updated parameter of generate.py in app.py

app.py

@@ -24,23 +24,22 @@ print(f"DEBUG: Using model repo ID: {MODEL_REPO_ID}")
 
 # In app.py
 
+# In app.py
+
 def call_generate_script(image_path: str, prompt_text: str) -> str:
     print(f"\n--- DEBUG (call_generate_script) ---")
     print(f"Timestamp: {time.strftime('%Y-%m-%d %H:%M:%S')}")
     print(f"Calling with image_path='{image_path}', prompt='{prompt_text}'")
     
-    # Arguments for nanoVLM's generate.py, VERIFIED against its source code
+    # Arguments for the provided nanoVLM's generate.py
     cmd_args = [
         "python", "-u", GENERATE_SCRIPT_PATH,
         "--hf_model", MODEL_REPO_ID,
-        "--image_path", image_path,      # VERIFIED: script expects --image_path
+        "--image", image_path,          # VERIFIED: script uses --image
         "--prompt", prompt_text,
-        "--num_samples", "1",           # VERIFIED: script expects --num_samples
-        "--max_new_tokens", "30",       # This was correct
-        "--device", "cpu"               # VERIFIED: script expects --device
-        # Optional args for generate.py that you can add if needed:
-        # "--temperature", "0.7",
-        # "--top_k", "200" # Default is 200 in script
+        "--generations", "1",           # VERIFIED: script uses --generations
+        "--max_new_tokens", "30"        # This was correct
+        # No --device argument, as it's not in the provided generate.py
     ]
     
     print(f"Executing command: {' '.join(cmd_args)}")
@@ -71,33 +70,37 @@ def call_generate_script(image_path: str, prompt_text: str) -> str:
 
         if process.returncode != 0:
             error_message = f"Error: Generation script failed (code {process.returncode})."
-            if "unrecognized arguments" in stderr:
+            if "unrecognized arguments" in stderr: # This shouldn't happen now
                 error_message += " Argument mismatch with script."
+            elif "syntax error" in stderr.lower():
+                error_message += " Syntax error in script."
             print(error_message)
-            return error_message + f" STDERR Snippet: {stderr[:300]}" # Show more stderr
+            return error_message + f" STDERR Snippet: {stderr[:300]}"
 
-        # --- Parse the output from nanoVLM's generate.py ---
-        # The original nanoVLM generate.py prints:
-        # > Sample 1: <generated text>
+        # --- Parse the output from the provided nanoVLM's generate.py ---
+        # The script prints:
+        #   >> Generation {i+1}: {out}
         output_lines = stdout.splitlines()
         generated_text = "[No parsable output from generate.py]"
         
         found_output_line = False
         for line_idx, line in enumerate(output_lines):
             stripped_line = line.strip()
-            # The actual generate.py from nanoVLM prints "> Sample 1:"
-            prefix_to_remove = None
-            if stripped_line.startswith("> Sample 1:"): 
-                prefix_to_remove = "> Sample 1:"
-            
-            if prefix_to_remove:
-                generated_text = stripped_line.replace(prefix_to_remove, "", 1).strip()
-                found_output_line = True
-                print(f"Parsed generated text: '{generated_text}'")
-                break 
-        
+            # Looking for the specific output format "  >> Generation X: text"
+            if stripped_line.startswith(">> Generation 1:"): # Assuming we only care about the first generation
+                # Extract text after ">> Generation 1: " (note the space after colon)
+                try:
+                    generated_text = stripped_line.split(">> Generation 1:", 1)[1].strip()
+                    found_output_line = True
+                    print(f"Parsed generated text: '{generated_text}'")
+                    break 
+                except IndexError:
+                    print(f"Could not split line for '>> Generation 1:': '{stripped_line}'")
+                    generated_text = f"[Parsing failed] Malformed 'Generation 1' line: {stripped_line}"
+                    break
+
         if not found_output_line:
-            print(f"Could not find '> Sample 1:' line in generate.py output. Raw STDOUT was:\n{stdout}")
+            print(f"Could not find '>> Generation 1:' line in generate.py output. Raw STDOUT was:\n{stdout}")
             if stdout:
                 generated_text = f"[Parsing failed] STDOUT: {stdout[:500]}"
             else: