feat: update onnxruntime to 1.20.1

.github/workflows/release.yml

@@ -33,11 +33,10 @@ jobs:
           file_pattern: pyproject.toml
       # Step 4: Attendere che il commit sia sincronizzato su GitHub
       - name: Wait for commit sync
-        run: sleep 15
+        run: sleep 10
       - uses: actions/checkout@v4
       # Step 5: Creare e taggare la nuova versione
       - name: Tag version
-
         id: tag_version
         uses: mathieudutour/[email protected]
         with:
@@ -49,4 +48,4 @@ jobs:
           tag: ${{ steps.tag_version.outputs.new_tag }}
           name: Release ${{ steps.tag_version.outputs.new_tag }}
           body: ${{ steps.tag_version.outputs.changelog }}
-          generateReleaseNotes: true
+          generateReleaseNotes: true

.github/workflows/test.yml

@@ -24,7 +24,7 @@ jobs:
           python-version: "3.12"
           cache: "pip"
       - name: Install dependencies
-        run: make install-dev
+        run: pip install .[dev]
       - name: Run test
         run: make test
       - name: Pytest coverage comment

.gitignore

@@ -88,4 +88,4 @@ ipython_config.py
 notebooks/.data
 .venv
 /data
-tests/junit.xml
+tests/junit.xml

.vscode/settings.json

@@ -1,4 +1,4 @@
 {
     "python.analysis.typeCheckingMode": "basic",
     "python.analysis.autoImportCompletions": true
-}
+}

Makefile

@@ -3,12 +3,10 @@
 install:
 	@pip install . --no-cache-dir
 install-dev:
-	@pip install -e ".[inference,dev]" --no-cache-dir
+	@pip install -e ".[dev]" --no-cache-dir
 
 install-pre-commit:
 	@pre-commit install
-
-
 lint:
 	@isort . --profile=black
 	@black .

README.md

@@ -17,16 +17,29 @@
 
 # Focoos SDK
 
+
+## Requirements
+For **local inference**, ensure that you have CUDA 12 and cuDNN 9 installed, as they are required for onnxruntime version 1.20.1.
+
+To install cuDNN 9:
+```bash
+apt-get -y install cudnn9-cuda-12
+```
+
+To perform inference using TensorRT, ensure you have TensorRT version 10.5 installed.
+
 # Install
 
 ```bash
 pip install .
 ```
 
-- local inference (CPU, COREML): `pip install .[inference]`
-- local inference with GPU (CUDA): `pip install .[inference-gpu]`
+- local inference (CPU, COREML): `pip install .`
+- local inference with GPU (CUDA): `pip install .[gpu]`
+
 
 ## 🤖 Cloud Inference
+
 ```python
 from focoos import Focoos
 

focoos/ports.py

@@ -234,7 +234,6 @@ class LatencyMetrics:
 
 class RuntimeTypes(str, Enum):
     ONNX_CUDA32 = "onnx_cuda32"
-    ONNX_CUDA16 = "onnx_cuda16"
     ONNX_TRT32 = "onnx_trt32"
     ONNX_TRT16 = "onnx_trt16"
     ONNX_CPU = "onnx_cpu"

focoos/runtime.py

@@ -287,10 +287,6 @@ def get_runtime(
         opts = OnnxEngineOpts(
             cuda=True, verbose=False, fp16=False, warmup_iter=warmup_iter
         )
-    elif runtime_type == RuntimeTypes.ONNX_CUDA16:
-        opts = OnnxEngineOpts(
-            cuda=True, verbose=False, fp16=True, warmup_iter=warmup_iter
-        )
     elif runtime_type == RuntimeTypes.ONNX_TRT32:
         opts = OnnxEngineOpts(
             cuda=False, verbose=False, trt=True, fp16=False, warmup_iter=warmup_iter

gradio/app.py

@@ -12,9 +12,7 @@
 focoos_models = []
 focoos = Focoos(api_key=os.getenv("FOCOOS_API_KEY"))
 focoos_models = [
-    model["ref"]
-    for model in focoos.list_focoos_models()
-    if model["status"] == "DEPLOYED"
+    model.ref for model in focoos.list_focoos_models() if model.status == "DEPLOYED"
 ]
 loaded_models = {}
 image_examples = [

pyproject.toml

@@ -20,6 +20,7 @@ dependencies = [
     "tqdm~=4.67.1",
     "numpy~=1.26.4",
     "scipy~=1.14.1",
+    "onnxruntime==1.20.1"
 ]
 
 authors = [{ name = "focoos.ai", email = "[email protected]" }]
@@ -42,6 +43,4 @@ dev = [
     "gradio~=5.3.0",
     "sniffio~=1.2.0",
 ]
-
-inference = ["onnxruntime==1.18.0"]
-inference-gpu = ["onnxruntime-gpu==1.18.0"]
+gpu = ["onnxruntime-gpu==1.20.1"]