Work around offline bug in optimum

Optimum raises if you try to use offline mode with hf hub.

huggingface/optimum#1796

Dockerfile.cache_test

@@ -4,7 +4,6 @@ WORKDIR /opt/whylogs-container
 RUN chown -R whylabs:whylabs /opt/whylogs-container
 USER whylabs
 
-ENV LLM_CONTAINER=True
 ENV CONTAINER_CACHE_BASE=/opt/whylogs-container/.cache
 ENV HF_HOME=$CONTAINER_CACHE_BASE/hf_home/
 ENV NLTK_DATA=$CONTAINER_CACHE_BASE/nltk_data/
@@ -38,9 +37,15 @@ RUN poetry config virtualenvs.in-project true
 RUN poetry install --no-root --extras "all" --without dev
 RUN rm -rf .venv/lib/python3.10/site-packages/pandas/tests # Pandas deploys a ton of tests to pypi
 
+ENV TRANSFORMERS_VERBOSITY=debug
 
 copy ./langkit ./langkit
 RUN bash -c "source .venv/bin/activate; python -m langkit.scripts.langkit_cache"
+RUN find  $CONTAINER_CACHE_BASE/
+
 # This step will fail if any network requests happen
+ENV TRANSFORMERS_OFFLINE=1
+ENV HF_DATASETS_OFFLINE=1
+ENV HF_HUB_OFFLINE=1
 RUN --network=none bash -c "source .venv/bin/activate; python -m langkit.scripts.langkit_cache --skip-downloads"
 

langkit/metrics/topic_onnx.py

@@ -2,6 +2,7 @@
 # pyright: reportUnknownVariableType=none
 # pyright: reportUnknownLambdaType=none
 
+import os
 from dataclasses import dataclass
 from functools import lru_cache, partial
 from typing import List, Optional, TypedDict
@@ -23,26 +24,46 @@
 
 _hypothesis_template = "This example is about {}"
 
+_model = "MoritzLaurer/xtremedistil-l6-h256-zeroshot-v1.1-all-33"
+_revision = "dea69e79cd6063916d08b883ea8a3c1823fd10b4"
+
 
 def _download_assets():
     ORTModelForSequenceClassification.from_pretrained(
-        "MoritzLaurer/xtremedistil-l6-h256-zeroshot-v1.1-all-33",
+        _model,
         subfolder="onnx",
         file_name="model.onnx",
+        revision=_revision,
         export=False,
     )
+    AutoTokenizer.from_pretrained(_model, revision=_revision)
 
 
 def _get_tokenizer() -> PreTrainedTokenizerBase:
-    return AutoTokenizer.from_pretrained("MoritzLaurer/xtremedistil-l6-h256-zeroshot-v1.1-all-33")
+    return AutoTokenizer.from_pretrained(_model, revision=_revision, local_files_only=True)
 
 
 def _get_model() -> PreTrainedModel:
+    # return ORTModelForSequenceClassification.from_pretrained(
+    #     _model,
+    #     subfolder="onnx",
+    #     file_name="model.onnx",
+    #     export=False,
+    #     revision=_revision,
+    #     local_files_only=True,
+    # )
+    # Optimum doesn't support offline mode https://github.com/huggingface/optimum/issues/1796
+    # workaround for now is to reference the actual model path after caching it. Uncomment the above code when the issue is resolved
+
+    model_name = _model.replace("/", "--")
+    home_dir = os.path.expanduser("~")
+    base = os.environ.get("HF_HOME", os.path.join(home_dir, ".cache/huggingface"))
+    model_path = f"{base}/hub/models--{model_name }/snapshots/{_revision}"
     return ORTModelForSequenceClassification.from_pretrained(
-        "MoritzLaurer/xtremedistil-l6-h256-zeroshot-v1.1-all-33",
-        subfolder="onnx",
-        file_name="model.onnx",
+        model_path,
+        file_name="onnx/model.onnx",
         export=False,
+        revision=_revision,
         local_files_only=True,
     )