Upload 9 files

app.py

@@ -103,18 +103,18 @@ with gr.Blocks(title="Salamandra Vision 7B · ZeroGPU") as demo:
             out = gr.Textbox(label="Descripción", lines=18)
 
     # UI
-    btn.click(_infer_one, [in_img, in_txt, max_new, temp], out, api_name="describe")
+    btn.click(_infer_one, [in_img, in_txt, max_new, temp], out, api_name="describe", concurrency_limit=1)
 
     # API simple (multipart) compatible con tu versión anterior
-    demo.load(
-        None,
-        [gr.Image(label="image", type="pil"),
-         gr.Textbox(value="Describe la imagen con detalle."),
-         gr.Slider(16, 1024, value=256),
-         gr.Slider(0.0, 1.5, value=0.7)],
-        describe_raw,
-        api_name="describe_raw"
-    )
+    # demo.load(
+    #     None,
+    #     [gr.Image(label="image", type="pil"),
+    #      gr.Textbox(value="Describe la imagen con detalle."),
+    #      gr.Slider(16, 1024, value=256),
+    #      gr.Slider(0.0, 1.5, value=0.7)],
+    #     describe_raw,
+    #     api_name="describe_raw"
+    # )
 
     # API BATCH para ENGINE (Gradio Client): images + context_json → list[str]
     # Firma que espera el VisionClient del engine (api_name="/predict")
@@ -127,7 +127,7 @@ with gr.Blocks(title="Salamandra Vision 7B · ZeroGPU") as demo:
 
     # Nota: Gradio Gallery entrega rutas/obj; nos apoyamos en el cliente para cargar archivos
     batch_btn.click(describe_batch, [batch_in_images, batch_context, batch_max, batch_temp], batch_out,
-                    api_name="predict")
+                    api_name="predict", concurrency_limit=1)
 
-demo.queue(concurrency_count=1, max_size=16).launch()
+demo.queue(max_size=16).launch()
 

requirements.txt

@@ -1,5 +1,3 @@
-gradio>=4.44.1
-spaces>=0.25.0
 transformers>=4.44.0
 torch>=2.2
 accelerate>=0.30.0