Spaces:
Running
on
Zero
Running
on
Zero
File size: 2,587 Bytes
5fc0263 f16a856 5fc0263 bbce3fa 5fc0263 2483355 bbce3fa 5fc0263 04683f7 5fc0263 04683f7 5fc0263 f16a856 fb4d7ba db2f499 f16a856 5fc0263 2483355 5fc0263 7a2d2cb 5fc0263 |
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 |
# Necessary imports
import sys
from typing import Any, Dict
import gradio as gr
import spaces
# Local imports
from src.utils.video_processing import encode_video
from src.config import (
device,
model_name,
sampling,
stream,
repetition_penalty,
)
from src.app.model import load_model_tokenizer_and_processor
from src.logger import logging
from src.exception import CustomExceptionHandling
# Model, tokenizer and processor
model, tokenizer, processor = load_model_tokenizer_and_processor(model_name, device)
@spaces.GPU(duration=120)
def describe_video(
video: str,
question: str,
temperature: float,
top_p: float,
top_k: int,
max_new_tokens: int,
) -> str:
"""
Describes a video by generating an answer to a given question.
Args:
- video (str): The path to the video file.
- question (str): The question to be answered about the video.
- temperature (float): The temperature parameter for the model.
- top_p (float): The top_p parameter for the model.
- top_k (int): The top_k parameter for the model.
- max_new_tokens (int): The max tokens to be generated by the model.
Returns:
str: The generated answer to the question.
"""
try:
# Check if video or question is None
if not video or not question:
gr.Warning("Please provide a video and a question.")
# Encode the video frames
frames = encode_video(video)
# Message format for the model
msgs = [{"role": "user", "content": frames + [question]}]
# Set decode params for video
params: Dict[str, Any] = {
"use_image_id": False,
"max_slice_nums": 1, # Use 1 if CUDA OOM and video resolution > 448*448
}
# Generate the answer
answer = model.chat(
image=None,
msgs=msgs,
tokenizer=tokenizer,
processor=processor,
sampling=sampling,
stream=stream,
top_p=top_p,
top_k=top_k,
temperature=temperature,
repetition_penalty=repetition_penalty,
max_new_tokens=max_new_tokens,
**params
)
# Log the successful generation of the answer
logging.info("Answer generated successfully.")
# Return the answer
return "".join(answer)
# Handle exceptions that may occur during answer generation
except Exception as e:
# Custom exception handling
raise CustomExceptionHandling(e, sys) from e
|