, const at::Tensor & boundaries, bool out_int32, bool right, at::Tensor & out);
|
| 59 |
+
};
|
| 60 |
+
|
| 61 |
+
}} // namespace at::_ops
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/copy_meta_dispatch.h
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace meta {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor copy(const at::Tensor & self, const at::Tensor & src, bool non_blocking=false);
|
| 21 |
+
|
| 22 |
+
} // namespace meta
|
| 23 |
+
} // namespace at
|
vllm/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_convolution_transpose_compositeexplicitautograd_dispatch.h
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
// @generated by torchgen/gen.py from DispatchKeyFunction.h
|
| 3 |
+
|
| 4 |
+
// NB: The implementing C++ file is RegisterDispatchKey.cpp
|
| 5 |
+
|
| 6 |
+
// The only #includes we need are for custom classes that have defaults in the C++ API
|
| 7 |
+
#include <c10/core/MemoryFormat.h>
|
| 8 |
+
#include <c10/core/Scalar.h>
|
| 9 |
+
#include <ATen/core/Reduction.h>
|
| 10 |
+
|
| 11 |
+
// Forward declarations of any types needed in the operator signatures.
|
| 12 |
+
// We can't directly include these classes because it will cause circular include dependencies.
|
| 13 |
+
// This file is included by TensorBody.h, which defines the Tensor class.
|
| 14 |
+
#include <ATen/core/ATen_fwd.h>
|
| 15 |
+
|
| 16 |
+
namespace at {
|
| 17 |
+
|
| 18 |
+
namespace compositeexplicitautograd {
|
| 19 |
+
|
| 20 |
+
TORCH_API at::Tensor & cudnn_convolution_transpose_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & weight, at::IntArrayRef padding, at::IntArrayRef output_padding, at::IntArrayRef stride, at::IntArrayRef dilation, int64_t groups, bool benchmark, bool deterministic, bool allow_tf32);
|
| 21 |
+
TORCH_API at::Tensor & cudnn_convolution_transpose_outf(const at::Tensor & self, const at::Tensor & weight, at::IntArrayRef padding, at::IntArrayRef output_padding, at::IntArrayRef stride, at::IntArrayRef dilation, int64_t groups, bool benchmark, bool deterministic, bool allow_tf32, at::Tensor & out);
|
| 22 |
+
TORCH_API at::Tensor & cudnn_convolution_transpose_symint_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & weight, c10::SymIntArrayRef padding, c10::SymIntArrayRef output_padding, c10::SymIntArrayRef stride, c10::SymIntArrayRef dilation, c10::SymInt groups, bool benchmark, bool deterministic, bool allow_tf32);
|
| 23 |
+
TORCH_API at::Tensor & cudnn_convolution_transpose_symint_outf(const at::Tensor & self, const at::Tensor & weight, c10::SymIntArrayRef padding, c10::SymIntArrayRef output_padding, c10::SymIntArrayRef stride, c10::SymIntArrayRef dilation, c10::SymInt groups, bool benchmark, bool deterministic, bool allow_tf32, at::Tensor & out);
|
| 24 |
+
|
| 25 |
+
} // namespace compositeexplicitautograd
|
| 26 |
+
} // namespace at
|