app.py

import gradio as gr
import os
import time
# from omegaconf import OmegaConf
import shutil
import os 
# import wget
import time 
variable = []
speech = ""
# context_2 = ""
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
from transformers import AutoTokenizer, AutoModel
import logging
import torch
import os
import base64

from pyannote.audio import Pipeline
from transformers import pipeline, AutoModelForCausalLM
from diarization_utils import diarize
from huggingface_hub import HfApi
from pydantic import ValidationError
from starlette.exceptions import HTTPException

# from config import model_settings, InferenceConfig

import logging

from pydantic import BaseModel
from pydantic_settings import BaseSettings
from typing import Optional, Literal

logger = logging.getLogger(__name__)


class ModelSettings(BaseSettings):
    asr_model: str
    assistant_model: Optional[str]
    diarization_model: Optional[str]
    hf_token: Optional[str]


class InferenceConfig(BaseModel):
    task: Literal["transcribe", "translate"] = "transcribe"
    batch_size: int = 24
    assisted: bool = False
    chunk_length_s: int = 30
    sampling_rate: int = 16000
    language: Optional[str] = None
    num_speakers: Optional[int] = None
    min_speakers: Optional[int] = None
    max_speakers: Optional[int] = None

# from nemo.collections.asr.parts.utils.diarization_utils import OfflineDiarWithASR
# from nemo.collections.asr.parts.utils.decoder_timestamps_utils import ASRDecoderTimeStamps
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
# logger.info(f"Using device: {device.type}")
torch_dtype = torch.float32 if device.type == "cpu" else torch.float16

tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm3-6b-32k", trust_remote_code=True)
model = AutoModel.from_pretrained("THUDM/chatglm3-6b-32k", trust_remote_code=True,device_map='auto')
# base_model = "lyogavin/Anima-7B-100K"
# tokenizer = AutoTokenizer.from_pretrained(base_model)
# model = AutoModelForCausalLM.from_pretrained(
#         base_model,
#         bnb_4bit_compute_dtype=torch.float16,
#         # torch_dtype=torch.float16,
#         trust_remote_code=True,
#         device_map="auto",
#         load_in_4bit=True 
#         )
# model.eval()

assistant_model = AutoModelForCausalLM.from_pretrained(
    "distil-whisper/distil-large-v3",
    torch_dtype=torch_dtype,
    low_cpu_mem_usage=True,
    use_safetensors=True
) 

assistant_model.to(device)

asr_pipeline = pipeline(
    "automatic-speech-recognition",
    model="openai/whisper-large-v3",
    torch_dtype=torch_dtype,
    device=device
)


HfApi().whoami(os.getenv('HF_TOKEN'))
diarization_pipeline = Pipeline.from_pretrained(
    checkpoint_path="pyannote/speaker-diarization-3.1",
    use_auth_token=os.getenv('HF_TOKEN'),
)
diarization_pipeline.to(device)


def upload_file(files):
    file_paths = [file.name for file in files]
    
    global variable
    variable = file_paths

    return file_paths




def audio_function():
    # Call the function and return its result to be displayed

    time_1 = time.time()
    paths = variable

    str1 = "processed speech"
    for i in paths:
        str1 = str1 + i
    
    str1=str1.replace("processed speech","")
    print("before processing ffmpeg ! ")

    command_to_mp4_to_wav =   "ffmpeg -i {}    current_out.wav -y"
    #-acodec pcm_s16le -ar 16000 -ac 1
    os.system(command_to_mp4_to_wav.format(str1))

    print("after ffmpeg")

    # os.system("insanely-fast-whisper  --file-name {}_new.wav  --task transcribe --hf_token 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".format(str1.replace("mp3","")))

    parameters = InferenceConfig()
    

    generate_kwargs = {
        "task": parameters.task, 
        "language": parameters.language,
        "assistant_model": assistant_model if parameters.assisted else None
    }

    with open("current_out.wav", 'rb') as f:
        audio_encoded = base64.b64encode(f.read()).decode("utf-8")
    file = base64.b64decode(audio_encoded)
    
    asr_outputs = asr_pipeline(
        file,
        chunk_length_s=parameters.chunk_length_s,
        batch_size=parameters.batch_size,
        generate_kwargs=generate_kwargs,
        return_timestamps=True,
    )
    

    
        
    transcript = diarize(diarization_pipeline, file, parameters, asr_outputs)
    global speech
    speech  = transcript
    return transcript,asr_outputs["chunks"],asr_outputs["text"]
    
def audio_function2():
    # Call the function and return its result to be displayed
    
    # global speech
    str2 = speech
    time_3   = time.time()

    
    # prompt = " {}  generate medical subjective objective assessment plan (soap) notes ?".format(str2) 
    prompt = " {}  summary of sales call ? is the agent qualified the lead properly ?".format(str2) 
    
    # model = model.eval()
    response, history = model.chat(tokenizer, prompt, history=[])
    print(response)
    # del model
    # del tokenizer
    # torch.cuda.empty_cache()
    time_4 = time.time()
    # response, history = model.chat(tokenizer, "晚上睡不着应该怎么办", history=history)
    # print(response)

    # inputs = tokenizer(prompt, return_tensors="pt")

    # inputs['input_ids'] = inputs['input_ids'].cuda()
    # inputs['attention_mask'] = inputs['attention_mask'].cuda()

    
    # generate_ids = model.generate(**inputs, max_new_tokens=4096,
    #                     only_last_logit=True, # to save memory
    #                     use_cache=False, # when run into OOM, enable this can save memory
    #                     xentropy=True)
    # output = tokenizer.batch_decode(generate_ids, 
    #                             skip_special_tokens=True,
    #                             clean_up_tokenization_spaces=False) 

    # tokenizer = AutoTokenizer.from_pretrained("togethercomputer/LLaMA-2-7B-32K")
    # model = AutoModelForCausalLM.from_pretrained("togethercomputer/LLaMA-2-7B-32K", trust_remote_code=True, torch_dtype=torch.float16,device_map="auto",bnb_4bit_compute_dtype=torch.float16,load_in_4bit=True)


    # input_context = "summarize "+" the following {}".format(str2) 
    # input_ids = tokenizer.encode(input_context, return_tensors="pt").cuda()
    # output = model.generate(input_ids, max_new_tokens=512, temperature=0.7)
    # output_text = tokenizer.decode(output[0], skip_special_tokens=True)
    # print(output_text,"wow what happened ")
    # return output
    return response,str(int(time_4-time_3)) + " seconds"


with gr.Blocks() as demo:
    file_output = gr.File()
    upload_button = gr.UploadButton("Click to Upload a File", file_types=["audio","video"], file_count="multiple")
    upload_button.upload(upload_file, upload_button, file_output)
    gr.Markdown("## Click process audio to display text from audio file")
    submit_button = gr.Button("Process Audio")
    output_text = gr.Textbox(label="Speech Diarization")
    output_text_2 = gr.Textbox(label="Speech chunks")
    submit_button.click(audio_function, outputs=[output_text,output_text_2,gr.Textbox(label=" asr_text :")])
    gr.Markdown("## Click the Summarize to display call summary")
    submit_button = gr.Button("Summarize")
    output_text = gr.Textbox(label="Sales Call Notes")
    submit_button.click(audio_function2, outputs=[output_text,gr.Textbox(label="Time Taken :")])

demo.launch()