feat(ml): export clip models to ONNX and host models on Hugging Face (#4700)

* export clip models

* export to hf

refactored export code

* export mclip, general refactoring

cleanup

* updated conda deps

* do transforms with pillow and numpy, add tokenization config to export, general refactoring

* moved conda dockerfile, re-added poetry

* minor fixes

* updated link

* updated tests

* removed `requirements.txt` from workflow

* fixed mimalloc path

* removed torchvision

* cleaner np typing

* review suggestions

* update default model name

* update test

machine-learning/app/conftest.py

@@ -1,5 +1,6 @@
 import json
-from typing import Any, Iterator, TypeAlias
+from pathlib import Path
+from typing import Any, Iterator
 from unittest import mock
 
 import numpy as np
@@ -8,8 +9,7 @@ from fastapi.testclient import TestClient
 from PIL import Image
 
 from .main import app, init_state
-
-ndarray: TypeAlias = np.ndarray[int, np.dtype[np.float32]]
+from .schemas import ndarray_f32
 
 
 @pytest.fixture
@@ -18,13 +18,13 @@ def pil_image() -> Image.Image:
 
 
 @pytest.fixture
-def cv_image(pil_image: Image.Image) -> ndarray:
+def cv_image(pil_image: Image.Image) -> ndarray_f32:
     return np.asarray(pil_image)[:, :, ::-1]  # PIL uses RGB while cv2 uses BGR
 
 
 @pytest.fixture
 def mock_get_model() -> Iterator[mock.Mock]:
-    with mock.patch("app.models.cache.InferenceModel.from_model_type", autospec=True) as mocked:
+    with mock.patch("app.models.cache.from_model_type", autospec=True) as mocked:
         yield mocked
 
 
@@ -37,3 +37,25 @@ def deployed_app() -> TestClient:
 @pytest.fixture(scope="session")
 def responses() -> dict[str, Any]:
     return json.load(open("responses.json", "r"))
+
+
+@pytest.fixture(scope="session")
+def clip_model_cfg() -> dict[str, Any]:
+    return {
+        "embed_dim": 512,
+        "vision_cfg": {"image_size": 224, "layers": 12, "width": 768, "patch_size": 32},
+        "text_cfg": {"context_length": 77, "vocab_size": 49408, "width": 512, "heads": 8, "layers": 12},
+    }
+
+
+@pytest.fixture(scope="session")
+def clip_preprocess_cfg() -> dict[str, Any]:
+    return {
+        "size": [224, 224],
+        "mode": "RGB",
+        "mean": [0.48145466, 0.4578275, 0.40821073],
+        "std": [0.26862954, 0.26130258, 0.27577711],
+        "interpolation": "bicubic",
+        "resize_mode": "shortest",
+        "fill_color": 0,
+    }