Merge remote-tracking branch 'origin' into kylesayrs/r3-only

Author: kylesayrs

.github/actions/test/action.yml

@@ -9,8 +9,7 @@ inputs:
     required: true
   code_coverage:
     description: whether to collect code coverage metrics during test run
-    type: boolean
-    default: false
+    default: 'false'
 outputs:
   status:
     description: "final status from test"
@@ -51,7 +50,7 @@ runs:
 
           if [[ "${ENABLE_COVERAGE}" == "true" ]]; then
             echo "::group::Installing code coverage requirements via pip"
-            pip install bashlex https://github.com/neuralmagic/pytest-nm-releng/archive/v0.4.0.tar.gz
+            pip install https://github.com/neuralmagic/pytest-nm-releng/archive/v0.4.0.tar.gz
             pip install coverage pytest-cov
 
             # Adding Code coverage to the tests
@@ -76,7 +75,7 @@ runs:
             fi
             echo "::endgroup::"
           fi
-          
+
           deactivate
           exit ${SUCCESS}
       shell: bash

.github/workflows/post-release-nightly-build.yml

@@ -0,0 +1,15 @@
+name: Post-release nightly build & release
+
+on:
+  release:
+    types: [released]
+
+jobs:
+  BUILD-TEST-NIGHTLY:
+    name: Post-release nightly build & release
+    uses: ./.github/workflows/trigger-all.yml
+    with:
+      wf_category: NIGHTLY
+      push_to_pypi: true
+      gitref: main
+    secrets: inherit

.github/workflows/test-check.yaml

@@ -12,7 +12,10 @@ on:
 
 jobs:
   python-tests:
-    runs-on: ubuntu-24.04
+    runs-on: k8s-util
+    env:
+        HF_HOME: /model-cache
+        HF_TOKEN: ${{ secrets.NM_HF_TOKEN_READ_ONLY }}
     steps:
         - uses: actions/setup-python@v5
           with:
@@ -21,11 +24,13 @@ jobs:
           with:
             fetch-depth: 0
             fetch-tags: true
+        - name: Install system dependencies
+          run: |-
+            sudo apt-get update
+            sudo apt-get install -y --no-install-recommends g++ gcc make
         - name: Set Env
-          run: |
-            pip3 install --upgrade pip && pip3 install --upgrade setuptools
+          run: pip3 install --upgrade pip setuptools
         - name: "⚙️ Install dependencies"
           run: pip3 install .[dev,accelerate]
         - name: "🔬 Running tests"
           run: make test
-

.github/workflows/test.yml

@@ -71,6 +71,9 @@ jobs:
             contents: 'read'
             id-token: 'write'
             pages: 'write'
+        env:
+            HF_HOME: /model-cache
+            HF_TOKEN: ${{ secrets.NM_HF_TOKEN_READ_ONLY }}
         environment:
             name: github-pages
             url: ${{ steps.coverage.outputs.page_url }}
@@ -89,6 +92,12 @@ jobs:
               with:
                   python-version: ${{ inputs.python }}
 
+            - name: install system dependencies
+              run: |-
+                  sudo apt-get update
+                  sudo apt-get install -y --no-install-recommends g++ gcc make
+              shell: bash
+
             - name: checkout code
               id: checkout
               uses: actions/checkout@v4
@@ -162,7 +171,7 @@ jobs:
               uses: actions/upload-artifact@v4
               if: success() || failure()
               with:
-                  name: report-${{ inputs.test_label }}.xml
+                  name: report-${{ inputs.python }}-${{ inputs.test_label }}.xml
                   path: test-results/report.xml
                   retention-days: 5
 

.github/workflows/trigger-all.yml

@@ -5,6 +5,21 @@ on:
       # * is a special character in YAML so you have to quote this string
       - cron: '30 0 * * *'  # nightly run
 
+    workflow_call:
+        inputs:
+            wf_category:
+                description: "workflow category, must be 'NIGHTLY' or 'RELEASE' (default: NIGHTLY)"
+                type: string
+                default: NIGHTLY
+            push_to_pypi:
+                description: "when set and tests pass, then '.whl' & '.tar.gz' will be pushed to public pypi"
+                type: boolean
+                default: false
+            gitref:
+                description: "git commit hash or tag name"
+                type: string
+                default: 'main'
+
     workflow_dispatch:
         inputs:
             wf_category:
@@ -19,7 +34,7 @@ on:
                 type: boolean
                 default: false
             gitref:
-                description: "git commit hash or tag name" 
+                description: "git commit hash or tag name"
                 type: string
                 default: 'main'
 
@@ -32,9 +47,8 @@ jobs:
             wf_category: ${{ inputs.wf_category || 'NIGHTLY' }}
             gitref: ${{ inputs.gitref || 'main' }}
             push_to_pypi: ${{ (github.event.schedule == '30 0 * * *') || inputs.push_to_pypi || false }}
-            test_configs: '[{"python":"3.11.4","label":"ubuntu-24.04","timeout":"40","code_coverage":true},
-                            {"python":"3.10.12","label":"ubuntu-22.04","timeout":"40"},
+            test_configs: '[{"python":"3.11.4","label":"k8s-util","timeout":"40","code_coverage":true},
+                            {"python":"3.10.12","label":"k8s-util","timeout":"40"},
                             {"python":"3.9.17","label":"k8s-h100-solo","timeout":"40"},
                             {"python":"3.12.6","label":"k8s-a100-duo","timeout":"40"}]'
-
         secrets: inherit

examples/llama_1.1b/ex_llmcompressor_quantization.py

@@ -24,7 +24,7 @@
 from pathlib import Path
 
 import torch
-from llmcompressor.transformers import oneshot
+from llmcompressor import oneshot
 
 
 recipe = str(Path(__file__).parent / "example_quant_recipe.yaml")

examples/quantize_and_pack_int4.ipynb

@@ -144,7 +144,7 @@
    "outputs": [],
    "source": [
     "quantization_config_dict = {\n",
-    "\t\"quant_method\": \"sparseml\",\n",
+    "\t\"quant_method\": \"compressed-tensors\",\n",
     "\t\"format\": \"pack-quantized\",\n",
     "\t\"global_compression_ratio\": None,\n",
     "\t\"config_groups\": {\n",

setup.py

@@ -88,7 +88,7 @@ def _setup_packages() -> List:
     )
 
 def _setup_install_requires() -> List:
-    return ["torch>=1.7.0", "transformers", "pydantic>=2.0", "frozendict"]
+    return ["torch>=1.7.0", "transformers", "pydantic>=2.0", "loguru"]
 
 def _setup_extras() -> Dict:
     return {

src/compressed_tensors/__init__.py

@@ -12,9 +12,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+# flake8: noqa
+# isort: off
+from .logger import LoggerConfig, configure_logger, logger
 from .base import *
 
-# flake8: noqa
 from .compressors import *
 from .config import *
 from .quantization import QuantizationConfig, QuantizationStatus