Merge pull request #54 from michaelfeil/docker-dependency-update

update dockerfile and tests

.github/workflows/release_docker_container.yaml

@@ -70,6 +70,7 @@ jobs:
           # We can only build for the intersection of platforms supported by
           # QEMU and base python image, for now build only for
           # linux/amd64 and linux/arm64
+          cache-from: type=registry,ref=${{ env.LATEST_TAG }}
           platforms: linux/amd64,linux/arm64
           tags: ${{ env.LATEST_TAG }},${{ env.VERSION_TAG }}
           push: true

libs/infinity_emb/.dockerignore

@@ -1,5 +1,3 @@
 .venv
-tests
 **/*_cache
-Dockerfile
 **/__pycache__

libs/infinity_emb/Dockerfile

@@ -1,9 +1,7 @@
 # Use the Python base image
-FROM nvidia/cuda:11.8.0-base-ubuntu22.04 AS base
+FROM nvidia/cuda:12.1.0-base-ubuntu22.04 AS base
 
 ENV PYTHONUNBUFFERED=1 \
-    # prevents python creating .pyc files
-    PYTHONDONTWRITEBYTECODE=1 \
     \
     # pip
     PIP_NO_CACHE_DIR=off \
@@ -18,11 +16,11 @@ ENV PYTHONUNBUFFERED=1 \
 
 RUN apt-get update && apt-get install python3.10 curl -y 
 # python3.10-venv 
+WORKDIR /app
 
 FROM base as builder
 
 # Set the working directory for the app
-WORKDIR /app
 
 # Define the version of Poetry to install (default is 1.6.1)
 # Define the directory to install Poetry to (default is /opt/poetry)
@@ -38,20 +36,34 @@ ENV PATH=$POETRY_HOME/bin:$PATH
 RUN echo "Poetry version:" && poetry --version
 
 # Copy the rest of the app source code (this layer will be invalidated and rebuilt whenever the source code changes)
-COPY poetry.lock poetry.toml pyproject.toml README.md .
+COPY poetry.lock poetry.toml pyproject.toml README.md /app/
+# Install dependencies only
+RUN poetry install --no-interaction --no-ansi --no-root --extras all --without lint,test
 COPY infinity_emb infinity_emb
-
-# Install dependencies and project
-RUN poetry config virtualenvs.create false
-RUN poetry install --no-interaction --no-ansi --extras all
+# Install dependency with infinity_emb package
+RUN poetry install --no-interaction --no-ansi --extras all --without lint,test
 # remove cache
 RUN poetry cache clear pypi --all
 
+FROM builder as testing
+# install lint and test dependencies
+RUN poetry install --no-interaction --no-ansi --extras all
+# lint 
+RUN poetry run ruff .
+RUN poetry run black --check .
+RUN poetry run mypy .
+# pytest
+COPY tests tests
+RUN poetry run python -m pytest -x
+# write a file 
+RUN echo "all tests passed" > "test_results.txt"
+
 # Use a multi-stage build -> production version
-from base AS production
+FROM base AS production
 
 COPY --from=builder /app /app
-WORKDIR /app
+# force testing stage to run
+COPY --from=testing /app/test_results.txt /app/test_results.txt
 
 ENV SENTENCE_TRANSFORMERS_HOME=/app/.cache/torch
 ENV PATH=/app/.venv/bin:$PATH

libs/infinity_emb/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "infinity_emb"
-version = "0.0.16"
+version = "0.0.17"
 description = "Infinity is a high-throughput, low-latency REST API for serving vector embeddings, supporting a wide range of sentence-transformer models and frameworks."
 authors = ["michaelfeil <[email protected]>"]
 license = "MIT"

libs/infinity_emb/tests/end_to_end/conftest.py

@@ -67,7 +67,7 @@ async def _post_batch(inputs):
             cosine_sim = np.dot(r, e) / (np.linalg.norm(e) * np.linalg.norm(r))
             assert cosine_sim > 0.95
         np.testing.assert_almost_equal(
-            np.array(responses), np.array(encodings), decimal=0
+            np.array(responses), np.array(encodings), decimal=3
         )
         assert time_api / time_st < 2.5
 
@@ -93,7 +93,9 @@ async def embedding_verify(client, model_base, prefix, model_name):
             want_embeddings = model_base.encode(inp)
 
             for embedding, st_embedding in zip(rdata["data"], want_embeddings):
-                np.testing.assert_almost_equal(embedding["embedding"], st_embedding)
+                np.testing.assert_almost_equal(
+                    embedding["embedding"], st_embedding, decimal=3
+                )
 
 
 @pytest.fixture

libs/infinity_emb/tests/end_to_end/test_ct2_sentence.py

@@ -58,13 +58,13 @@ async def test_model_route(client):
 
 @pytest.mark.anyio
 async def test_embedding(client, model_base, helpers):
-    helpers.embedding_verify(client, model_base, prefix=PREFIX, model_name=MODEL)
+    await helpers.embedding_verify(client, model_base, prefix=PREFIX, model_name=MODEL)
 
 
 @pytest.mark.performance
 @pytest.mark.anyio
 async def test_batch_embedding(client, get_sts_bechmark_dataset, model_base, helpers):
-    helpers.util_batch_embedding(
+    await helpers.util_batch_embedding(
         client=client,
         sts_bechmark_dataset=get_sts_bechmark_dataset,
         model_base=model_base,

libs/infinity_emb/tests/end_to_end/test_fastembed.py

@@ -44,13 +44,13 @@ async def test_model_route(client):
 
 @pytest.mark.anyio
 async def test_embedding(client, model_base, helpers):
-    helpers.embedding_verify(client, model_base, prefix=PREFIX, model_name=MODEL)
+    await helpers.embedding_verify(client, model_base, prefix=PREFIX, model_name=MODEL)
 
 
 @pytest.mark.performance
 @pytest.mark.anyio
 async def test_batch_embedding(client, get_sts_bechmark_dataset, model_base, helpers):
-    helpers.util_batch_embedding(
+    await helpers.util_batch_embedding(
         client=client,
         sts_bechmark_dataset=get_sts_bechmark_dataset,
         model_base=model_base,

libs/infinity_emb/tests/end_to_end/test_sentence_transformers.py

@@ -50,13 +50,13 @@ async def test_model_route(client):
 
 @pytest.mark.anyio
 async def test_embedding(client, model_base, helpers):
-    helpers.embedding_verify(client, model_base, prefix=PREFIX, model_name=MODEL)
+    await helpers.embedding_verify(client, model_base, prefix=PREFIX, model_name=MODEL)
 
 
 @pytest.mark.performance
 @pytest.mark.anyio
 async def test_batch_embedding(client, get_sts_bechmark_dataset, model_base, helpers):
-    helpers.util_batch_embedding(
+    await helpers.util_batch_embedding(
         client=client,
         sts_bechmark_dataset=get_sts_bechmark_dataset,
         model_base=model_base,

libs/infinity_emb/tests/unit_test/inference/test_models.py

@@ -47,8 +47,8 @@ def _pretrained_model_score(
         ("sentence-transformers/all-MiniLM-L6-v2", 82.03, "default"),
         ("sentence-transformers/all-MiniLM-L6-v2", 81.73, "int8"),
         ("sentence-transformers/all-MiniLM-L6-v2", 82.03, "default"),
-        ("BAAI/bge-small-en-v1.5", 86.03, None),
-        ("BAAI/bge-small-en-v1.5", 86.03, "int8"),
+        ("BAAI/bge-small-en-v1.5", 86.00, None),
+        ("BAAI/bge-small-en-v1.5", 86.00, "int8"),
     ],
 )
 def test_bert(get_sts_bechmark_dataset, model, score, compute_type):

libs/infinity_emb/tests/unit_test/transformer/classifier/test_torch_classifer.py

@@ -3,11 +3,9 @@
 from infinity_emb.transformer.classifier.torch import SentenceClassifier
 
 
-def test_classifier():
-    model = SentenceClassifier("SamLowe/roberta-base-go_emotions")
-    pipe = pipeline(
-        model="SamLowe/roberta-base-go_emotions", task="text-classification"
-    )
+def test_classifier(model_name: str = "SamLowe/roberta-base-go_emotions"):
+    model = SentenceClassifier(model_name)
+    pipe = pipeline(model=model_name, task="text-classification")
 
     sentences = ["This is awesome.", "I am depressed."]
 
@@ -30,6 +28,8 @@ def test_classifier():
 
     for pred_orig, pred in zip(preds_orig, preds):
         assert len(pred_orig) == len(pred)
-        for pred_orig_i, pred_i in zip(pred_orig, pred):
-            assert pred_orig_i["label"] == pred_i["label"]
-            assert abs(pred_orig_i["score"] - pred_i["score"]) < 0.1
+        for pred_orig_i, pred_i in zip(pred_orig[:5], pred[:5]):
+            assert abs(pred_orig_i["score"] - pred_i["score"]) < 0.05
+
+            if pred_orig_i["score"] > 0.005:
+                assert pred_orig_i["label"] == pred_i["label"]