# torchao

[![Open In Colab: Torchao Demo](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/transformers_doc/en/quantization/torchao.ipynb)

[torchao](https://github.com/pytorch/ao) は、カスタムな高パフォーマンスデータ型、量子化、スパース性をサポートする PyTorch アーキテクチャ最適化ライブラリです。[torch.compile](https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html) などの PyTorch ネイティブ機能と組み合わせることで、さらに高速な推論と学習を実現できます。

torchaoの追加機能については、以下の表を参照してください。

| 機能 | 説明 |
|--------|-------------|
| **量子化認識トレーニング (QAT)** | 精度の低下を最小限に抑えて量子化モデルを学習します（[QAT README](https://github.com/pytorch/ao/blob/main/torchao/quantization/qat/README.md)を参照） |
| **Float8 学習** | float8形式による高スループットな学習（[torchtitan](https://github.com/pytorch/torchtitan/blob/main/docs/float8.md)および[Accelerate](https://huggingface.co/docs/accelerate/usage_guides/low_precision_training#configuring-torchao)のドキュメントを参照） |
| **スパース性のサポート** | 高速な推論のための半構造化(2:4)スパース性(詳細は[Accelerating Neural Network Training with Semi-Structured (2:4) Sparsity](https://pytorch.org/blog/accelerating-neural-network-training/)ブログ記事をご参照ください) |
| **オプティマイザー量子化** | 4ビットおよび8ビットのAdamバリアントを使用して、オプティマイザ状態のメモリを削減 |
| **KVキャッシュの量子化** | メモリを低減しながら長文脈推論を有効にします（[KV Cache Quantization](https://github.com/pytorch/ao/blob/main/torchao/_models/llama/README.md)を参照） |
| **カスタムカーネルのサポート** | 独自の`torch.compile`互換オペレーションを使用します |
| **FSDP2** | FSDP2と組み合わせ可能で、トレーニングに対応|

> [!TIP]
> ライブラリの詳細については、torchaoの[README.md](https://github.com/pytorch/ao#torchao-pytorch-architecture-optimization)を参照してください。

torchao は、以下の[量子化技術](https://github.com/pytorch/ao/blob/main/torchao/quantization/README.md)をサポートしています。

- A16W8 Float8 動的量子化
- A16W8 Float8 WeightOnly 量子化
- A8W8 Int8 動的量子化
- A16W8 Int8 重みのみ量子化
- A16W4 Int4 重みのみ量子化
- A16W4 Int4 重みのみ量子化 + 2:4 スパース
- Autoquantization

torchao は、モジュールの完全修飾名とその対応する量子化設定の辞書を指定することで、モジュールレベルの設定もサポートしています。これにより、特定のレイヤーの量子化をスキップしたり、異なるモジュールに異なる量子化設定を使用したりできます。

以下の表を確認し、お使いのハードウェアが互換性があるかどうかをお確かめください。

| コンポーネント | 互換性 |
|----------|----------------|
| CUDA バージョン | ✅ cu118、cu126、cu128 |
| XPU バージョン | ✅ pytorch2.8 |
| CPU | ✅ `device_map="cpu"` を変更します（以下の例を参照してください） |

以下のコマンドで、PyPi または PyTorch インデックスから torchao をインストールします。

```bash
# Updating 🤗 Transformers to the latest version, as the example script below uses the new auto compilation
# Stable release from Pypi which will default to CUDA 12.6
pip install --upgrade torchao transformers
```

PyTorchインデックスからの安定版リリース

```bash
pip install torchao --index-url https://download.pytorch.org/whl/cu126 # options are cpu/cu118/cu126/cu128
```

torchao >= 0.15.0 is required. The string-based API (e.g., `TorchAoConfig("int4_weight_only")`) has been removed — use `AOBaseConfig` objects instead (see examples below).

## 量子化の例

TorchAO は、さまざまな量子化設定を提供しています。各設定は、`group_size`、`scheme`、`layout` などのパラメータでさらにカスタマイズでき、特定のハードウェアやモデルアーキテクチャに合わせて最適化できます。

利用可能な設定の完全なリストについては、[量子化 API のドキュメント](https://github.com/pytorch/ao/blob/main/torchao/quantization/quant_api.py)を参照してください。

量子化の種類と設定を手動で選択するか、量子化の種類を自動選択するかを選択できます。

[TorchAoConfig](/docs/transformers/main/ja/main_classes/quantization#transformers.TorchAoConfig) を作成し、量子化対象の重みの量子化タイプと `group_size`（int8 weight only および int4 weight only の場合）を指定します。フォワードメソッドを自動的に [torch.compile](https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html) するように、`cache_implementation` を `"static"` に設定します。

ハードウェア（例: A100 GPU、H100 GPU、CPU）に基づいた推奨される量子化手法の例を示します。

> [!WARNING]
> `cache_implementation="static"` を設定すると、torchao は最初の推論時にモデルを自動的にコンパイルします。バッチサイズまたは `max_new_tokens` が変更されるたびに、モデルは再コンパイルされます。コンパイルせずに量子化を行う場合は、[generate()](/docs/transformers/main/ja/main_classes/text_generation#transformers.GenerationMixin.generate) に `disable_compile=True` を渡してください。

### H100 GPU

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import Float8DynamicActivationFloat8WeightConfig, Float8WeightOnlyConfig

quant_config = Float8DynamicActivationFloat8WeightConfig()
# or float8 weight only quantization
# quant_config = Float8WeightOnlyConfig()
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device, quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import GemliteUIntXWeightOnlyConfig

# We integrated with gemlite, which optimizes for batch size N on A100 and H100
quant_config = GemliteUIntXWeightOnlyConfig(group_size=128)
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device, quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

### A100 GPU

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import Int8DynamicActivationInt8WeightConfig, Int8WeightOnlyConfig

quant_config = Int8DynamicActivationInt8WeightConfig()
# or int8 weight only quantization
# quant_config = Int8WeightOnlyConfig()
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device, quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import GemliteUIntXWeightOnlyConfig, Int4WeightOnlyConfig

# For batch size N, we recommend gemlite, which may require autotuning
# default is 4 bit, 8 bit is also supported by passing `bit_width=8`
quant_config = GemliteUIntXWeightOnlyConfig(group_size=128)

# For batch size 1, we also have custom tinygemm kernel that's only optimized for this
# We can set `use_hqq` to `True` for better accuracy
# quant_config = Int4WeightOnlyConfig(group_size=128, use_hqq=True)

quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device, quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

### Intel XPU

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import Int8DynamicActivationInt8WeightConfig, Int8WeightOnlyConfig

quant_config = Int8DynamicActivationInt8WeightConfig()
# or int8 weight only quantization
# quant_config = Int8WeightOnlyConfig()
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device, quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import Int4WeightOnlyConfig

quant_config = Int4WeightOnlyConfig(group_size=128, int4_packing_format="plain_int32")
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device).to(quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

### CPU

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import Int8DynamicActivationInt8WeightConfig, Int8WeightOnlyConfig

quant_config = Int8DynamicActivationInt8WeightConfig()
# quant_config = Int8WeightOnlyConfig()
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="cpu",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device, quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

> [!TIP]
> CPU int4 weight-only quantizationには、torchaoのバージョン0.15.0以降が必要です。

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.prototype.quantization.int4 import PrototypeInt4WeightOnlyConfig

quantization_config = TorchAoConfig(PrototypeInt4WeightOnlyConfig(group_size=128, int4_choose_qparams_algorithm="tinygemm"))

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="cpu",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(quantized_model.device).to(quantized_model.dtype)

# auto-compile the quantized model with `cache_implementation="static"` to get speed up
output = quantized_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))
```

### モジュールごとの量子化

#### 1. 特定のレイヤーの量子化をスキップする

`FqnToConfig` を使用すると、すべてのレイヤーにデフォルトの構成を指定しつつ、特定のレイヤーの量子化をスキップできます。

```py
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TorchAoConfig

model_id = "meta-llama/Llama-3.1-8B-Instruct"

from torchao.quantization import Int4WeightOnlyConfig, FqnToConfig
config = Int4WeightOnlyConfig(group_size=128)

# set default to int4 (for linears), and skip quantizing `model.layers.0.self_attn.q_proj`
quant_config = FqnToConfig({"_default": config, "model.layers.0.self_attn.q_proj": None})
quantization_config = TorchAoConfig(quant_type=quant_config)
quantized_model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", dtype=torch.bfloat16, quantization_config=quantization_config)
# lm_head is not quantized and model.layers.0.self_attn.q_proj is not quantized
print("quantized model:", quantized_model)
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Manual Testing
prompt = "Hey, are you conscious? Can you talk to me?"
inputs = tokenizer(prompt, return_tensors="pt").to(quantized_model.device, quantized_model.dtype)
generated_ids = quantized_model.generate(**inputs, max_new_tokens=128)
output_text = tokenizer.batch_decode(
    generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)
```

#### 2. 異なるレイヤーに異なる量子化設定を適用する（正規表現なし）

```py
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TorchAoConfig

model_id = "facebook/opt-125m"

from torchao.quantization import Int4WeightOnlyConfig, FqnToConfig, Int8DynamicActivationInt4WeightConfig, IntxWeightOnlyConfig, PerAxis, MappingType

weight_dtype = torch.int8
granularity = PerAxis(0)
mapping_type = MappingType.ASYMMETRIC
embedding_config = IntxWeightOnlyConfig(
    weight_dtype=weight_dtype,
    granularity=granularity,
    mapping_type=mapping_type,
)
linear_config = Int8DynamicActivationInt4WeightConfig(group_size=128)
quant_config = FqnToConfig({"_default": linear_config, "model.decoder.embed_tokens": embedding_config, "model.decoder.embed_positions": None})
# set `include_embedding` to True in order to include embedding in quantization
# when `include_embedding` is True, we'll remove input embedding from `modules_not_to_convert` as well
quantization_config = TorchAoConfig(quant_type=quant_config, include_embedding=True)
quantized_model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cpu", dtype=torch.bfloat16, quantization_config=quantization_config)
print("quantized model:", quantized_model)
# make sure embedding is quantized
print("embed_tokens weight:", quantized_model.model.decoder.embed_tokens.weight)
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Manual Testing
prompt = "Hey, are you conscious? Can you talk to me?"
inputs = tokenizer(prompt, return_tensors="pt").to("cpu", quantized_model.dtype)
generated_ids = quantized_model.generate(**inputs, max_new_tokens=128, cache_implementation="static")
output_text = tokenizer.batch_decode(
    generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)
```

#### 3. 正規表現を使用して、異なるレイヤーに異なる量子化設定を適用する

`module_fqn` が正規表現に一致するすべてのモジュールの config を指定するために、正規表現を使用することもできます。すべての正規表現は `re:` で始まる必要があります。例えば、`re:layers\..*\.gate_proj` は `layers.0.gate_proj` のようなすべてのレイヤーに一致します。ドキュメントは[こちら](https://github.com/pytorch/ao/blob/2fe0ca0899c730c528efdbec8886feaa38879f39/torchao/quantization/quant_api.py#L2392)を参照してください。

```py
import logging

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TorchAoConfig

# Configure logging to see warnings and debug information
logging.basicConfig(
    level=logging.INFO, format="%(name)s - %(levelname)s - %(message)s"
)

# Enable specific loggers that might contain the serialization warnings
logging.getLogger("transformers").setLevel(logging.INFO)
logging.getLogger("torchao").setLevel(logging.INFO)
logging.getLogger("safetensors").setLevel(logging.INFO)
logging.getLogger("huggingface_hub").setLevel(logging.INFO)

model_id = "facebook/opt-125m"

from torchao.quantization import (
    Float8DynamicActivationFloat8WeightConfig,
    Int4WeightOnlyConfig,
    IntxWeightOnlyConfig,
    PerRow,
    PerAxis,
    FqnToConfig,
    Float8Tensor,
    Int4TilePackedTo4dTensor,
    IntxUnpackedToInt8Tensor,
)

float8dyn = Float8DynamicActivationFloat8WeightConfig(granularity=PerRow())
int4wo = Int4WeightOnlyConfig(int4_packing_format="tile_packed_to_4d")
intxwo = IntxWeightOnlyConfig(weight_dtype=torch.int8, granularity=PerAxis(0))

qconfig_dict = {
    # highest priority
    "model.decoder.layers.3.self_attn.q_proj": int4wo,
    "model.decoder.layers.3.self_attn.k_proj": int4wo,
    "model.decoder.layers.3.self_attn.v_proj": int4wo,
    # vllm
    "model.decoder.layers.3.self_attn.qkv_proj": int4wo,

    "re:model\.decoder\.layers\..+\.self_attn\.q_proj": float8dyn,
    "re:model\.decoder\.layers\..+\.self_attn\.k_proj": float8dyn,
    "re:model\.decoder\.layers\..+\.self_attn\.v_proj": float8dyn,
    # this should not take effect and we'll fallback to _default
    # since no full mach (missing `j` in the end)
    "re:model\.decoder\.layers\..+\.self_attn\.out_pro": float8dyn,
    # vllm
    "re:model\.decoder\.layers\..+\.self_attn\.qkv_proj": float8dyn,

    "_default": intxwo,
}
quant_config = FqnToConfig(qconfig_dict)
quantization_config = TorchAoConfig(quant_type=quant_config)
quantized_model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype=torch.bfloat16,
    quantization_config=quantization_config,
)
print("quantized model:", quantized_model)
tokenizer = AutoTokenizer.from_pretrained(model_id)
for i in range(12):
    if i == 3:
        assert isinstance(quantized_model.model.decoder.layers[i].self_attn.q_proj.weight, Int4TilePackedTo4dTensor)
        assert isinstance(quantized_model.model.decoder.layers[i].self_attn.k_proj.weight, Int4TilePackedTo4dTensor)
        assert isinstance(quantized_model.model.decoder.layers[i].self_attn.v_proj.weight, Int4TilePackedTo4dTensor)
    else:
        assert isinstance(quantized_model.model.decoder.layers[i].self_attn.q_proj.weight, Float8Tensor)
        assert isinstance(quantized_model.model.decoder.layers[i].self_attn.k_proj.weight, Float8Tensor)
        assert isinstance(quantized_model.model.decoder.layers[i].self_attn.v_proj.weight, Float8Tensor)
    assert isinstance(quantized_model.model.decoder.layers[i].self_attn.out_proj.weight, IntxUnpackedToInt8Tensor)

# Manual Testing
prompt = "What are we having for dinner?"
print("Prompt:", prompt)
inputs = tokenizer(
    prompt,
    return_tensors="pt",
).to(quantized_model.device, quantized_model.dtype)
# setting temperature to 0 to make sure result deterministic
generated_ids = quantized_model.generate(**inputs, max_new_tokens=128, temperature=0)

correct_output_text = tokenizer.batch_decode(
    generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print("Response:", correct_output_text[0][len(prompt) :])

# Load model from saved checkpoint
reloaded_model = AutoModelForCausalLM.from_pretrained(
    save_to,
    device_map=torch.accelerator.current_accelerator(),
    torch_dtype=torch.bfloat16,
    # quantization_config=quantization_config,
)

generated_ids = reloaded_model.generate(**inputs, max_new_tokens=128, temperature=0)
output_text = tokenizer.batch_decode(
    generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print("Response:", output_text[0][len(prompt) :])

assert(correct_output_text == output_text)
```

## シリアライズ

`save_pretrained` による量子化モデルの保存（[safetensors](https://huggingface.co/docs/safetensors/en/index) 形式）は、torchao >= v0.15 のみサポートされています。それ以前のバージョンでは、[torch.save](https://docs.pytorch.org/docs/stable/generated/torch.save.html) を使用して、非安全な `.bin` チェックポイントとして手動で保存するしかありません。

```py
# torchao >= 0.15
output_dir = "llama3-8b-int4wo-128"
quantized_model.save_pretrained("llama3-8b-int4wo-128")
```

```py
# torchao >= 0.15
USER_ID = "your_huggingface_user_id"
REPO_ID = "llama3-8b-int4wo-128"
quantized_model.push_to_hub(f"{USER_ID}/llama3-8b-int4wo-128")
tokenizer.push_to_hub(f"{USER_ID}/llama3-8b-int4wo-128")
```

```py
# torchao < 0.15 -> unsafe serialization
filename = "llama3-8b-int4wo-128/pytorch_model.bin"
torch.save(quantized_model.state_dict(), filename)
```

## 量子化モデルの読み込み

量子化モデルの読み込み方法は、量子化スキームによって異なります。int8やfloat8などの量子化スキームでは、モデルを任意のデバイス上で量子化し、また任意のデバイス上で読み込むことができます。以下の例では、CPU上でモデルを量子化し、その後CUDAまたはXPU上で読み込む手順を示しています。

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import Int8WeightOnlyConfig

quant_config = Int8WeightOnlyConfig(group_size=128)
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="cpu",
    quantization_config=quantization_config
)
# save the quantized model
output_dir = "llama-3.1-8b-torchao-int8"
quantized_model.save_pretrained(output_dir)

# reload the quantized model
reloaded_model = AutoModelForCausalLM.from_pretrained(
    output_dir,
    device_map="auto",
    dtype=torch.bfloat16
)
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(reloaded_model.device.type)

output = reloaded_model.generate(**input_ids, max_new_tokens=10)
print(tokenizer.decode(output[0], skip_special_tokens=True))

```

int4の場合、レイアウトがデバイスに依存するため、モデルは量子化が行われたデバイスでのみ読み込むことができます。以下の例では、CPU上でモデルを量子化して読み込む方法を示しています。

```py
import torch
from transformers import TorchAoConfig, AutoModelForCausalLM, AutoTokenizer
from torchao.quantization import Int4WeightOnlyConfig
from torchao.quantization.quantize_.workflows import Int4PackingFormat

quant_config = Int4WeightOnlyConfig(group_size=128, int4_packing_format=Int4PackingFormat.PLAIN_INT32)
quantization_config = TorchAoConfig(quant_type=quant_config)

# Load and quantize the model
quantized_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    dtype="auto",
    device_map="cpu",
    quantization_config=quantization_config
)
# save the quantized model
output_dir = "llama-3.1-8b-torchao-int4-cpu"
quantized_model.save_pretrained(output_dir)

# reload the quantized model
reloaded_model = AutoModelForCausalLM.from_pretrained(
    output_dir,
    device_map="cpu",
    dtype=torch.bfloat16
)
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to(reloaded_model.device.type)

output = reloaded_model.generate(**input_ids, max_new_tokens=10)
print(tokenizer.decode(output[0], skip_special_tokens=True))

```

## リソース

期待されるパフォーマンスをより正確に把握するために、CUDA および XPU バックエンドでの各種モデルの[ベンチマーク](https://github.com/pytorch/ao/tree/main/torchao/quantization#benchmarks)をご参照ください。また、以下のコードを実行して、ご自身でモデルのベンチマークを行うこともできます。

```py
from torch._inductor.utils import do_bench_using_profiling
from typing import Callable

def benchmark_fn(func: Callable, *args, **kwargs) -> float:
    """Thin wrapper around do_bench_using_profiling"""
    no_args = lambda: func(*args, **kwargs)
    time = do_bench_using_profiling(no_args)
    return time * 1e3

MAX_NEW_TOKENS = 1000
print("int4wo-128 model:", benchmark_fn(quantized_model.generate, **input_ids, max_new_tokens=MAX_NEW_TOKENS, cache_implementation="static"))

bf16_model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", dtype=torch.bfloat16)
output = bf16_model.generate(**input_ids, max_new_tokens=10, cache_implementation="static") # auto-compile
print("bf16 model:", benchmark_fn(bf16_model.generate, **input_ids, max_new_tokens=MAX_NEW_TOKENS, cache_implementation="static"))
```

> [!TIP]
> 最適なパフォーマンスを得るには、`torchao.quantization.utils.recommended_inductor_config_setter()` を呼び出して推奨設定を使用できます。

その他の例やドキュメントについては、[Other Available Quantization Techniques](https://github.com/pytorch/ao/tree/main/torchao/quantization#other-available-quantization-techniques)を参照してください。

## 問題

Transformers 統合に関する問題が発生した場合は、[Transformers](https://github.com/huggingface/transformers/issues) リポジトリで issue を作成してください。torchao に直接関連する問題については、[torchao](https://github.com/pytorch/ao/issues) リポジトリで issue を作成してください。