v1

app.py

@@ -32,7 +32,7 @@ freeze_model(meteor)
 # previous length
 previous_length = 0
 
-def threading_function(inputs, image_token_number, streamer, device):
+def threading_function(inputs, image_token_number, streamer, device, temperature, new_max_token, top_p):
 
     # Meteor Mamba
     mmamba_inputs = mmamba.eval_process(inputs=inputs, tokenizer=tok_meteor, device=device, img_token_number=image_token_number)
@@ -50,14 +50,14 @@ def threading_function(inputs, image_token_number, streamer, device):
     generation_kwargs = meteor_inputs
     generation_kwargs.update({'streamer': streamer})
     generation_kwargs.update({'do_sample': True})
-    generation_kwargs.update({'max_new_tokens': 128})
-    generation_kwargs.update({'top_p': 0.95})
-    generation_kwargs.update({'temperature': 0.9})
+    generation_kwargs.update({'max_new_tokens': new_max_token})
+    generation_kwargs.update({'top_p': top_p})
+    generation_kwargs.update({'temperature': temperature})
     generation_kwargs.update({'use_cache': True})
     return meteor.generate(**generation_kwargs)
 
 @spaces.GPU
-def bot_streaming(message, history):
+def bot_streaming(message, history, temperature, new_max_token, top_p):
 
     # param
     for param in mmamba.parameters():
@@ -80,7 +80,13 @@ def bot_streaming(message, history):
         streamer = TextIteratorStreamer(tok_meteor, skip_special_tokens=True)
 
         # Threading generation
-        thread = Thread(target=threading_function, kwargs=dict(inputs=inputs, image_token_number=image_token_number, streamer=streamer, device=accel.device))
+        thread = Thread(target=threading_function, kwargs=dict(inputs=inputs,
+                                                               image_token_number=image_token_number,
+                                                               streamer=streamer,
+                                                               device=accel.device,
+                                                               temperature=temperature,
+                                                               new_max_token=new_max_token,
+                                                               top_p=top_p))
         thread.start()
 
         # generated text
@@ -98,7 +104,9 @@ def bot_streaming(message, history):
         time.sleep(0.02)
         yield buffer
 
-demo = gr.ChatInterface(fn=bot_streaming, title="☄️ Meteor",
+demo = gr.ChatInterface(fn=bot_streaming,
+                        additional_inputs = [gr.Slider(0, 1, 0.9, label="temperature"), gr.Slider(1, 2048, 128, label="new_max_token"), gr.Slider(0, 1, 0.95, label="top_p")],
+                        title="☄️ Meteor",
                         description="Meteor is efficient 7B size Large Language and Vision Model built on the help of traversal of rationale",
                         stop_btn="Stop Generation", multimodal=True)
 demo.launch()