Spaces:
Runtime error
Runtime error
File size: 4,393 Bytes
4b4ae5c 1d54733 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 | # imports
import os
from huggingface_hub import login
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline, BitsAndBytesConfig, AutoProcessor, AutoModelForSpeechSeq2Seq
import gradio as gr
# Log in to HuggingFace
login(token=os.getenv("HF_TOKEN"))
# quantization setup
quant_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_use_double_quant=True,
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_quant_type="nf4"
)
# constants
LLAMA = "meta-llama/Meta-Llama-3.1-8B-Instruct"
device = "cuda:0" if torch.cuda.is_available() else "cpu"
tokenizer = AutoTokenizer.from_pretrained(LLAMA)
tokenizer.pad_token = tokenizer.eos_token
model = AutoModelForCausalLM.from_pretrained(LLAMA, use_auth_token=os.getenv("HF_TOKEN"), device_map="cuda:0", quantization_config=quant_config)
# load Speech model, read audio file and convert to text
def transcript_audio(audio_file):
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
model_id = "openai/whisper-large-v3-turbo"
model = AutoModelForSpeechSeq2Seq.from_pretrained(
model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
)
model.to(device)
processor = AutoProcessor.from_pretrained(model_id)
pipe = pipeline(
"automatic-speech-recognition",
model=model,
tokenizer=processor.tokenizer,
feature_extractor=processor.feature_extractor,
torch_dtype=torch_dtype,
device=device,
)
result = pipe(audio_file, return_timestamps=True)
return result
# promopts and summarizing the text
def summarize_with_llama(transcript, context="Phone Call", language="English"):
if context == "Phone Call":
system_message = (
"You are an assistant that produces minutes of phone calls from transcripts, "
"with summary and key discussion points, in markdown."
)
user_prompt = (
"Below is an extract transcript of a phone call. "
"Please write minutes in markdown, including a summary, location, and discussion points."
)
elif context == "Meeting":
system_message = (
"You are an assistant that produces minutes of meetings from transcripts, "
"with summary, key discussion points, takeaways, and action items with owners, in markdown."
)
user_prompt = (
"Below is an extract transcript of a meeting. "
"Please write minutes in markdown, including a summary with attendees, location, and date; "
"discussion points; takeaways; and action items with owners."
)
else:
raise ValueError(f"Unknown context: {context}")
# Add language instruction
if language.lower() == "hebrew":
user_prompt += "\n\nWrite the entire generatated summary and points in Hebrew."
elif language.lower() == "english":
user_prompt += "\n\nWrite the entire summary in English."
else:
raise ValueError(f"Unsupported language: {language}")
messages = [
{"role": "system", "content": system_message},
{"role": "user", "content": f"{system_message}\n\n{user_prompt}\n\nTranscript:\n{transcript}"}
]
input_features = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda")
output_ids = model.generate(input_features, max_new_tokens=2000)
summary = tokenizer.decode(output_ids[0], skip_special_tokens=True).strip()
return summary
# combining the two functions
def transcribe_and_summarize(audio, choices, language):
try:
result = transcript_audio(audio)
transcript = result["text"]
return summarize_with_llama(transcript, context=choices, language=language)
except Exception as e:
return f"**Error:** {str(e)}"
# Gradio UI setup
demo = gr.Interface(
fn=transcribe_and_summarize,
inputs=[
gr.Audio(type="filepath", label="Upload Audio"),
gr.Radio(["Phone Call", "Meeting"], label="Select Audio Type"),
gr.Radio(["English", "Hebrew"], label="Select Language Output")
],
outputs=gr.Markdown(height=750),
title="Audio Summarizer"
)
# Required for HF Spaces
if __name__ == "__main__":
demo.launch() |