Merge branch 'main' into edwardk/fix-config-embed

Author: charlesbluca

.github/workflows/ci-pull-request.yml

@@ -9,7 +9,7 @@ on:
       - labeled
 
 concurrency:
-  group: ${{ github.workflow }}-pr-${{ github.event.pull_request_target.number }}
+  group: ${{ github.workflow }}-pr-${{ github.event.pull_request.number }}
   cancel-in-progress: true
 
 jobs:
@@ -32,7 +32,7 @@ jobs:
     with:
       platform: 'linux/amd64'
       target: 'test'
-      tags: 'nv-ingest:pr-${{ github.event.pull_request_target.number }}'
+      tags: 'nv-ingest:pr-${{ github.event.pull_request.number }}'
       base-image: 'ubuntu'
       base-image-tag: 'jammy-20250415.1'
       test-selection: 'full'
@@ -54,10 +54,10 @@ jobs:
         id: check
         run: |
           HAS_ACCESS="false"
-          if [[ "${{ github.event.pull_request_target.author_association }}" == "MEMBER" ]] || \
-             [[ "${{ github.event.pull_request_target.author_association }}" == "COLLABORATOR" ]] || \
-             [[ "${{ github.event.pull_request_target.author_association }}" == "OWNER" ]] || \
-             [[ "${{ contains(github.event.pull_request_target.labels.*.name, 'ok-to-test') }}" == "true" ]]; then
+          if [[ "${{ github.event.pull_request.author_association }}" == "MEMBER" ]] || \
+             [[ "${{ github.event.pull_request.author_association }}" == "COLLABORATOR" ]] || \
+             [[ "${{ github.event.pull_request.author_association }}" == "OWNER" ]] || \
+             [[ "${{ contains(github.event.pull_request.labels.*.name, 'ok-to-test') }}" == "true" ]]; then
             HAS_ACCESS="true"
           fi
           echo "has-access=$HAS_ACCESS" >> $GITHUB_OUTPUT
@@ -72,7 +72,7 @@ jobs:
     with:
       runner: 'ubuntu-latest'
       timeout-minutes: 60
-      source-ref: ${{ github.event.pull_request_target.head.sha }}
+      source-ref: ${{ github.event.pull_request.head.sha }}
     secrets:
       AUDIO_FUNCTION_ID: ${{ secrets.AUDIO_FUNCTION_ID }}
       EMBEDDING_NIM_MODEL_NAME: ${{ secrets.EMBEDDING_NIM_MODEL_NAME }}

.github/workflows/reusable-integration-test.yml

@@ -72,6 +72,7 @@ jobs:
           uv venv .venv
           source .venv/bin/activate
           uv pip install -e ./src -e ./api -e ./client
+          # block pulling in breaking changes in pypdfium2 v5
           uv pip install \
             opencv-python \
             llama-index-embeddings-nvidia \
@@ -80,7 +81,7 @@ jobs:
             'milvus-lite==2.4.12' \
             'nvidia-riva-client==2.20.0' \
             'unstructured-client>=0.25.9' \
-            'pypdfium2<5' \ # block pulling in breaking changes in pypdfium2 v5
+            'pypdfium2<5' \
             tritonclient \
             markitdown \
             glom
@@ -109,4 +110,4 @@ jobs:
           source .venv/bin/activate
           echo 'Running integration tests...'
           unset LD_LIBRARY_PATH
-          python -m pytest -rsx tests/integration
+          python -m pytest -rsx -m integration tests/integration

nemo_retriever/src/nemo_retriever/model/local/llama_nemotron_embed_1b_v2_embedder.py

@@ -45,7 +45,7 @@ def __post_init__(self) -> None:
         MODEL_ID = self.model_id or "nvidia/llama-3.2-nv-embedqa-1b-v2"
         dev = torch.device(self.device or ("cuda" if torch.cuda.is_available() else "cpu"))
         hf_cache_dir = self.hf_cache_dir or str(Path.home() / ".cache" / "huggingface")
-        self._tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, cache_dir=hf_cache_dir)
+        self._tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True, cache_dir=hf_cache_dir)
         self._model = AutoModel.from_pretrained(MODEL_ID, trust_remote_code=True, cache_dir=hf_cache_dir)
         self._model = self._model.to(dev)
         self._model.eval()