Update app.py

app.py

@@ -4,108 +4,101 @@ from threading import Thread
 import gradio as gr
 import torch
 from PIL import Image
-from transformers import AutoProcessor, LlavaForConditionalGeneration, TextIteratorStreamer, TextStreamer
+from transformers import AutoProcessor, LlavaForConditionalGeneration
+from transformers import TextIteratorStreamer
 
 import spaces
-import argparse
-
-from llava_llama3.model.builder import load_pretrained_model
-from llava_llama3.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN, DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN
-from llava_llama3.conversation import conv_templates, SeparatorStyle
-from llava_llama3.utils import disable_torch_init
-from llava_llama3.mm_utils import process_images, tokenizer_image_token, get_model_name_from_path
-from llava_llama3.serve.cli import chat_llava
-
-import requests
-from io import BytesIO
-import base64
-import os
-import glob
-import pandas as pd
-from tqdm import tqdm
-import json
-
-root_path = os.path.dirname(os.path.abspath(__file__))
-print(f'\033[92m{root_path}\033[0m')
-os.environ['GRADIO_TEMP_DIR'] = root_path
-
-parser = argparse.ArgumentParser()
-parser.add_argument("--model-path", type=str, default="TheFinAI/FinLLaVA")
-parser.add_argument("--device", type=str, default="cuda")
-parser.add_argument("--conv-mode", type=str, default="llama_3")
-parser.add_argument("--temperature", type=float, default=0.7)
-parser.add_argument("--max-new-tokens", type=int, default=512)
-parser.add_argument("--load-8bit", action="store_true")
-parser.add_argument("--load-4bit", action="store_true")
-args = parser.parse_args()
-
-# Load model
-tokenizer, llava_model, image_processor, context_len = load_pretrained_model(
-    args.model_path, 
-    None, 
-    'llava_llama3', 
-    args.load_8bit, 
-    args.load_4bit, 
-    device=args.device)
+
+
+PLACEHOLDER = """
+<div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">
+   <img src="https://cdn-uploads.huggingface.co/production/uploads/64ccdc322e592905f922a06e/DDIW0kbWmdOQWwy4XMhwX.png" style="width: 80%; max-width: 550px; height: auto; opacity: 0.55;  "> 
+   <h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">LLaVA-Llama-3-8B</h1>
+   <p style="font-size: 18px; margin-bottom: 2px; opacity: 0.65;">Llava-Llama-3-8b is a LLaVA model fine-tuned from Meta-Llama-3-8B-Instruct and CLIP-ViT-Large-patch14-336 with ShareGPT4V-PT and InternVL-SFT by XTuner</p>
+</div>
+"""
+
+
+model_id = "TheFinAI/FinLLaVA"
+
+processor = AutoProcessor.from_pretrained(model_id)
+
+model = LlavaForConditionalGeneration.from_pretrained(
+    model_id,
+    torch_dtype=torch.float16,
+    low_cpu_mem_usage=True,
+)
+
+model.to("cuda:0")
+model.generation_config.eos_token_id = 128009
+
 
 @spaces.GPU
 def bot_streaming(message, history):
     print(message)
-    image_file = None
     if message["files"]:
+        # message["files"][-1] is a Dict or just a string
         if type(message["files"][-1]) == dict:
-            image_file = message["files"][-1]["path"]
+            image = message["files"][-1]["path"]
         else:
-            image_file = message["files"][-1]
+            image = message["files"][-1]
     else:
+        # if there's no image uploaded for this turn, look for images in the past turns
+        # kept inside tuples, take the last one
         for hist in history:
             if type(hist[0]) == tuple:
-                image_file = hist[0][0]
-                
-    if image_file is None:
+                image = hist[0][0]
+    try:
+        if image is None:
+            # Handle the case where image is None
+            gr.Error("You need to upload an image for LLaVA to work.")
+    except NameError:
+        # Handle the case where 'image' is not defined at all
         gr.Error("You need to upload an image for LLaVA to work.")
-        return
-    
-    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
-    def generate():
-        print('\033[92mRunning chat\033[0m')
-        output = chat_llava(
-                    args=args,
-                    image_file=image_file,
-                    text=message['text'],
-                    tokenizer=tokenizer,
-                    model=llava_model,
-                    image_processor=image_processor,
-                    context_len=context_len,
-                    streamer=streamer)
-        return output
-
-    thread = Thread(target=generate)
+
+    prompt = f"<|start_header_id|>user<|end_header_id|>\n\n<image>\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
+    # print(f"prompt: {prompt}")
+    image = Image.open(image)
+    inputs = processor(prompt, image, return_tensors='pt').to(0, torch.float16)
+
+    streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": False, "skip_prompt": True})
+    generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024, do_sample=False)
+
+    thread = Thread(target=model.generate, kwargs=generation_kwargs)
     thread.start()
-    # thread.join()
+
+    text_prompt = f"<|start_header_id|>user<|end_header_id|>\n\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
+    # print(f"text_prompt: {text_prompt}")
 
     buffer = ""
-    # output = generate()
+    time.sleep(0.5)
     for new_text in streamer:
+        # find <|eot_id|> and remove it from the new_text
+        if "<|eot_id|>" in new_text:
+            new_text = new_text.split("<|eot_id|>")[0]
         buffer += new_text
+
+        # generated_text_without_prompt = buffer[len(text_prompt):]
         generated_text_without_prompt = buffer
+        # print(generated_text_without_prompt)
         time.sleep(0.06)
+        # print(f"new_text: {generated_text_without_prompt}")
         yield generated_text_without_prompt
 
-chatbot = gr.Chatbot(scale=1)
+
+chatbot=gr.Chatbot(placeholder=PLACEHOLDER,scale=1)
 chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)
-with gr.Blocks(fill_height=True) as demo:
+with gr.Blocks(fill_height=True, ) as demo:
     gr.ChatInterface(
-        fn=bot_streaming,
-        title="FinLLaVA Demo",
-        examples=[
-            {"text": "What is in this picture?", "files": ["http://images.cocodataset.org/val2017/000000039769.jpg"]},
-        ],
-        description="",
-        stop_btn="Stop Generation",
-        multimodal=True,
-        textbox=chat_input,
-        chatbot=chatbot,
+    fn=bot_streaming,
+    title="LLaVA Llama-3-8B",
+    examples=[{"text": "What is on the flower?", "files": ["./bee.jpg"]},
+             ],
+    description="Try [LLaVA Llama-3-8B](https://huggingface.co/xtuner/llava-llama-3-8b-v1_1-transformers). Upload an image 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,
+    textbox=chat_input,
+    chatbot=chatbot,
     )
 
 demo.queue(api_open=False)