[ci] update docker workflow (#9231)

.github/workflows/docker.yml

@@ -18,6 +18,9 @@ on:
       - "requirements.txt"
       - "docker/**"
       - ".github/workflows/*.yml"
+  release:
+    types:
+      - published
 
 jobs:
   build:
@@ -40,7 +43,7 @@ jobs:
 
     steps:
       - name: Free up disk space
-        uses: jlumbroso/free-disk-space@54081f138730dfa15788a46383842cd2f914a1be # v1.3.1
+        uses: jlumbroso/free-disk-space@v1.3.1
         with:
           tool-cache: true
           docker-images: false
@@ -51,12 +54,16 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.9"
+          python-version: "3.10"
 
       - name: Get llamafactory version
         id: version
         run: |
-          echo "tag=$(python setup.py --version | sed 's/\.dev0//')" >> "$GITHUB_OUTPUT"
+          if [ "${{ github.event_name }}" = "release" ]; then
+            echo "tag=$(python setup.py --version)" >> "$GITHUB_OUTPUT"
+          else
+            echo "tag=latest" >> "$GITHUB_OUTPUT"
+          fi
 
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
@@ -86,7 +93,6 @@ jobs:
             EXTRAS=metrics,deepspeed,liger-kernel
           push: ${{ github.event_name != 'pull_request' }}
           tags: |
-            docker.io/hiyouga/llamafactory:latest
             docker.io/hiyouga/llamafactory:${{ steps.version.outputs.tag }}
           cache-from: type=gha
           cache-to: type=gha,mode=max
@@ -100,9 +106,7 @@ jobs:
           file: ./docker/docker-npu/Dockerfile
           push: ${{ github.event_name != 'pull_request' }}
           tags: |
-            docker.io/hiyouga/llamafactory:latest-npu-a2
             docker.io/hiyouga/llamafactory:${{ steps.version.outputs.tag }}-npu-a2
-            quay.io/ascend/llamafactory:latest-npu-a2
             quay.io/ascend/llamafactory:${{ steps.version.outputs.tag }}-npu-a2
           cache-from: type=gha
           cache-to: type=gha,mode=max

.github/workflows/tests.yml

@@ -30,7 +30,7 @@ jobs:
         os:
           - "ubuntu-latest"
           - "windows-latest"
-          - "macos-13"
+          - "macos-latest"
         transformers:
           - null
         include:  # test backward compatibility
@@ -75,10 +75,11 @@ jobs:
         run: |
           python -m pip install "transformers==${{ matrix.transformers }}"
 
-      - name: Install transformers to avoid mac os ci errors
+      - name: Update accelerate to avoid mac os ci errors (before accelerate 1.11.0)
-        if: ${{ matrix.os == 'macos-13' }}
+        if: ${{ matrix.os == 'macos-latest' }}
         run: |
-          python -m pip install "transformers<=4.51.3"
+          python -m pip uninstall -y accelerate
+          python -m pip install "git+https://github.com/huggingface/accelerate.git"
 
       - name: Cache files
         id: hf-hub-cache

requirements.txt

@@ -1,7 +1,8 @@
 # core deps
-transformers>=4.49.0,<=4.56.2,!=4.52.0
+transformers>=4.49.0,<=4.56.2,!=4.52.0; python_version < '3.10'
+transformers>=4.49.0,<=4.57.0,!=4.52.0; python_version >= '3.10'
 datasets>=2.16.0,<=4.0.0
-accelerate>=1.3.0,<=1.10.1
+accelerate>=1.3.0,<=1.11.0
 peft>=0.14.0,<=0.17.1
 trl>=0.8.6,<=0.9.6
 # gui

src/llamafactory/extras/misc.py

@@ -94,9 +94,9 @@ def check_version(requirement: str, mandatory: bool = False) -> None:
 
 def check_dependencies() -> None:
     r"""Check the version of the required packages."""
-    check_version("transformers>=4.49.0,<=4.56.2")
+    check_version("transformers>=4.49.0,<=4.57.0")
     check_version("datasets>=2.16.0,<=4.0.0")
-    check_version("accelerate>=1.3.0,<=1.10.1")
+    check_version("accelerate>=1.3.0,<=1.11.0")
     check_version("peft>=0.14.0,<=0.17.1")
     check_version("trl>=0.8.6,<=0.9.6")
 

src/llamafactory/model/model_utils/moe.py

@@ -40,7 +40,7 @@ def add_z3_leaf_module(model: "PreTrainedModel") -> None:
 
     model_type = getattr(model.config, "model_type", None)
     text_config = getattr(model.config, "text_config", None)
-    text_architectures = getattr(text_config, "architectures", None)
+    text_model_type = getattr(text_config, "model_type", None)
 
     if model_type == "dbrx":
         from transformers.models.dbrx.modeling_dbrx import DbrxFFN
@@ -105,11 +105,21 @@ def add_z3_leaf_module(model: "PreTrainedModel") -> None:
 
         _set_z3_leaf_modules(model, [Qwen2MoeSparseMoeBlock])
 
-    if model_type == "qwen3_moe" or text_architectures == "Qwen3MoeForCausalLM":
+    if model_type == "qwen3_moe" or text_model_type == "qwen3_moe":  # internvl 3.5
         from transformers.models.qwen3_moe.modeling_qwen3_moe import Qwen3MoeSparseMoeBlock
 
         _set_z3_leaf_modules(model, [Qwen3MoeSparseMoeBlock])
 
+    if model_type == "qwen3_vl_moe":
+        from transformers.models.qwen3_vl_moe.modeling_qwen3_vl_moe import Qwen3VLMoeTextSparseMoeBlock
+
+        _set_z3_leaf_modules(model, [Qwen3VLMoeTextSparseMoeBlock])
+
+    if model_type == "qwen3_omni_moe":
+        from transformers.models.qwen3_omni_moe.modeling_qwen3_omni_moe import Qwen3OmniMoeThinkerTextSparseMoeBlock
+
+        _set_z3_leaf_modules(model, [Qwen3OmniMoeThinkerTextSparseMoeBlock])
+
 
 def configure_moe(config: "PretrainedConfig", model_args: "ModelArguments", is_trainable: bool) -> None:
     if not is_trainable or not model_args.moe_aux_loss_coef:

tests/data/test_mm_plugin.py

@@ -376,7 +376,7 @@ def test_qwen2_vl_plugin():
 @pytest.mark.skipif(not is_transformers_version_greater_than("4.57.0"), reason="Requires transformers>=4.57.0")
 def test_qwen3_vl_plugin():
     frame_seqlen = 1
-    tokenizer_module = _load_tokenizer_module(model_name_or_path="Qwen/Qwen3-VL-235B-A22B-Instruct")
+    tokenizer_module = _load_tokenizer_module(model_name_or_path="Qwen/Qwen3-VL-30B-A3B-Instruct")
     qwen3_vl_plugin = get_mm_plugin(name="qwen3_vl", video_token="<|video_pad|>")
     check_inputs = {"plugin": qwen3_vl_plugin, **tokenizer_module}
     check_inputs["expected_mm_messages"] = [

tests/version.txt

@@ -1,2 +1,2 @@
 # change if test fails or cache is outdated
-0.9.4.102
+0.9.4.103