create app.py

app.py

@@ -0,0 +1,96 @@
+import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline, StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer
+import time
+import numpy as np
+from torch.nn import functional as F
+import os
+from threading import Thread
+
+# init
+tok = AutoTokenizer.from_pretrained("togethercomputer/RedPajama-INCITE-Chat-3B-v1")
+m = AutoModelForCausalLM.from_pretrained("togethercomputer/RedPajama-INCITE-Chat-3B-v1", torch_dtype=torch.float16)
+m = model.to('cuda:0')
+
+class StopOnTokens(StoppingCriteria):
+    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
+        #stop_ids = [[29, 13961, 31], [29, 12042, 31], 1, 0]
+        stop_ids = [29, 0]
+        for stop_id in stop_ids:
+            #print(f"^^input ids - {input_ids}")
+            if input_ids[0][-1] == stop_id:
+                return True
+        return False
+
+        
+def user(message, history):
+    # Append the user's message to the conversation history
+    return "", history + [[message, ""]]
+
+
+
+def chat(curr_system_message, history):
+    # Initialize a StopOnTokens object
+    stop = StopOnTokens()
+
+    # Construct the input message string for the model by concatenating the current system message and conversation history
+    messages = "".join(["".join(["\n<human>:"+item[0], "\n<bot>:"+item[1]])  #curr_system_message + 
+                for item in history])
+
+    # Tokenize the messages string
+    model_inputs = tok([messages], return_tensors="pt").to("cuda")
+    streamer = TextIteratorStreamer(
+        tok, timeout=10., skip_prompt=False, skip_special_tokens=True)
+    generate_kwargs = dict(
+        model_inputs,
+        streamer=streamer,
+        max_new_tokens=1024,
+        do_sample=True,
+        top_p=0.95,
+        top_k=1000,
+        temperature=1.0,
+        num_beams=1,
+        stopping_criteria=StoppingCriteriaList([stop])
+    )
+    t = Thread(target=m.generate, kwargs=generate_kwargs)
+    t.start()
+
+    # print(history)
+    # Initialize an empty string to store the generated text
+    partial_text = ""
+    for new_text in streamer:
+        print(new_text)
+        if new_text != '<':
+            partial_text += new_text
+            history[-1][1] = partial_text.split('<bot>:')[-1]
+            # Yield an empty string to cleanup the message textbox and the updated conversation history
+            yield history
+    return partial_text
+
+
+with gr.Blocks() as demo:
+    gr.Markdown("## RedPajama-INCITE-Chat-3B-v1")
+    gr.HTML('''<center><a href="https://huggingface.co/spaces/ysharma/RedPajama-Chat-3B?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space to skip the queue and run in a private space</center>''')
+    chatbot = gr.Chatbot().style(height=500)
+    with gr.Row():
+        with gr.Column():
+            msg = gr.Textbox(label="Chat Message Box", placeholder="Chat Message Box",
+                             show_label=False).style(container=False)
+        with gr.Column():
+            with gr.Row():
+                submit = gr.Button("Submit")
+                stop = gr.Button("Stop")
+                clear = gr.Button("Clear")
+    system_msg = gr.Textbox(
+        start_message, label="System Message", interactive=False, visible=False)
+
+    submit_event = msg.submit(fn=user, inputs=[msg, chatbot], outputs=[msg, chatbot], queue=False).then(
+        fn=chat, inputs=[system_msg, chatbot], outputs=[chatbot], queue=True)
+    submit_click_event = submit.click(fn=user, inputs=[msg, chatbot], outputs=[msg, chatbot], queue=False).then(
+        fn=chat, inputs=[system_msg, chatbot], outputs=[chatbot], queue=True)
+    stop.click(fn=None, inputs=None, outputs=None, cancels=[
+               submit_event, submit_click_event], queue=False)
+    clear.click(lambda: None, None, [chatbot], queue=False)
+
+demo.queue(max_size=32, concurrency_count=2)
+demo.launch(debug=True)