Merge branch 'fix-quantization-lib' of https://github.com/huggingface…

…/optimum-benchmark into fix-quantization-lib

.github/workflows/build_and_publish_docker_images.yaml → .github/workflows/images.yaml

@@ -7,7 +7,7 @@ on:
       - main
     paths:
       - docker/**
-      - .github/workflows/build_and_publish_docker_images.yaml
+      - .github/workflows/images.yaml
   schedule:
     - cron: "0 0 * * *"
 
@@ -29,24 +29,27 @@ jobs:
     runs-on: ubuntu-latest
 
     permissions:
-      contents: read
+      contents: write
       packages: write
       id-token: write
 
     steps:
-      - name: Checkout
+      - name: Free Disk Space
+        uses: jlumbroso/free-disk-space@main
+
+      - name: Checkout code
         uses: actions/checkout@v4
 
       - name: Login to GitHub Container Registry
         uses: docker/login-action@v3
         with:
-          registry: ${{ env.REGISTRY }}
+          registry: ghcr.io
           username: ${{ github.actor }}
-          password: ${{ secrets.GH_TOKEN }}
+          password: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Extract metadata (tags, labels) for Docker images
         id: meta
-        uses: docker/metadata-action@v4
+        uses: docker/metadata-action@v5
         with:
           flavor: |
             latest=false

.github/workflows/quality.yaml

@@ -23,8 +23,7 @@ jobs:
       - name: Install quality requirements
         run: |
           pip install --upgrade pip
-          pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
-          pip install -e .[quality]
+          pip install ruff
 
       - name: Check style
         run: make quality

.gitignore

@@ -174,7 +174,5 @@ amdsmi/
 amd-*
 
 external_repos/
-outputs/
-
-# Mac specific
 .DS_Store
+outputs/

Makefile

@@ -123,7 +123,7 @@ install_cli_cuda_onnxruntime:
 # Run tests
 
 test_api_misc:
-	pytest -s -k "api and not (cpu or cuda)
+	pytest -s -k "api and not (cpu or cuda or mps)"
 
 test_api_cpu:
 	pytest -s -k "api and cpu"
@@ -135,7 +135,7 @@ test_api_rocm:
 	pytest -s -k "api and cuda"
 
 test_cli_misc:
-	pytest -s -k "cli and not (cpu or cuda)"
+	pytest -s -k "cli and not (cpu or cuda or mps)"
 
 test_cli_cpu_pytorch:
 	pytest -s -k "cli and cpu and pytorch"

optimum_benchmark/backends/neural_compressor/config.py

@@ -37,7 +37,6 @@
     "quant_level": "auto",
     "accuracy_criterion": ACCURACY_CRITERION_CONFIG,
     "tuning_criterion": TUNING_CRITERION_CONFIG,
-    "diagnosis": False,
 }
 
 

optimum_benchmark/system_utils.py

@@ -95,7 +95,10 @@ def get_gpus():
         pynvml.nvmlInit()
         for i in range(pynvml.nvmlDeviceGetCount()):
             handle = pynvml.nvmlDeviceGetHandleByIndex(i)
-            gpus.append(pynvml.nvmlDeviceGetName(handle))
+            gpu = pynvml.nvmlDeviceGetName(handle)
+            # Older pynvml versions may return bytes
+            gpu = gpu.decode("utf-8") if isinstance(gpu, bytes) else gpu
+            gpus.append(gpu)
         pynvml.nvmlShutdown()
 
     elif is_rocm_system():