update 4B

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,13 +1,14 @@
 ---
 title: Ovis2 4B
-emoji: 🐢
-colorFrom: purple
-colorTo: yellow
+emoji: 🦫
+colorFrom: yellow
+colorTo: purple
 sdk: gradio
-sdk_version: 5.16.0
+sdk_version: 5.1.0
 app_file: app.py
 pinned: false
+license: apache-2.0
 short_description: Ovis2-4B
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,188 @@
+import subprocess
+subprocess.run('pip install flash-attn==2.7.0.post2 --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
+
+import spaces
+import os
+import re
+import logging
+from typing import List, Any
+from threading import Thread
+
+import torch
+import gradio as gr
+from transformers import AutoModelForCausalLM, TextIteratorStreamer
+
+model_name = 'AIDC-AI/Ovis2-4B'
+use_thread = True
+
+# load model
+model = AutoModelForCausalLM.from_pretrained(model_name,
+                                             torch_dtype=torch.bfloat16,
+                                             multimodal_max_length=8192,
+                                             trust_remote_code=True).to(device='cuda')
+text_tokenizer = model.get_text_tokenizer()
+visual_tokenizer = model.get_visual_tokenizer()
+streamer = TextIteratorStreamer(text_tokenizer, skip_prompt=True, skip_special_tokens=True)
+image_placeholder = '<image>'
+cur_dir = os.path.dirname(os.path.abspath(__file__))
+
+logging.getLogger("httpx").setLevel(logging.WARNING)
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+def initialize_gen_kwargs():
+    return {
+        "max_new_tokens": 1536,
+        "do_sample": False,
+        "top_p": None,
+        "top_k": None,
+        "temperature": None,
+        "repetition_penalty": 1.05,
+        "eos_token_id": model.generation_config.eos_token_id,
+        "pad_token_id": text_tokenizer.pad_token_id,
+        "use_cache": True
+    }
+
+def submit_chat(chatbot, text_input):
+    response = ''
+    chatbot.append((text_input, response))
+    return chatbot ,''
+
+@spaces.GPU
+def ovis_chat(chatbot: List[List[str]], image_input: Any):
+    conversations, model_inputs = prepare_inputs(chatbot, image_input)
+    gen_kwargs = initialize_gen_kwargs()
+
+    with torch.inference_mode():
+        generate_func = lambda: model.generate(**model_inputs, **gen_kwargs, streamer=streamer)
+        
+        if use_thread:
+            thread = Thread(target=generate_func)
+            thread.start()
+        else:
+            generate_func()
+
+        response = ""
+        for new_text in streamer:
+            response += new_text
+            chatbot[-1][1] = response
+            yield chatbot
+
+        if use_thread:
+            thread.join()
+
+    log_conversation(chatbot)
+
+    
+def prepare_inputs(chatbot: List[List[str]], image_input: Any):
+    # conversations = [{
+    #     "from": "system",
+    #     "value": "You are a helpful assistant, and your task is to provide reliable and structured responses to users."
+    # }]
+    conversations= []
+
+    for query, response in chatbot[:-1]:
+        conversations.extend([
+            {"from": "human", "value": query},
+            {"from": "gpt", "value": response}
+        ])
+    
+    last_query = chatbot[-1][0].replace(image_placeholder, '')
+    conversations.append({"from": "human", "value": last_query})
+
+    if image_input is not None:
+        for conv in conversations:
+            if conv["from"] == "human":
+                conv["value"] = f'{image_placeholder}\n{conv["value"]}'
+                break
+
+    logger.info(conversations)
+    
+    prompt, input_ids, pixel_values = model.preprocess_inputs(conversations, [image_input], max_partition=16)
+    attention_mask = torch.ne(input_ids, text_tokenizer.pad_token_id)
+    
+    model_inputs = {
+        "inputs": input_ids.unsqueeze(0).to(device=model.device),
+        "attention_mask": attention_mask.unsqueeze(0).to(device=model.device),
+        "pixel_values": [pixel_values.to(dtype=visual_tokenizer.dtype, device=visual_tokenizer.device)] if image_input is not None else [None]
+    }
+    
+    return conversations, model_inputs
+
+def log_conversation(chatbot):
+    logger.info("[OVIS_CONV_START]")
+    [print(f'Q{i}:\n {request}\nA{i}:\n {answer}') for i, (request, answer) in enumerate(chatbot, 1)]
+    logger.info("[OVIS_CONV_END]")
+
+def clear_chat():
+    return [], None, ""
+
+with open(f"{cur_dir}/resource/logo.svg", "r", encoding="utf-8") as svg_file:
+    svg_content = svg_file.read()
+font_size = "2.5em"
+svg_content = re.sub(r'(<svg[^>]*)(>)', rf'\1 height="{font_size}" style="vertical-align: middle; display: inline-block;"\2', svg_content)
+html = f"""
+<p align="center" style="font-size: {font_size}; line-height: 1;">
+    <span style="display: inline-block; vertical-align: middle;">{svg_content}</span>
+    <span style="display: inline-block; vertical-align: middle;">{model_name.split('/')[-1]}</span>
+</p>
+<center><font size=3><b>Ovis</b> has been open-sourced on <a href='https://huggingface.co/{model_name}'>😊 Huggingface</a> and <a href='https://github.com/AIDC-AI/Ovis'>🌟 GitHub</a>. If you find Ovis useful, a like❤️ or a star🌟 would be appreciated.</font></center>
+"""
+
+latex_delimiters_set = [{
+        "left": "\\(",
+        "right": "\\)",
+        "display": False 
+    }, {
+        "left": "\\begin{equation}",
+        "right": "\\end{equation}",
+        "display": True 
+    }, {
+        "left": "\\begin{align}",
+        "right": "\\end{align}",
+        "display": True
+    }, {
+        "left": "\\begin{alignat}",
+        "right": "\\end{alignat}",
+        "display": True
+    }, {
+        "left": "\\begin{gather}",
+        "right": "\\end{gather}",
+        "display": True
+    }, {
+        "left": "\\begin{CD}",
+        "right": "\\end{CD}",
+        "display": True
+    }, {
+        "left": "\\[",
+        "right": "\\]",
+        "display": True
+    }]
+
+text_input = gr.Textbox(label="prompt", placeholder="Enter your text here...", lines=1, container=False)
+with gr.Blocks(title=model_name.split('/')[-1], theme=gr.themes.Ocean()) as demo:
+    gr.HTML(html)
+    with gr.Row():
+        with gr.Column(scale=3):
+            image_input = gr.Image(label="image", height=350, type="pil")
+            gr.Examples(
+                examples=[
+                    [f"{cur_dir}/examples/ovis2_math2.png", "Find the area of the shaded region."],
+                    [f"{cur_dir}/examples/ovis2_figure2.png", "What is net profit margin as a percentage of total revenue?"],
+                    [f"{cur_dir}/examples/ovis2_table0.png", "Convert the table to markdown."],
+                    [f"{cur_dir}/examples/ovis2_ocr0.jpeg", "OCR:"],
+                ],
+                inputs=[image_input, text_input]
+            )
+        with gr.Column(scale=7):
+            chatbot = gr.Chatbot(label="Ovis", layout="panel", height=600, show_copy_button=True, latex_delimiters=latex_delimiters_set)
+            text_input.render()
+            with gr.Row():
+                send_btn = gr.Button("Send", variant="primary")
+                clear_btn = gr.Button("Clear", variant="secondary")
+
+    send_click_event = send_btn.click(submit_chat, [chatbot, text_input], [chatbot, text_input]).then(ovis_chat,[chatbot, image_input],chatbot)
+    submit_event = text_input.submit(submit_chat, [chatbot, text_input], [chatbot, text_input]).then(ovis_chat,[chatbot, image_input],chatbot)
+    clear_btn.click(clear_chat, outputs=[chatbot, image_input, text_input])
+
+demo.launch()

requirements.txt

@@ -0,0 +1,4 @@
+numpy==1.25.0
+torch==2.4.0
+transformers==4.46.2
+pillow==10.3.0