Add ONNX model inference implementations for MoGe-2, BEN2, and DA3

- Introduced `onnx_moge2.py` for MoGe-2 ONNX inference with NumPy post-processing.
- Added `onnx_ben2.py` for BEN2 ONNX inference with Pillow and NumPy preprocessing.
- Created `onnx_da3.py` for Depth Anything 3 inference with ONNX Runtime and NumPy.
- Implemented shared ONNX Runtime session helpers in `onnx_runtime.py`.
- Developed unit tests for server functionalities in `test_server.py`, including model caching and job management.
- Enhanced job server to handle frame extraction without requiring models.

tests/test_download_progress.py

@@ -1,7 +1,5 @@
-import importlib.util
-import json
 import os
-import tempfile
+import sys
 import unittest
 from pathlib import Path
 from unittest.mock import patch
@@ -10,14 +8,16 @@ from unittest.mock import patch
 PROJECT_ROOT = Path(__file__).resolve().parents[1]
 
 
-def load_worker():
-    specification = importlib.util.spec_from_file_location(
-        "snapform_download_worker",
-        PROJECT_ROOT / "src" / "snapform" / "ai_service" / "jobs.py",
-    )
-    module = importlib.util.module_from_spec(specification)
-    specification.loader.exec_module(module)
-    return module
+def load_download_module():
+    addon_directory = PROJECT_ROOT / "src" / "snapform"
+    sys.path.insert(0, str(addon_directory))
+    try:
+        return __import__(
+            "server.model_download",
+            fromlist=["model_download"],
+        )
+    finally:
+        sys.path.remove(str(addon_directory))
 
 
 class FakeTqdm:
@@ -36,52 +36,41 @@ class FakeTqdm:
 class DownloadProgressTest(unittest.TestCase):
     @classmethod
     def setUpClass(cls):
-        cls.worker = load_worker()
+        cls.download = load_download_module()
+
+    def create_progress(self, description="Downloading bytes", **options):
+        reports = []
+        progress_class = self.download.create_download_progress_class(
+            lambda value, message: reports.append((value, message)),
+            "owner/model",
+            tqdm_type=options.pop("tqdm_type", FakeTqdm),
+            clock=options.pop("clock", lambda: 1.0),
+            **options,
+        )
+        progress = progress_class(
+            desc=description,
+            total=100,
+            initial=0,
+            unit="B",
+        )
+        return progress, reports
 
     def test_reports_aggregated_download_bytes(self):
-        with tempfile.TemporaryDirectory() as directory:
-            status_path = Path(directory) / "status.json"
-            progress_class = self.worker.create_download_progress_class(
-                status_path,
-                "owner/model",
-                tqdm_type=FakeTqdm,
-                clock=lambda: 1.0,
-            )
-            progress = progress_class(
-                desc="Downloading bytes",
-                total=100,
-                initial=0,
-                unit="B",
-            )
+        progress, reports = self.create_progress()
 
-            progress.update(25)
+        progress.update(25)
 
-            status = json.loads(status_path.read_text(encoding="utf-8"))
-            self.assertAlmostEqual(status["progress"], 0.275)
-            self.assertEqual(status["message"], "Downloading owner/model")
+        self.assertAlmostEqual(reports[-1][0], 0.275)
+        self.assertEqual(reports[-1][1], "Downloading owner/model")
 
-    def test_reports_byte_progress_for_aggregated_bar_description(self):
-        # huggingface_hub 1.x 用 "Downloading (incomplete total...)" 作为字节进度条描述
-        with tempfile.TemporaryDirectory() as directory:
-            status_path = Path(directory) / "status.json"
-            progress_class = self.worker.create_download_progress_class(
-                status_path,
-                "owner/model",
-                tqdm_type=FakeTqdm,
-                clock=lambda: 1.0,
-            )
-            progress = progress_class(
-                desc="Downloading (incomplete total...)",
-                total=100,
-                initial=0,
-                unit="B",
-            )
+    def test_reports_hugging_face_aggregated_bar(self):
+        progress, reports = self.create_progress(
+            "Downloading (incomplete total...)"
+        )
 
-            progress.update(25)
+        progress.update(25)
 
-            status = json.loads(status_path.read_text(encoding="utf-8"))
-            self.assertAlmostEqual(status["progress"], 0.275)
-            self.assertEqual(status["message"], "Downloading owner/model")
+        self.assertAlmostEqual(reports[-1][0], 0.275)
 
     def test_reports_download_speed(self):
         class FakeTqdmWithRate(FakeTqdm):
@@ -89,129 +78,48 @@ class DownloadProgressTest(unittest.TestCase):
             def format_dict(self):
                 return {"rate": 14_000_000}
 
-        with tempfile.TemporaryDirectory() as directory:
-            status_path = Path(directory) / "status.json"
-            progress_class = self.worker.create_download_progress_class(
-                status_path,
-                "owner/model",
-                tqdm_type=FakeTqdmWithRate,
-                clock=lambda: 1.0,
-                label="da3 small",
-            )
-            progress = progress_class(
-                desc="Downloading bytes",
-                total=100,
-                initial=0,
-                unit="B",
-            )
+        progress, reports = self.create_progress(
+            tqdm_type=FakeTqdmWithRate,
+            label="da3 small",
+        )
 
-            progress.update(50)
+        progress.update(50)
 
-            status = json.loads(status_path.read_text(encoding="utf-8"))
-            self.assertEqual(
-                status["message"],
-                "Downloading da3 small 13.4 MB/s",
-            )
+        self.assertEqual(
+            reports[-1][1],
+            "Downloading da3 small 13.4 MB/s",
+        )
 
-    def test_does_not_report_estimated_time_remaining(self):
+    def test_progress_callback_has_no_eta_or_state_protocol(self):
         times = iter((0.0, 10.0))
-        with tempfile.TemporaryDirectory() as directory:
-            status_path = Path(directory) / "status.json"
-            progress_class = self.worker.create_download_progress_class(
-                status_path,
-                "owner/model",
-                tqdm_type=FakeTqdm,
-                clock=lambda: next(times),
-            )
-            progress = progress_class(
-                desc="Downloading bytes",
-                total=100,
-                initial=0,
-                unit="B",
-            )
+        progress, reports = self.create_progress(clock=lambda: next(times))
 
-            progress.update(50)
+        progress.update(50)
 
-            status = json.loads(status_path.read_text(encoding="utf-8"))
-            self.assertNotIn("eta_seconds", status)
-
-    def test_staged_download_keeps_model_label(self):
-        class FakeTqdmWithRate(FakeTqdm):
-            @property
-            def format_dict(self):
-                return {"rate": 14_000_000}
-
-        with tempfile.TemporaryDirectory() as directory:
-            status_path = Path(directory) / "status.json"
-            progress_class = self.worker.create_download_progress_class(
-                status_path,
-                "owner/model",
-                tqdm_type=FakeTqdmWithRate,
-                clock=lambda: 1.0,
-                label="da3 small",
-                stage=6,
-                stages=8,
-                stage_label="da3 small",
-            )
-            progress = progress_class(
-                desc="Downloading bytes",
-                total=100,
-                initial=0,
-                unit="B",
-            )
-
-            progress.update(50)
-
-            status = json.loads(status_path.read_text(encoding="utf-8"))
-            self.assertEqual(
-                status["message"],
-                "Downloading da3 small 13.4 MB/s",
-            )
-            self.assertEqual(status["stage"], 6)
-            self.assertEqual(status["stages"], 8)
-            self.assertEqual(status["stage_label"], "da3 small")
+        self.assertEqual(len(reports[-1]), 2)
 
     def test_ignores_unrelated_progress_bars(self):
-        with tempfile.TemporaryDirectory() as directory:
-            status_path = Path(directory) / "status.json"
-            progress_class = self.worker.create_download_progress_class(
-                status_path,
-                "owner/model",
-                tqdm_type=FakeTqdm,
-                clock=lambda: 1.0,
-            )
-            progress = progress_class(
-                desc="Reconstructing",
-                total=100,
-                initial=0,
-                unit="B",
-            )
+        progress, reports = self.create_progress("Reconstructing")
 
-            progress.update(25)
+        progress.update(25)
 
-            self.assertFalse(status_path.exists())
+        self.assertEqual(reports, [])
 
     def test_download_endpoints_uses_direct_endpoint(self):
         with patch.dict(os.environ, {"HF_ENDPOINT": ""}):
-            endpoints = self.worker.download_endpoints()
+            endpoints = self.download.download_endpoints()
 
-        self.assertEqual(
-            endpoints,
-            ["https://huggingface.co"],
-        )
+        self.assertEqual(endpoints, ["https://huggingface.co"])
 
     def test_download_endpoints_respects_environment_override(self):
         with patch.dict(os.environ, {"HF_ENDPOINT": "https://hf.example"}):
-            endpoints = self.worker.download_endpoints()
+            endpoints = self.download.download_endpoints()
 
-        self.assertEqual(
-            endpoints,
-            ["https://hf.example"],
-        )
+        self.assertEqual(endpoints, ["https://hf.example"])
 
     def test_download_endpoints_strips_trailing_slash(self):
         with patch.dict(os.environ, {"HF_ENDPOINT": "https://hf.example/"}):
-            endpoints = self.worker.download_endpoints()
+            endpoints = self.download.download_endpoints()
 
         self.assertEqual(endpoints, ["https://hf.example"])