| <!--Copyright 2024 The HuggingFace Team. All rights reserved. | |
| Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with | |
| the License. You may obtain a copy of the License at | |
| http://www.apache.org/licenses/LICENSE-2.0 | |
| Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on | |
| an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the | |
| specific language governing permissions and limitations under the License. | |
| β οΈ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be | |
| rendered properly in your Markdown viewer. | |
| --> | |
| # EETQ [[eetq]] | |
| [EETQ](https://github.com/NetEase-FuXi/EETQ) λΌμ΄λΈλ¬λ¦¬λ NVIDIA GPUμ λν΄ int8 μ±λλ³(per-channel) κ°μ€μΉ μ μ© μμν(weight-only quantization)μ μ§μν©λλ€. κ³ μ±λ₯ GEMM λ° GEMV 컀λμ FasterTransformer λ° TensorRT-LLMμμ κ°μ Έμμ΅λλ€. κ΅μ (calibration) λ°μ΄ν°μ μ΄ νμ μμΌλ©°, λͺ¨λΈμ μ¬μ μ μμνν νμλ μμ΅λλ€. λν, μ±λλ³ μμν(per-channel quantization) λλΆμ μ νλ μ νκ° λ―Έλ―Έν©λλ€. | |
| [λ¦΄λ¦¬μ€ νμ΄μ§](https://github.com/NetEase-FuXi/EETQ/releases)μμ eetqλ₯Ό μ€μΉνλμ§ νμΈνμΈμ. | |
| ``` | |
| pip install --no-cache-dir https://github.com/NetEase-FuXi/EETQ/releases/download/v1.0.0/EETQ-1.0.0+cu121+torch2.1.2-cp310-cp310-linux_x86_64.whl | |
| ``` | |
| λλ μμ€ μ½λ https://github.com/NetEase-FuXi/EETQ μμ μ€μΉν μ μμ΅λλ€. EETQλ CUDA κΈ°λ₯μ΄ 8.9 μ΄νμ΄κ³ 7.0 μ΄μμ΄μ΄μΌ ν©λλ€. | |
| ``` | |
| git clone https://github.com/NetEase-FuXi/EETQ.git | |
| cd EETQ/ | |
| git submodule update --init --recursive | |
| pip install . | |
| ``` | |
| λΉμμν λͺ¨λΈμ "from_pretrained"λ₯Ό ν΅ν΄ μμνν μ μμ΅λλ€. | |
| ```py | |
| from transformers import AutoModelForCausalLM, EetqConfig | |
| path = "/path/to/model". | |
| quantization_config = EetqConfig("int8") | |
| model = AutoModelForCausalLM.from_pretrained(path, device_map="auto", quantization_config=quantization_config) | |
| ``` | |
| μμνλ λͺ¨λΈμ "save_pretrained"λ₯Ό ν΅ν΄ μ μ₯ν μ μμΌλ©°, "from_pretrained"λ₯Ό ν΅ν΄ λ€μ μ¬μ©ν μ μμ΅λλ€. | |
| ```py | |
| quant_path = "/path/to/save/quantized/model" | |
| model.save_pretrained(quant_path) | |
| model = AutoModelForCausalLM.from_pretrained(quant_path, device_map="auto") | |
| ``` |