Merge branch 'main' into CSY/fix-to

Author: CSY-ModelCloud

.github/scripts/ci_release_build_sdist.sh

@@ -1,5 +1,6 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
-python -m pip install -U build twine setuptools
+python -m pip install uv
+uv pip install -U build twine setuptools
 python -m build --sdist

.github/scripts/ci_release_test_install.sh

@@ -4,6 +4,7 @@ set -euo pipefail
 pkg_name="${1:?package filename is required}"
 venv_dir="${2:-local_uv_env}"
 
-uv venv "$venv_dir"
+python -m venv "$venv_dir"
 source "$venv_dir/bin/activate"
-uv pip install "dist/$pkg_name" torch
+pip install uv
+uv pip install "dist/$pkg_name"

.github/scripts/ci_write_runner_outputs.sh

@@ -3,23 +3,19 @@ set -euo pipefail
 
 runner_ip="${1:?runner ip is required}"
 github_run_id="${2:?github run id is required}"
-artifact_id="${3:-}"
-max_parallel="${4:-}"
+max_parallel="${3:-}"
 
 if [[ -z "${GITHUB_OUTPUT:-}" ]]; then
   echo "GITHUB_OUTPUT is required" >&2
   exit 1
 fi
 
 run_id="$github_run_id"
-if [[ -n "$artifact_id" ]]; then
-  run_id="$artifact_id"
-fi
 
 echo "ip=$runner_ip" >> "$GITHUB_OUTPUT"
 echo "ip: $runner_ip"
 echo "run_id=$run_id" >> "$GITHUB_OUTPUT"
-echo "artifact_id=$run_id"
+echo "run_id=$run_id"
 
 if [[ -n "$max_parallel" ]]; then
   max_parallel_json="{\"size\": ${max_parallel:-20}}"

.github/scripts/deps.yaml

@@ -107,6 +107,9 @@ tests:
     - gguf
 
 tests/models:
+  test_chatglm:
+    - sentencepiece
+
   test_cohere2:
     - jieba
 
@@ -123,6 +126,7 @@ tests/models:
     - http://10.0.13.31/files/causal_conv1d-1.6.0-cp313-cp313-linux_x86_64.whl
     - mamba_ssm
     - tiktoken
+    - attn_gym
 
   test_internlm.py:
     - transformers<=4.44.2
@@ -137,10 +141,26 @@ tests/models:
     - backoff>=2.2.1
     - optimum>=1.24.0
     - scipy
+    - backoff
 
   test_llama3_2_exllamav3.py:
     - pydantic
 
+  test_internvl_chat.py:
+    - timm
+    - sentencepiece
+    - tiktoken
+
+  test_ling_2_6_flash_support:
+    - flash-linear-attention
+
+test_nemotron_omni_support.py:
+    - einops
+    - timm
+    - open-clip-torch
+    - librosa
+    - mamba_ssm
+
 tests/models/awq:
   test_qwen3_8b_base_awq.py:
     - peft

.github/scripts/test.yaml

@@ -16,6 +16,9 @@ tests:
 tests/models:
   py: 3.13
 
+  test_llama3_2_fp8:
+    sm: '12.0'
+
 tests/kernels:
   test_asymmetric_real_models:
     sm: '9.0'

.github/workflows/release.yml

@@ -54,59 +54,57 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  check-vm:
-    runs-on: [ self-hosted, linux ]
-    container:
-      image: modelcloud/gptqmodel:alpine-ci-v1
-    outputs:
-      ip: ${{ steps.get_ip.outputs.ip }}
-      run_id: ${{ steps.get_ip.outputs.run_id }}
-    if: ${{ inputs.github_vm == false }}
-    steps:
-      - name: Checkout Codes
-        uses: actions/checkout@v6
-        with:
-          repository: ${{ env.repo }}
-          ref: ${{ env.ref }}
-
-      - name: Print env
-        run: |
-          echo "event name: ${{ github.event_name }}"
-          echo "repo: ${{ env.repo }}"
-          echo "ref: ${{ env.ref }}"
-          echo "upload_release: ${{ inputs.upload_release }}"
-          echo "upload_pypi: ${{ inputs.upload_pypi }}"
-
-      - name: Select server
-        id: get_ip
-        run: |
-          bash .github/scripts/ci_write_runner_outputs.sh "$RUNNER" "${{ github.run_id }}"
-
   release-source:
     permissions:
       contents: write
     runs-on: [ self-hosted, xeon5 ]
-    needs:
-      - check-vm
     if: ${{ inputs.github_vm == false }}
+    env:
+      UV_TORCH_BACKEND: cu130
+      TORCH_VERSION: 2.11.0
+      PYTHON_VERSION: 3.14
+      UV_PYTHON: 3.14
     container:
-      image: ${{ needs.check-vm.outputs.ip }}:5000/nvidia/cuda:130-ubuntu24.04_0415
+      image: 10.0.13.31:5000/nvidia/cuda:130-ubuntu24.04_0415
       volumes:
         - /monster/ci/env/entrypoint.sh:/etc/profile.d/01-entrypoint.sh
-        - /monster/ci/uv:/opt/uv
+        - /github/workspace/uv:/opt/uv
+        - /github/workspace/tmp:/opt/uv/tmp
+        - /monster/ci/uv/python:/opt/uv/python
+        - /monster/ci/uv/cache/python:/opt/uv/cache/python
+        - /monster/ci/uv/setup_uv_venv.sh:/opt/uv/setup_uv_venv.sh
+        - /monster/ci/uv/uv:/opt/uv/uv
+        - /monster/ci/uv/uvx:/opt/uv/uvx
+        - /monster/ci/uv/env:/opt/uv/env
+        - /monster/ci/uv/uv.toml:/opt/uv/uv.toml
+        - /monster/ci/env:/opt/env
     steps:
       - name: Checkout Codes
         uses: actions/checkout@v6
         with:
           repository: ${{ env.repo }}
           ref: ${{ env.ref }}
 
+      - name: Activate uv env
+        run: |
+          echo "::group::-- dust uv cache dir --"
+            dust $(uv cache dir)
+          echo "::endgroup::"
+
+          bash .github/scripts/ci_unit_activate_uv_env.sh \
+            "gptqmodel_py314t_release_source" \
+            130 \
+            "${{ env.TORCH_VERSION }}" \
+            "${{ github.run_id }}" \
+            "${{ github.run_attempt }}"
+
       - name: Setup release env
         run: |
-          export UV_PYTHON=3.14
-          echo "UV_PYTHON=3.14" >> "$GITHUB_ENV"
+          echo "::group::-- uv envs --"
+            printenv | grep UV
+          echo "::endgroup::"
 
-          env_name="gptqmodel_py314_release_source"
+          env_name="gptqmodel_py314t_release_source"
           /opt/uv/setup_uv_venv.sh $env_name
 
       - name: Run shared release source flow

.github/workflows/unit_tests.yml

@@ -30,10 +30,6 @@ on:
         description: 'Regex to filter test files'
         required: false
         default: ''
-      artifact_id:
-        description: 'Run id for artifact to be downloaded'
-        required: false
-        default: ''
       max-parallel:
         description: 'Parallel jobs'
         required: false
@@ -113,7 +109,6 @@ jobs:
         run: |
           echo "repo: ${{ env.repo }}"
           echo "ref: ${{ env.ref }}"
-          echo "artifact_id: ${{ github.event.inputs.artifact_id }}"
           echo "selected server: ${{ github.event.inputs.server }}"
 
       - name: Set server
@@ -122,7 +117,6 @@ jobs:
           bash .github/scripts/ci_write_runner_outputs.sh \
             "$RUNNER" \
             "${{ github.run_id }}" \
-            "${{ github.event.inputs.artifact_id }}" \
             "${{ github.event.inputs['max-parallel'] }}"
 
   list-test-files:

.github/workflows/unit_tests_reusable.yml

@@ -71,6 +71,7 @@ env:
   repo: ${{ inputs.repo || github.repository }}
   ref: ${{ inputs.ref || github.ref }}
   HF_TOKEN: ${{ secrets.HF_TOKEN }}
+  DEBUG_ON: 1
 
 jobs:
   test:
@@ -244,7 +245,7 @@ jobs:
           rm -rf ./* .[^.] .??* || true
           echo "cleaning venv: ${{ env.VIRTUAL_ENV }}"
           rm -rf "${{ env.VIRTUAL_ENV }}"
-          
+
           echo "::group::-- dust uv cache dir --"
             dust $(uv cache dir)
           echo "::endgroup::"

.gitignore

@@ -194,4 +194,6 @@ example.py
 /gptqmodel_ext/machete/generated/
 AGENT.md
 AGENTs.md
-.codex
+.codex
+temp
+tmp