first version, only llms are functional

app.py

@@ -1,64 +1,93 @@
+#!/usr/bin/env python3
+
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoTokenizer, AutoModelForCausalLM
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+llms = {
+    "Qwen2-1.5B":     {"model": "Qwen/Qwen2-1.5B-Instruct", "prefix": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
+    "Qwen2-3B":       {"model": "Qwen/Qwen2-3B-Instruct",   "prefix": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
+    "Qwen2-7B":       {"model": "Qwen/Qwen2-7B-Instruct",   "prefix": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
+    "Qwen2.5-1.5B":   {"model": "Qwen/Qwen2.5-1.5B-Instruct", "prefix": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
+    "Qwen2.5-3B":     {"model": "Qwen/Qwen2.5-3B-Instruct",   "prefix": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
+    "DeepSeek-Coder": {"model": "DeepSeek/DeepSeek-Coder", "prefix": "You are a helpful assistant."},
+}
 
+vlms = dict()
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+def run_example(text_input, model_id="Qwen2-1.5B"):
+    global messages
+    tokenizer = AutoTokenizer.from_pretrained(llms[model_id]["model"], trust_remote_code=True)
+    model = AutoModelForCausalLM.from_pretrained(llms[model_id]["model"], trust_remote_code=True)
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+    system_prompt = llms[model_id]["prefix"]
+    
+    if messages is None:
+        messages = [
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": text_input},
+        ]
+    else:
+        messages.append({"role": "user", "content": text_input})
+    
 
-    messages.append({"role": "user", "content": message})
+    text = tokenizer.apply_chat_template(
+        messages,
+        tokenize=False,
+        add_generation_prompt=True,
+    )
 
-    response = ""
+    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
+
+    generated_ids = model.generate(
+        **model_inputs,
+        max_new_tokens=512,
+    )
+
+    generated_ids = [
+        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
+    ]
+
+    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+
+    return response
+
+
+messages = list()
+def reset_conversation():
+    global messages
+    messages = list()
+
+with gr.Blocks() as demo:
+    gr.Markdown(
+    """
+    # LLM & VLM Demo
+    Use the different LLMs or VLMs to experience the different models.
+    """)
+    with gr.Tab(label="LLM"):
+        with gr.Row():
+            with gr.Column():
+                model_selector = gr.Dropdown(choices=list(llms.keys()), label="Model", value="Qwen2-1.5B")
+                text_input = gr.Textbox(label="User Prompt")
+                submit_btn = gr.Button(value="Submit")
+                reset_btn = gr.Button(value="Reset conversation")
+            with gr.Column():
+                model_output_text = gr.Textbox(label="Model Output Text")
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
 
+        submit_btn.click(run_example, 
+                         [text_input, model_selector], 
+                         [model_output_text])
+        
+        reset_btn.click(reset_conversation)
+                        
+    with gr.Tab(label="VLM (WIP)"):
+        with gr.Row():
+            with gr.Column():
+               input_img = gr.Image(label="Input Image", type="pil")
+               model_selector = gr.Dropdown(choices=list(vlms.keys()), label="Model", value="Qwen2-1.5B")
+               text_input = gr.Textbox(label="User Prompt")
+               submit_btn = gr.Button(value="Submit")
+               reset_btn = gr.Button(value="Reset conversation")
 
 if __name__ == "__main__":
     demo.launch()

requirements.txt

@@ -1 +1,4 @@
-huggingface_hub==0.25.2
+huggingface_hub==0.25.2
+torch
+transformers
+gradio