Spaces:
Runtime error
Runtime error
File size: 3,263 Bytes
2ac2001 cff67c8 2ac2001 babdbe7 143381d 2ac2001 babdbe7 2ac2001 586fe20 babdbe7 2ac2001 2bf6cea 104d077 2ac2001 8ea9610 babdbe7 b111794 babdbe7 2ac2001 02ca315 818cb74 2ac2001 |
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 |
import gradio as gr
import os
import time
import spaces
import torch
import re
import gradio as gr
from threading import Thread
from transformers import TextIteratorStreamer, AutoTokenizer, AutoModelForCausalLM
from PIL import Image
import subprocess
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
model_id = "vikhyatk/moondream2"
revision = "2024-04-02"
tokenizer = AutoTokenizer.from_pretrained(model_id, revision=revision)
moondream = AutoModelForCausalLM.from_pretrained(
model_id, trust_remote_code=True, revision=revision,
torch_dtype=torch.bfloat16, device_map={"": "cuda"},
attn_implementation="flash_attention_2"
)
moondream.eval()
def print_like_dislike(x: gr.LikeData):
print(x.index, x.value, x.liked)
def add_message(history, message):
# Handle image and text input
if message["files"]:
for x in message["files"]:
history.append(((x,), None))
if message["text"] is not None:
history.append((message["text"], None))
return history, gr.MultimodalTextbox(value=None, interactive=False)
@spaces.GPU(duration=10)
def bot(history):
# Reverse search through the last 5 messages for an image file
last_five_messages = history[-5:] # Get the last five messages
image_path = None
last_message = None
for message in last_five_messages:
if isinstance(message[0], tuple) and isinstance(message[0][0], str):
image_path = message[0][0]
if isinstance(message[0],str):
last_message = message[0]
if image_path:
try:
image = Image.open(image_path) # Try to open the image using Pillow
image_embeds = moondream.encode_image(image)
print(image_embeds.shape)
response = moondream.answer_question(image_embeds, last_message, tokenizer)
except IOError:
response = "Failed to open image. Please check the image path or file permissions."
else:
image_embeds = torch.zeros(1, 729, 2048, dtype=torch.bfloat16, device='cuda')
response = moondream.answer_question(image_embeds, last_message, tokenizer)
history[-1][1] = ""
for character in response:
history[-1][1] += character
yield history
with gr.Blocks(theme=gr.themes.Default(primary_hue="green", secondary_hue="emerald")) as demo:
gr.Markdown(
"""
# AskMoondream: Moondream 2 Demonstration Space
## Modularity AI presents this open source huggingface space for running fast experimental inferences on Moondream2.
"""
)
chatbot = gr.Chatbot(
[],
elem_id="chatbot",
bubble_full_width=False,
height = 550
)
chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)
chat_msg = chat_input.submit(add_message, inputs=[chatbot, chat_input], outputs=[chatbot, chat_input])
bot_msg = chat_msg.then(bot, inputs=chatbot, outputs=chatbot, api_name="bot_response")
bot_msg.then(lambda: gr.MultimodalTextbox(interactive=True), None, outputs=[chat_input])
chatbot.like(print_like_dislike, None, None)
demo.queue()
demo.launch()
|