Upload server.py with huggingface_hub

server.py

@@ -58,6 +58,21 @@ def health():
         return jsonify({"status": "healthy"}), 200
     return jsonify({"status": "loading"}), 503
 
+@app.route("/debug", methods=["GET"])
+def debug():
+    if pipe is None:
+        return jsonify({"status": "loading"}), 503
+    import inspect
+    sig = inspect.signature(pipe.__call__)
+    params = {}
+    for name, param in sig.parameters.items():
+        params[name] = str(param.default) if param.default != inspect.Parameter.empty else "required"
+    return jsonify({
+        "pipeline_class": type(pipe).__name__,
+        "parameters": params,
+        "device": str(pipe.device) if hasattr(pipe, 'device') else 'unknown',
+    })
+
 @app.route("/", methods=["POST"])
 def generate():
     try:
@@ -83,7 +98,8 @@ def generate():
         print(f"[video-long] Generating {num_frames} frames ({num_frames/fps:.1f}s): {prompt[:100]}", flush=True)
         print(f"[video-long] Params: {width}x{height}, steps={steps}, ar_step={ar_step}", flush=True)
 
-        output = pipe(
+        # Build kwargs - only pass causal_block_size when ar_step > 0
+        call_kwargs = dict(
             prompt=prompt,
             num_inference_steps=steps,
             height=height,
@@ -91,11 +107,14 @@ def generate():
             num_frames=num_frames,
             base_num_frames=base_num_frames,
             ar_step=ar_step,
-            causal_block_size=causal_block_size,
             overlap_history=overlap_history,
             addnoise_condition=addnoise_condition,
-            guidance_scale=guidance_scale,
-        ).frames[0]
+        )
+        if ar_step > 0:
+            call_kwargs["causal_block_size"] = causal_block_size
+
+        print(f"[video-long] Pipeline call kwargs: {call_kwargs}", flush=True)
+        output = pipe(**call_kwargs).frames[0]
 
         # Write to temp file
         with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmp: