Spaces:
Runtime error
Runtime error
File size: 4,678 Bytes
dd2a3b9 80dce9b dd2a3b9 80dce9b dd2a3b9 80dce9b dd2a3b9 80dce9b dd2a3b9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 |
import subprocess
# Installing flash_attn
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
import gradio as gr
from PIL import Image
from transformers import AutoModelForCausalLM
from transformers import AutoProcessor
from transformers import TextIteratorStreamer
import time
from threading import Thread
import torch
import spaces
model_id = "microsoft/Phi-3-vision-128k-instruct"
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cuda", trust_remote_code=True, torch_dtype="auto")
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
model.to("cuda:0")
# Enhanced Placeholder HTML with instructions
PLACEHOLDER = """
<div style="padding: 20px; text-align: center; display: flex; flex-direction: column; align-items: center;">
<h1 style="font-size: 32px; margin-bottom: 10px;">Get Ripped with Arnold's AI Coach</h1>
<p style="font-size: 18px; margin-bottom: 5px;">Welcome to the ultimate fitness companion! 💪</p>
<ul style="text-align: left; font-size: 16px;">
<li>📸 <strong>Upload</strong> a photo of your exercise.</li>
<li>⚡ <strong>Get instant feedback</strong> to perfect your form.</li>
<li>🔥 <strong>Improve your workouts</strong> with expert tips!</li>
</ul>
</div>
"""
@spaces.GPU
def bot_streaming(message, history):
print(f'message is - {message}')
print(f'history is - {history}')
if message["files"]:
if type(message["files"][-1]) == dict:
image = message["files"][-1]["path"]
else:
image = message["files"][-1]
else:
for hist in history:
if type(hist[0]) == tuple:
image = hist[0][0]
try:
if image is None:
raise gr.Error("You need to upload an image for Phi3-Vision to work. Close the error and try again with an Image.")
except NameError:
raise gr.Error("You need to upload an image for Phi3-Vision to work. Close the error and try again with an Image.")
conversation = []
flag = False
for user, assistant in history:
if assistant is None:
flag = True
conversation.extend([{"role": "user", "content": ""}])
continue
if flag == True:
conversation[0]['content'] = f"<|image_1|>\n{user}"
conversation.extend([{"role": "assistant", "content": assistant}])
flag = False
continue
conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
if len(history) == 0:
conversation.append({"role": "user", "content": f"<|image_1|>\n{message['text']}"})
else:
conversation.append({"role": "user", "content": message['text']})
print(f"prompt is -\n{conversation}")
prompt = processor.tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True)
image = Image.open(image)
inputs = processor(prompt, image, return_tensors="pt").to("cuda:0")
streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": True, "skip_prompt": True, 'clean_up_tokenization_spaces': False,})
generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=280, do_sample=False, temperature=0.0, eos_token_id=processor.tokenizer.eos_token_id,)
thread = Thread(target=model.generate, kwargs=generation_kwargs)
thread.start()
buffer = ""
for new_text in streamer:
buffer += new_text
yield buffer
chatbot = gr.Chatbot(scale=1, placeholder=PLACEHOLDER)
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:
gr.ChatInterface(
fn=bot_streaming,
title="Get Ripped with Arnold's AI Coach",
examples=[
{"text": "Identify and provide coaching cues for this exercise.", "files": ["./squat.jpg"]},
{"text": "What improvements can I make?", "files": ["./pushup.jpg"]},
{"text": "How is my form?", "files": ["./plank.jpg"]},
{"text": "Give me some tips to improve my deadlift.", "files": ["./deadlift.jpg"]}
],
description="Welcome to the ultimate fitness companion! 💪\nUpload a photo of your exercise and get instant feedback to perfect your form. Improve your workouts with expert tips!",
stop_btn="Stop Generation",
multimodal=True,
textbox=chat_input,
chatbot=chatbot,
cache_examples=False,
examples_per_page=3
)
demo.queue()
demo.launch(debug=True, quiet=True)
|