Added CLI

Author: jianshen92

README.md

@@ -16,13 +16,13 @@ This library provides you with an instant, easy-to-use interface for CLIP, allow
 Ensure that you have Python 3.8 or newer and `pip` installed on your system. We highly recommend using a Virtual Environment to avoid any potential package conflicts.
 
 To install the service, enter the following command:
-```
+```bash
 pip install clip-api-service
 ```
 
 Once the installation process is complete, you can start the service by running:
-```
-clip-api-service serve --model_name=ViT-B-32:openai
+```bash
+clip-api-service serve --model-name=ViT-B-32:openai
 ```
 Your service is now running! Interact with it via the Swagger UI at `localhost:3000`
 ![SwaggerUI](images/swagger-ui.png)
@@ -67,7 +67,7 @@ bentoml yatai login --api-token <your-api-token> --endpoint <bento-cloud-endpoin
 Next, build your BentoML service using the `build` command:
 
 ```bash
-clip-api-service build --model_name=ViT-B-32:openai
+clip-api-service build --model-name=ViT-B-32:openai
 ```
 
 Lastly, push your freshly-built Bento service to BentoCloud using the `push` command:
@@ -181,13 +181,13 @@ And the response looks like:
 Spins up a HTTP Server with the model of your choice.
 
 Arguments:
-* `--model_name` : Name of the CLIP model. Use `list_models` to see the list of available model. Default: `openai/clip-vit-large-patch14`
+* `--model-name` : Name of the CLIP model. Use `list_models` to see the list of available model. Default: `openai/clip-vit-large-patch14`
 
 #### `build`
 Builds a Bento with the model of your choice
 
 Arguments:
-* `--model_name` : Name of the CLIP model. Use `list_models` to see the list of available model. Default: `openai/clip-vit-large-patch14`
+* `--model-name` : Name of the CLIP model. Use `list_models` to see the list of available model. Default: `openai/clip-vit-large-patch14`
 
 #### `list_models`
 List all available CLIP models.

pyproject.toml

@@ -31,13 +31,23 @@ keywords = [
 ]
 dependencies = [
     "bentoml>=1.0.20",
+    "bentoml",
+    "transformers",
+    "accelerate",
+    "optimum",
+    "pydantic",
+    "Pillow",
+    "open-clip-torch",
+    "torch",
+    "typer",
 ]
 
 [project.urls]
 "Homepage" = "https://github.com/bentoml/CLIP-API-service"
 "Bug Tracker" = "https://github.com/bentoml/CLIP-API-service/issues"
 
-
+[project.scripts]
+clip-api-service = "cli:app"
 
 [tool.pdm.dev-dependencies]
 test = [

requirements-dev.txt

@@ -6,3 +6,4 @@ pydantic
 Pillow
 open_clip_torch
 torch
+typer

src/cli.py

@@ -0,0 +1,34 @@
+import os
+import typer
+import subprocess
+
+from typing_extensions import Annotated
+from clip_api_service import list_models as _list_models
+from clip_api_service.build import build_bento
+from clip_api_service.models import MODEL_ENV_VAR_KEY, DEFAULT_MODEL_NAME
+app = typer.Typer()
+
+@app.command()
+def serve(model_name: Annotated[str, typer.Option(help="CLIP Model name")] = DEFAULT_MODEL_NAME):
+    env = os.environ.copy()
+    env[MODEL_ENV_VAR_KEY] = model_name
+
+    try:
+        subprocess.run(["bentoml", "serve", "clip_api_service.service:svc"], env=env, check=True)
+    except subprocess.CalledProcessError as e:
+        typer.echo(f"Command 'bentoml serve {model_name}' failed with error code {e.returncode}")
+
+
+@app.command()
+def build(
+    model_name: Annotated[str, typer.Option(help="CLIP Model name")] = DEFAULT_MODEL_NAME,
+    use_gpu: Annotated[bool, typer.Option(help="Use GPU for build")] = False,
+):
+    build_bento(model_name=model_name, use_gpu=use_gpu)
+
+@app.command()
+def list_models():
+    print(_list_models())
+
+if __name__ == "__main__":
+    app()

src/clip_api_service/bentofiles/bentofile.openai.cpu.yaml

@@ -2,4 +2,10 @@ service: "clip_api_service.service:svc"
 include:
 - "clip_api_service/*.py"
 python:
-    requirements_txt: "./requirements.txt"
+    packages:
+    - bentoml
+    - transformers
+    - accelerate
+    - optimum
+    - pydantic
+    - Pillow

src/clip_api_service/bentofiles/bentofile.openai.gpu.yaml

@@ -2,4 +2,12 @@ service: "clip_api_service.service:svc"
 include:
 - "clip_api_service/*.py"
 python:
-    requirements_txt: "./requirements.txt"
+    packages:
+    - bentoml
+    - transformers
+    - accelerate
+    - optimum
+    - pydantic
+    - Pillow
+docker:
+    cuda_version: "11.6.2"

src/clip_api_service/models/openai.py

@@ -42,6 +42,16 @@ def download_model(model_name: str) -> bentoml.Model:
     )
     return bentoml.models.get(bento_model_tag)
 
+def bentofile_path(use_gpu: bool = False) -> str:
+    import os
+    build_ctx = os.path.dirname(os.path.dirname(__file__))
+
+    declaration = {
+        False: os.path.join(build_ctx, "bentofiles", "bentofile.openai.cpu.yaml"),
+        True: os.path.join(build_ctx, "bentofiles", "bentofile.openai.gpu.yaml"),
+    }
+
+    return declaration[use_gpu]
 
 class OpenAICLIPRunnable(CLIPRunnable):
     SUPPORTED_RESOURCES = ("nvidia.com/gpu", "cpu")

src/clip_api_service/models/openclip.py

@@ -102,6 +102,7 @@
 }
 
 
+
 def get_bento_model_tag(model_name: str) -> bentoml.Tag:
     model_version = model_name.replace(":", ".")
     return bentoml.Tag("openclip", model_version.lower())
@@ -131,6 +132,16 @@ def download_model(model_name: str) -> bentoml.Model:
     )
     return bentoml.models.get(bento_model_tag)
 
+def bentofile_path(use_gpu: bool = False) -> str:
+    import os
+    build_ctx = os.path.dirname(os.path.dirname(__file__))
+
+    declaration = {
+        False: os.path.join(build_ctx, "bentofiles", "bentofile.openclip.cpu.yaml"),
+        True: os.path.join(build_ctx, "bentofiles", "bentofile.openclip.gpu.yaml"),
+    }
+
+    return declaration[use_gpu]
 
 class OpenClipRunnable(CLIPRunnable):
     SUPPORTED_RESOURCES = ("nvidia.com/gpu", "cpu")

src/main.py

@@ -41,3 +41,9 @@ def test():
     # server = bentoml.HTTPServer(svc)
     # server.start(blocking=False)
     test()
+
+#%%
+from clip_api_service.build import build_bento
+# %%
+build_bento(model_name="ViT-B-32:openai")
+# %%