Use threads for concurrency

app.py

@@ -397,7 +397,11 @@ with gr.Blocks() as demo:
 
     get_input_img_btn = gr.Button(value="Get Input Image")
     get_input_img_btn.click(
-        get_img, inputs=[img_number], outputs=[get_img_out], api_name="get-img"
+        get_img,
+        inputs=[img_number],
+        outputs=[get_img_out],
+        api_name="get-img",
+        concurrency_limit=10,
     )
 
     ###################
@@ -457,4 +461,5 @@ with gr.Blocks() as demo:
     )
 
 if __name__ == "__main__":
+    demo.queue(default_concurrency_limit=2, max_size=32)
     demo.launch()

pyproject.toml

@@ -9,6 +9,7 @@ dependencies = [
     "einops>=0.8.0",
     "gradio>=5.3.0",
     "numpy>=2.2.2",
+    "requests>=2.32.3",
     "saev",
     "torch>=2.6.0",
     "torchvision>=0.21.0",

requirements.txt

@@ -4,7 +4,7 @@ aiofiles==23.2.1
     # via gradio
 aiohappyeyeballs==2.4.6
     # via aiohttp
-aiohttp==3.11.12
+aiohttp==3.11.13
     # via
     #   datasets
     #   fsspec
@@ -25,7 +25,7 @@ attrs==25.1.0
     #   aiohttp
     #   jsonschema
     #   referencing
-beartype==0.19.0
+beartype==0.20.0
     # via
     #   saev-semantic-segmentation (pyproject.toml)
     #   saev
@@ -84,9 +84,9 @@ fsspec==2024.12.0
     #   torch
 ftfy==6.3.1
     # via open-clip-torch
-gradio==5.16.2
+gradio==5.18.0
     # via saev-semantic-segmentation (pyproject.toml)
-gradio-client==1.7.1
+gradio-client==1.7.2
     # via gradio
 h11==0.14.0
     # via
@@ -129,7 +129,7 @@ jsonschema-specifications==2024.10.1
     # via jsonschema
 kiwisolver==1.4.8
     # via matplotlib
-marimo==0.11.7
+marimo==0.11.9
     # via saev
 markdown==3.7
     # via
@@ -153,7 +153,7 @@ multidict==6.1.0
     #   yarl
 multiprocess==0.70.16
     # via datasets
-narwhals==1.27.1
+narwhals==1.28.0
     # via
     #   altair
     #   marimo
@@ -202,7 +202,7 @@ nvidia-nvjitlink-cu12==12.4.127
     #   torch
 nvidia-nvtx-cu12==12.4.127
     # via torch
-open-clip-torch==2.30.0
+open-clip-torch==2.31.0
     # via saev
 orjson==3.10.15
     # via gradio
@@ -227,7 +227,7 @@ pillow==11.1.0
     #   matplotlib
     #   saev
     #   torchvision
-polars==1.22.0
+polars==1.23.0
     # via saev
 propcache==0.3.0
     # via
@@ -279,13 +279,14 @@ regex==2024.11.6
     # via open-clip-torch
 requests==2.32.3
     # via
+    #   saev-semantic-segmentation (pyproject.toml)
     #   datasets
     #   huggingface-hub
 rich==13.9.4
     # via
     #   typer
     #   tyro
-rpds-py==0.23.0
+rpds-py==0.23.1
     # via
     #   jsonschema
     #   referencing
@@ -293,7 +294,7 @@ ruff==0.9.7
     # via
     #   gradio
     #   marimo
-saev @ git+https://github.com/samuelstevens/saev@298cabdb6b771c76b402d0fdddab6907d1941d7a
+saev @ git+https://github.com/samuelstevens/saev@d03db3bf6a2e104ef4758c439194e500016e0366
     # via saev-semantic-segmentation (pyproject.toml)
 safehttpx==0.1.6
     # via gradio
@@ -303,7 +304,7 @@ safetensors==0.5.2
     #   timm
 semantic-version==2.10.0
     # via gradio
-setuptools==75.8.0
+setuptools==75.8.1
     # via torch
 shellingham==1.5.4
     # via typer
@@ -322,7 +323,7 @@ submitit==1.5.2
     # via saev
 sympy==1.13.1
     # via torch
-timm==1.0.14
+timm==1.0.15
     # via open-clip-torch
 tomlkit==0.13.2
     # via
@@ -382,7 +383,7 @@ wadler-lindig==0.1.3
     # via jaxtyping
 wcwidth==0.2.13
     # via ftfy
-websockets==14.2
+websockets==15.0
     # via
     #   gradio-client
     #   marimo