Merge pull request #274 from michaelfeil/ci-nightly

add nightly and pypi pull
michaelfeil · Jun 20, 2024 · b4dab11 · b4dab11
2 parents bb7e700 + c9fb19b
commit b4dab11
Show file tree

Hide file tree

Showing 10 changed files with 55 additions and 32 deletions.
diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
@@ -12,6 +12,8 @@ on:
     branches:
     - main
     - dev
+  schedule: # nightly
+    - cron: '0 0 * * *'
   workflow_dispatch:  # Allows to trigger the workflow manually in GitHub UI
 
 # If another push to the same PR or branch happens while this workflow is still running,

diff --git a/.github/workflows/docs.yaml b/.github/workflows/docs.yaml
@@ -12,16 +12,16 @@ jobs:
   deploy:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
       - name: Configure Git Credentials
         run: |
           git config user.name github-actions[bot]
           git config user.email 41898282+github-actions[bot]@users.noreply.github.com
       - uses: actions/setup-python@v5
         with:
           python-version: 3.11
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
       - run: echo "cache_id=$(date --utc '+%V')" >> $GITHUB_ENV 
       - uses: actions/cache@v4
         with:

diff --git a/.github/workflows/release_modal_com.yaml b/.github/workflows/release_modal_com.yaml
@@ -8,7 +8,7 @@ on:
       unused:
         required: false
         type: string
-        description: "From which folder this pipeline executes"
+        description: "To be replaced."
 
 jobs:
   modal:
@@ -46,4 +46,5 @@ jobs:
       - name: Modal Deploy
         run: |
           cd infra/modal
+          INFINITY_VERSION=$ {{ steps.get-latest-tag.outputs.tag }}
           modal deploy --tag ${{ steps.get-latest-tag.outputs.tag }}-github-ci --env prod webserver.py
diff --git a/.github/workflows/test.yaml b/.github/workflows/test.yaml
@@ -11,6 +11,7 @@ on:
         type: boolean
         default: true
 
+
 env:
   POETRY_VERSION: "1.7.1"
 
@@ -23,13 +24,14 @@ jobs:
     strategy:
       # fail-fast: true
       matrix:
-        os: [ubuntu-latest, windows-latest, ] # macos-latest
+        os: [ubuntu-latest, windows-latest] # macos-latest
         python-version:
           - "3.9"
           - "3.10"
           - "3.11"
           - "3.12"
         coverage_tests: ["unit_test", "end_to_end"]
+        source: ["local", "pypi"]
         exclude:
           # Exclude unit tests on macOS due to compatibility issues
           - python-version: "3.9"
@@ -48,7 +50,7 @@ jobs:
 
 
     runs-on: ${{ matrix.os }}
-    name: Py${{ matrix.python-version }}-${{ matrix.os }}-${{ matrix.coverage_tests }}
+    name: Py${{ matrix.python-version }}-${{ matrix.os }}-${{ matrix.coverage_tests }}-${{ matrix.source }}
     steps:
       - name: Free Disk Space
         uses: jlumbroso/[email protected]
@@ -70,11 +72,20 @@ jobs:
           python-version: ${{ matrix.python-version }}
           poetry-version: ${{ env.POETRY_VERSION }}
           working-directory: ${{ inputs.working-directory }}
-          cache-key: infinity-emb-test-${{ matrix.coverage_tests }}
+          cache-key: infinity-emb-test-${{ matrix.coverage_tests }}-${{ matrix.source }}
 
-      - name: Install dependencies
+      - name: Install dependencies from local
+        # only if source is local
+        if: matrix.source == 'local'
         shell: bash
         run: poetry install --extras all --with test
+
+      - name: Install from PyPI
+        # only if source is pypi
+        if: matrix.source == 'pypi'
+        run: |
+          poetry install --only test
+          poetry run pip install infinity-emb[all] --upgrade
 
       - name: Run Pytest Coverage
         run: |

diff --git a/README.md b/README.md
@@ -233,7 +233,7 @@ asyncio.run(classifier(array["SamLowe/roberta-base-go_emotions"]))
 
 Example models:
 - [ProsusAI/finbert](https://huggingface.co/ProsusAI/finbert)
-- [SamLowe/roberta-base-go_emotions](SamLowe/roberta-base-go_emotions)
+- [SamLowe/roberta-base-go_emotions](https://huggingface.co/SamLowe/roberta-base-go_emotions)
 
 ## Integrations:
 - [Serverless deployments at Runpod](https://github.com/runpod-workers/worker-infinity-embedding)

diff --git a/docs/benchmarks/simple_app.py b/docs/benchmarks/simple_app.py
@@ -61,7 +61,7 @@ def encode_sentence_transformer(text: list[str]):
 
 
 async def encode_infinity(text: list[str]):
-    return (await engine.embed(text))[0]
+    return (await engine.embed(sentences=text))[0]
 
 
 @asynccontextmanager

diff --git a/docs/docs/python_engine.md b/docs/docs/python_engine.md
@@ -134,5 +134,5 @@ asyncio.run(classifier(array["SamLowe/roberta-base-go_emotions"]))
 
 Example models:
 - [ProsusAI/finbert](https://huggingface.co/ProsusAI/finbert)
-- [SamLowe/roberta-base-go_emotions](SamLowe/roberta-base-go_emotions)
+- [SamLowe/roberta-base-go_emotions](https://huggingface.co/SamLowe/roberta-base-go_emotions)
 
diff --git a/infra/modal/functional.py b/infra/modal/functional.py
@@ -4,8 +4,10 @@
 """
 
 from modal import App, Image, build, enter, method
+import os
 
-image = Image.from_registry("michaelf34/infinity:0.0.45").entrypoint([]).pip_install("timm")
+INFINITY_VERSION = os.environ.get("INFINITY_VERSION", "0.0.46")
+image = Image.from_registry(f"michaelf34/infinity:{INFINITY_VERSION}").entrypoint([])
 
 app = App("infinity-functional", image=image)
 
@@ -15,6 +17,7 @@
 
 class BaseInfinityModel:
     """Base class for the infinity model."""
+
     def _get_array(self):
         return AsyncEngineArray.from_args(
             [
@@ -39,10 +42,10 @@ async def enter(self):
 @app.cls(gpu="any", allow_concurrent_inputs=500)
 class InfinityModal(BaseInfinityModel):
     """Model for embedding text (via clip or Bert) and images (via clip) ."""
+
     def __init__(self, model_id: tuple[str]) -> None:
         self.model_id = model_id
         super().__init__()
-
 
     @method()
     async def embed(self, sentences: list[str], model: str | int = 0):
@@ -68,24 +71,32 @@ async def classify(self, sentences: list[str], model: str | int = 0):
         classes, usage = await engine.classify(sentences=sentences)
         return classes
 
+
 @app.local_entrypoint()
 def main():
-    model_id=(
-            "jinaai/jina-clip-v1",
-            "michaelfeil/bge-small-en-v1.5",
-            "mixedbread-ai/mxbai-rerank-xsmall-v1",
-            "philschmid/tiny-bert-sst2-distilled",
-        )
-    deployment = InfinityModal(
-        model_id=model_id
+    model_id = (
+        "jinaai/jina-clip-v1",
+        "michaelfeil/bge-small-en-v1.5",
+        "mixedbread-ai/mxbai-rerank-xsmall-v1",
+        "philschmid/tiny-bert-sst2-distilled",
     )
+    deployment = InfinityModal(model_id=model_id)
     embeddings_1 = deployment.embed.remote(sentences=["hello world"], model=model_id[1])
-    embeddings_2 = deployment.image_embed.remote(urls=["http://images.cocodataset.org/val2017/000000039769.jpg"], model=model_id[0])
+    embeddings_2 = deployment.image_embed.remote(
+        urls=["http://images.cocodataset.org/val2017/000000039769.jpg"],
+        model=model_id[0],
+    )
+
+    rerankings_1 = deployment.rerank.remote(
+        query="Where is Paris?",
+        docs=["Paris is the capital of France.", "Berlin is a city in Europe."],
+        model=model_id[2],
+    )
+
+    classifications_1 = deployment.classify.remote(
+        sentences=["I feel great today!"], model=model_id[3]
+    )
 
-    rerankings_1 =deployment.rerank.remote(query="Where is Paris?", docs=["Paris is the capital of France.", "Berlin is a city in Europe."], model=model_id[2])
-
-    classifications_1 = deployment.classify.remote(sentences=["I feel great today!"], model=model_id[3])
-
     print(
         "Success, all tasks submitted! Embeddings:",
         embeddings_1[0].shape,
@@ -94,4 +105,4 @@ def main():
         rerankings_1,
         "Classifications:",
         classifications_1,
-    )
+    )
diff --git a/infra/modal/webserver.py b/infra/modal/webserver.py
@@ -11,7 +11,7 @@
 # CONFIG.
 # Writing the configuration to a .env file to have the example in one file.
 PORT = 7997
-INFINITY_PIPY_VERSION = "0.0.45"
+INFINITY_VERSION = os.environ.get("INFINITY_VERSION", "0.0.45")
 pathlib.Path(".env").write_text(
     f"""
 # Auto-generated by webserver.py
@@ -44,9 +44,7 @@ def download_models():
 # We'll start from a recommended Docker Hub image and install `infinity`.
 image = (
     Image.from_registry("nvidia/cuda:12.1.1-base-ubuntu22.04", add_python="3.11")
-    .pip_install(
-        f"infinity_emb[all,vision]=={INFINITY_PIPY_VERSION}"
-    )
+    .pip_install(f"infinity_emb[all]=={INFINITY_VERSION}")
     .run_function(
         download_models,
         secrets=[

diff --git a/libs/infinity_emb/tests/unit_test/test_engine.py b/libs/infinity_emb/tests/unit_test/test_engine.py
@@ -25,7 +25,7 @@ async def test_async_api_debug():
         EngineArgs(engine=InferenceEngine.debugengine)
     )
     async with engine:
-        embeddings, usage = await engine.embed(sentences=sentences)
+        embeddings, usage = await engine.embed(sentences)
         embeddings = np.array(embeddings)
         assert usage == sum([len(s) for s in sentences])
         assert embeddings.shape[0] == len(sentences)