Update app.py

app.py

@@ -16,13 +16,14 @@ model = ChameleonForConditionalGeneration.from_pretrained("facebook/chameleon-7b
 def bot_streaming(message, history):
 
   txt = message.text
+  ext_buffer = f"{txt}"
 
   if message.files:
     if len(message.files) == 1:
-      img = [message.files[0].path]
-    # interleaved images
+      image = [message.files[0].path]
+    # interleaved images or video
     elif len(message.files) > 1:
-      img = [msg.path for msg in message.files]
+      image = [msg.path for msg in message.files]
   else:
       
     def has_file_data(lst):
@@ -37,31 +38,32 @@ def bot_streaming(message, history):
         if all(isinstance(sub_item, str) for sub_item in item):
             latest_text_only_index = i
 
-    img = [path for i, item in enumerate(history) if i < latest_text_only_index and has_file_data(item) for path in extract_paths(item)]
+    image = [path for i, item in enumerate(history) if i < latest_text_only_index and has_file_data(item) for path in extract_paths(item)]
 
   if message.files is None:
-      gr.Error("You need to upload an image or multiple images at least once for LLaVA to work.")
+      gr.Error("You need to upload an image or video for LLaVA to work.")
       
   image_extensions = Image.registered_extensions()
   image_extensions = tuple([ex for ex, f in image_extensions.items()])
-  if len(img) == 1:
-      image = Image.open(img[0]).convert("RGB")
-      prompt = f"{txt}<image>"
+  if len(image) == 1:
+      image = Image.open(image[0]).convert("RGB")
+      prompt = f"{message.text}<image>"
 
-  elif len(img) > 1:
+  elif len(image) > 1:
     image_list = []
-    user_prompt = txt
+    user_prompt = message.text
 
-    for im in img:
-      image_list.append(Image.open(im).convert("RGB"))
+    for img in image:
+      img = Image.open(img).convert("RGB")
+      image_list.append(img)
       
     toks = "<image>" * len(image_list)
     prompt = user_prompt + toks
 
-    img = image_list
+    image = image_list
 
 
-  inputs = processor(prompt, img, return_tensors="pt").to("cuda", torch.float16)
+  inputs = processor(prompt, image, return_tensors="pt").to("cuda", torch.float16)
   streamer = TextIteratorStreamer(processor, {"skip_special_tokens": True})
   generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=250)
   generated_text = ""
@@ -76,7 +78,7 @@ def bot_streaming(message, history):
     
     buffer += new_text
     
-    generated_text_without_prompt = buffer
+    generated_text_without_prompt = buffer#[len(ext_buffer):]
     time.sleep(0.01)
     yield buffer
 
@@ -89,4 +91,4 @@ demo = gr.ChatInterface(fn=bot_streaming, title="Chameleon 🦎", examples=[
       textbox=gr.MultimodalTextbox(file_count="multiple"), 
       description="Try [Chameleon-7B](https://huggingface.co/facebook/chameleon-7b) by Meta with transformers in this demo. Upload image(s), and start chatting about it, or simply try one of the examples below. If you don't upload an image, you will receive an error. ",
       stop_btn="Stop Generation", multimodal=True)
-demo.launch(debug=True)
+demo.launch(debug=True)