app.py

import gradio as gr
import os, gc, torch
from datetime import datetime
from huggingface_hub import hf_hub_download
from pynvml import *
nvmlInit()
gpu_h = nvmlDeviceGetHandleByIndex(0)
ctx_limit = 1024
title1 = "RWKV-4-Raven-7B-v8-Eng-20230408-ctx4096"

os.environ["RWKV_JIT_ON"] = '1'
os.environ["RWKV_CUDA_ON"] = '1' # if '1' then use CUDA kernel for seq mode (much faster)

from rwkv.model import RWKV
model_path = hf_hub_download(repo_id="BlinkDL/rwkv-4-raven", filename=f"{title1}.pth")
model = RWKV(model=model_path, strategy='cuda fp16i8 *8 -> cuda fp16')
from rwkv.utils import PIPELINE, PIPELINE_ARGS
pipeline = PIPELINE(model, "20B_tokenizer.json")

from TTS.api import TTS
tts = TTS(model_name="tts_models/multilingual/multi-dataset/your_tts", progress_bar=False, gpu=True)
import whisper
model1 = whisper.load_model("small")

os.system('pip install voicefixer --upgrade')
from voicefixer import VoiceFixer
voicefixer = VoiceFixer()

import torchaudio
from speechbrain.pretrained import SpectralMaskEnhancement

enhance_model = SpectralMaskEnhancement.from_hparams(
source="speechbrain/metricgan-plus-voicebank",
savedir="pretrained_models/metricgan-plus-voicebank",
run_opts={"device":"cuda"},
)

def generate_prompt(instruction, input=None):
    if input:
        return f"""Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
# Instruction:
{instruction}
# Input:
{input}
# Response:
"""
    else:
        return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
# Instruction:
{instruction}
# Response:
"""

def evaluate(
    upload,
    audio,
#    instruction,
#    input=None,
#    token_count=200,
#    temperature=1.0,
#    top_p=0.7,
#    presencePenalty = 0.1,
#    countPenalty = 0.1,
):

    audio = whisper.load_audio(audio)
    audio = whisper.pad_or_trim(audio)

    # make log-Mel spectrogram and move to the same device as the model1
    mel = whisper.log_mel_spectrogram(audio).to(model1.device)

    # detect the spoken language
    _, probs = model1.detect_language(mel)
    print(f"Detected language: {max(probs, key=probs.get)}")

    # decode the audio
    options = whisper.DecodingOptions()
    result = whisper.decode(model1, mel, options)

    
    res = []
    args = PIPELINE_ARGS(temperature = max(0.2, float(1)), top_p = float(0.5),
                     alpha_frequency = 0.4,
                     alpha_presence = 0.4,
                     token_ban = [], # ban the generation of some tokens
                     token_stop = [0]) # stop generation whenever you see any token here

    instruction = result.text.strip()
    input=None
#    input = input.strip()
    ctx = generate_prompt(instruction, input)
    
    gpu_info = nvmlDeviceGetMemoryInfo(gpu_h)
    print(f'vram {gpu_info.total} used {gpu_info.used} free {gpu_info.free}')
    
    all_tokens = []
    out_last = 0
    out_str = ''
    occurrence = {}
    state = None
    for i in range(int(150)):
        out, state = model.forward(pipeline.encode(ctx)[-ctx_limit:] if i == 0 else [token], state)
        for n in occurrence:
            out[n] -= (args.alpha_presence + occurrence[n] * args.alpha_frequency)

        token = pipeline.sample_logits(out, temperature=args.temperature, top_p=args.top_p)
        if token in args.token_stop:
            break
        all_tokens += [token]
        if token not in occurrence:
            occurrence[token] = 1
        else:
            occurrence[token] += 1
        
        tmp = pipeline.decode(all_tokens[out_last:])
        if '\ufffd' not in tmp:
            out_str += tmp
            yield out_str.strip()
            out_last = i + 1
    gc.collect()
    torch.cuda.empty_cache()

    res.append(out_str.strip())

#    res1 = ''.join(str(x) for x in res)

    tts.tts_to_file(res, speaker_wav = upload, language="en", file_path="output.wav")
    
    voicefixer.restore(input="output.wav", # input wav file path
                    output="audio1.wav", # output wav file path
                    cuda=True, # whether to use gpu acceleration
                    mode = 0) # You can try out mode 0, 1, or 2 to find out the best result
    
    noisy = enhance_model.load_audio(
    "audio1.wav"
    ).unsqueeze(0)

    enhanced = enhance_model.enhance_batch(noisy, lengths=torch.tensor([1.]))
    torchaudio.save("enhanced.wav", enhanced.cpu(), 16000)
    
    return [result.text, res, "enhanced.wav"]
    
#    yield out_str.strip()

g = gr.Interface(
    fn=evaluate,
    inputs=[
        gr.Audio(source="upload", label = "请上传您喜欢的声音(wav文件)", type="filepath"),
        gr.Audio(source="microphone", label = "和您的专属AI聊天吧！", type="filepath"),
#        gr.components.Textbox(lines=2, label="Instruction", value="Tell me about ravens."),
#        gr.components.Textbox(lines=2, label="Input", placeholder="none"),
#        gr.components.Slider(minimum=10, maximum=200, step=10, value=150), # token_count
#        gr.components.Slider(minimum=0.2, maximum=2.0, step=0.1, value=1.0), # temperature
#        gr.components.Slider(minimum=0, maximum=1, step=0.05, value=0.5), # top_p
#        gr.components.Slider(0.0, 1.0, step=0.1, value=0.4),  # presencePenalty
#        gr.components.Slider(0.0, 1.0, step=0.1, value=0.4),  # countPenalty        
    ],
    outputs=[
        gr.Textbox(label="Speech to Text"),
        gr.Textbox(label="Raven Output"),
        gr.Audio(label="Audio with Custom Voice"),
    ],
    title="🥳💬💕 - TalktoAI，随时随地，谈天说地！",
    description="🤖 - 让有人文关怀的AI造福每一个人！AI向善，文明璀璨！TalktoAI - Enable the future！",
    article = "Powered by the RWKV Language Model"
)
g.queue(concurrency_count=1, max_size=10)
g.launch(show_error=True)