Build from source quantization packages

CONTRIBUTING.md

@@ -55,9 +55,9 @@ If you would like to work on any of the open Issues:
 	For a better development experience, we recommend using isolated docker containers to run tests:
 
 	```bash
-	make build_docker_cpu
-	make run_docker_cpu
-	make install_cli_cpu_pytorch_extras
+	make build_cpu_image
+	make run_cpu_container
+	make install_cli_cpu_pytorch
 	make test_cli_cpu_pytorch
 	```
 

Makefile

@@ -4,6 +4,7 @@
 PWD := $(shell pwd)
 USER_ID := $(shell id -u)
 GROUP_ID := $(shell id -g)
+TORCH_VERSION := 2.3.1
 
 quality:
 	ruff check .
@@ -23,7 +24,7 @@ build_cpu_image:
 	docker build --build-arg IMAGE=optimum-benchmark:latest-cpu --build-arg USER_ID=$(USER_ID) --build-arg GROUP_ID=$(GROUP_ID) -t optimum-benchmark:latest-cpu docker/unroot
 
 build_cuda_image:
-	docker build -t optimum-benchmark:latest-cuda docker/cuda
+	docker build --build-arg TORCH_VERSION=$(TORCH_VERSION) -t optimum-benchmark:latest-cuda docker/cuda
 	docker build --build-arg IMAGE=optimum-benchmark:latest-cuda --build-arg USER_ID=$(USER_ID) --build-arg GROUP_ID=$(GROUP_ID) -t optimum-benchmark:latest-cuda docker/unroot
 
 build_cuda_ort_image:

docker/cuda/Dockerfile

@@ -25,7 +25,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     python3.10 python3-pip python3.10-dev && \
     apt-get clean && rm -rf /var/lib/apt/lists/* && \
     update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1 && \
-    pip install --no-cache-dir --upgrade pip setuptools wheel 
+    pip install --no-cache-dir --upgrade pip setuptools wheel requests
 
 # Install PyTorch
 ARG TORCH_CUDA=cu121

setup.py

@@ -62,8 +62,10 @@
     AUTOAWQ = "autoawq@https://github.com/casper-hansen/AutoAWQ/releases/download/v0.2.1/autoawq-0.2.1+rocm571-cp310-cp310-linux_x86_64.whl"
     AUTOGPTQ = "auto-gptq@https://huggingface.github.io/autogptq-index/whl/rocm573/auto-gptq/auto_gptq-0.7.1%2Brocm5.7.3-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl"
 else:
-    AUTOAWQ = "autoawq==0.2.1"
-    AUTOGPTQ = "auto-gptq==0.7.1"
+    AUTOAWQ = "autoawq@git+https://github.com/casper-hansen/AutoAWQ.git"
+    AUTOAWQ_KERNELS = "autoawq-kernels@git+https://github.com/casper-hansen/AutoAWQ_kernels.git"
+
+    AUTOGPTQ = "auto-gptq@git+https://github.com/PanQiWei/AutoGPTQ.git"
 
 EXTRAS_REQUIRE = {
     "quality": ["ruff"],
@@ -79,7 +81,7 @@
     "py-txi": ["py-txi"],
     "vllm": ["vllm"],
     # optional dependencies
-    "autoawq": [AUTOAWQ],
+    "autoawq": [AUTOAWQ_KERNELS, AUTOAWQ],
     "auto-gptq": ["optimum", AUTOGPTQ],
     "sentence-transformers": ["sentence-transformers"],
     "bitsandbytes": ["bitsandbytes"],
@@ -112,7 +114,7 @@
         "License :: OSI Approved :: Apache Software License",
         "Topic :: Scientific/Engineering :: Artificial Intelligence",
     ],
-    keywords="benchmaek, transformers, quantization, pruning, optimization, training, inference, onnx, onnx runtime, intel, "
+    keywords="benchmark, transformers, quantization, pruning, optimization, training, inference, onnx, onnx runtime, intel, "
     "habana, graphcore, neural compressor, ipex, ipu, hpu, llm-swarm, py-txi, vllm, auto-gptq, autoawq, "
     "sentence-transformers, bitsandbytes, codecarbon, flash-attn, deepspeed, diffusers, timm, peft",
     long_description=open("README.md", "r", encoding="utf-8").read(),