Spaces:
Build error
Build error
File size: 10,549 Bytes
405b910 5950dc0 5949885 5950dc0 5949885 5950dc0 5949885 405b910 5950dc0 405b910 5950dc0 405b910 5950dc0 405b910 5950dc0 405b910 5950dc0 405b910 5950dc0 405b910 a4bf7ee e90bc6e 405b910 e90bc6e 5949885 586a110 a4bf7ee 586a110 405b910 586a110 5950dc0 9c9c777 5950dc0 405b910 5949885 5950dc0 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 |
import gradio as gr
import random
import torch
from transformers import AutoConfig, AutoTokenizer, AutoModelWithLMHead
from transformers import GPT2Tokenizer, GPT2LMHeadModel
from itertools import chain
import os
import tempfile
from typing import Optional
from TTS.config import load_config
import numpy as np
from TTS.utils.manage import ModelManager
from TTS.utils.synthesizer import Synthesizer
#emotion_tokenizer = AutoTokenizer.from_pretrained("mrm8488/t5-base-finetuned-emotion")
#emotion_model = AutoModelWithLMHead.from_pretrained("mrm8488/t5-base-finetuned-emotion")
def get_emotion(text):
input_ids = tokenizer.encode(text + '</s>', return_tensors='pt')
output = model.generate(input_ids=input_ids,max_length=2)
dec = [tokenizer.decode(ids) for ids in output]
label = dec[0]
return label.split()[1]
config = AutoConfig.from_pretrained('gorkemgoknar/gpt2chatbotenglish')
model = GPT2LMHeadModel.from_pretrained('gorkemgoknar/gpt2chatbotenglish', config=config)
tokenizer = GPT2Tokenizer.from_pretrained('gorkemgoknar/gpt2chatbotenglish')
tokenizer.model_max_length = 1024
#Dynamic Temperature
#See experiment https://www.linkedin.com/pulse/ai-goes-job-interview-g%25C3%25B6rkem-g%25C3%25B6knar
base_temperature = 1.2
dynamic_temperature_range = 0.15
rand_range = random.uniform(-1 * dynamic_temperature_range , dynamic_temperature_range )
temperature = base_temperature + rand_range
SPECIAL_TOKENS = ["<bos>", "<eos>", "<speaker1>", "<speaker2>", "<pad>"]
#See document for experiment https://www.linkedin.com/pulse/ai-goes-job-interview-g%C3%B6rkem-g%C3%B6knar/
def get_chat_response(name,history=[], input_txt = "Hello , what is your name?"):
ai_history = history.copy()
#ai_history.append(input_txt)
ai_history_e = [tokenizer.encode(e) for e in ai_history]
personality = "My name is " + name
bos, eos, speaker1, speaker2 = tokenizer.convert_tokens_to_ids(SPECIAL_TOKENS[:-1])
#persona first, history next, input text must be at the end
#[[bos, persona] , [history] , [input]]
sequence = [[bos] + tokenizer.encode(personality)] + ai_history_e + [tokenizer.encode(input_txt)]
##[[bos, persona] , [speaker1 .., speakser2 .., speaker1 ... speaker2 ... , [input]]
sequence = [sequence[0]] + [[speaker2 if (len(sequence)-i) % 2 else speaker1] + s for i, s in enumerate(sequence[1:])]
sequence = list(chain(*sequence))
#bot_input_ids = tokenizer.encode(personality + tokenizer.eos_token + input_txt + tokenizer.eos_token , return_tensors='pt')
sequence_len = len(sequence)
#optimum response and speed
chat_history_ids = model.generate(
torch.tensor(sequence).unsqueeze(0), max_length=50,
pad_token_id=tokenizer.eos_token_id,
no_repeat_ngram_size=3,
do_sample=True,
top_k=60,
top_p=0.8,
temperature = 1.3
)
out_str = tokenizer.decode(chat_history_ids[0][sequence_len:], skip_special_tokens=True)
#out_str = tokenizer.decode(chat_history_ids[:, sequence.shape[-1]:][0], skip_special_tokens=False)
return out_str
##you can use anyone from below
'''
| Macleod | Moran | Brenda | Ramirez | Peter Parker | Quentin Beck | Andy
| Red | Norton | Willard | Chief | Chef | Kilgore | Kurtz | Westley | Buttercup
| Vizzini | Fezzik | Inigo | Man In Black | Taylor | Zira | Zaius | Cornelius
| Bud | Lindsey | Hippy | Erin | Ed | George | Donna | Trinity | Agent Smith
| Morpheus | Neo | Tank | Meryl | Truman | Marlon | Christof | Stromboli | Bumstead
| Schreber | Walker | Korben | Cornelius | Loc Rhod | Anakin | Obi-Wan | Palpatine
| Padme | Superman | Luthor | Dude | Walter | Donny | Maude | General | Starkiller
| Indiana | Willie | Short Round | John | Sarah | Terminator | Miller | Sarge | Reiben
| Jackson | Upham | Chuckie | Will | Lambeau | Sean | Skylar | Saavik | Spock
| Kirk | Bones | Khan | Kirk | Spock | Sybok | Scotty | Bourne | Pamela | Abbott
| Nicky | Marshall | Korshunov | Troy | Vig | Archie Gates | Doc | Interrogator
| Ellie | Ted | Peter | Drumlin | Joss | Macready | Childs | Nicholas | Conrad
| Feingold | Christine | Adam | Barbara | Delia | Lydia | Cathy | Charles | Otho
| Schaefer | Han | Luke | Leia | Threepio | Vader | Yoda | Lando | Elaine | Striker
| Dr. Rumack | Kramer | David | Saavik | Kirk | Kruge | Holden | Deckard | Rachael
| Batty | Sebastian | Sam | Frodo | Pippin | Gandalf | Kay | Edwards | Laurel
| Edgar | Zed | Jay | Malloy | Plissken | Steve Rogers | Tony Stark | Scott Lang
| Bruce Banner | Bruce | Edward | Two-Face | Batman | Chase | Alfred | Dick
| Riddler | Din Djarin | Greef Karga | Kuiil | Ig-11 | Cara Dune | Peli Motto
| Toro Calican | Ripley | Meredith | Dickie | Marge | Peter | Lambert | Kane
| Dallas | Ripley | Ash | Parker | Threepio | Luke | Leia | Ben | Han | Common Bob
| Common Alice | Jack | Tyler | Marla | Dana | Stantz | Venkman | Spengler | Louis
| Fry | Johns | Riddick | Kirk | Decker | Spock | "Ilia | Indy | Belloq | Marion
| Brother | Allnut | Rose | Qui-Gon | Jar Jar
'''
MODEL_NAME= "tts_models/multilingual/multi-dataset/your_tts"
def greet(character,your_voice,message,history):
#gradios set_state/get_state had problems on embedded html!
history = history or {"character": character, "message_history" : [] }
#gradios set_state/get_state does not persist session for now using global
#global history
if history["character"] != character:
#switching character
history = {"character": character, "message_history" : [] }
response = get_chat_response(character,history=history["message_history"],input_txt=message)
os.system('tts --text "'+response+'" --model_name tts_models/multilingual/multi-dataset/your_tts --speaker_wav '+your_voice+' --language_idx "en"')
history["message_history"].append((message, response))
#emotion = get_emotion(response)
html = "<div class='chatbot'>"
for user_msg, resp_msg in history["message_history"]:
html += f"<div class='user_msg'>You: {user_msg}</div>"
html += f"<div class='resp_msg'>{character}: {resp_msg}</div>"
html += "</div>"
return html,history,"tts_output.wav"
def greet_textonly(character,message,history):
#gradios set_state/get_state had problems on embedded html!
history = history or {"character": character, "message_history" : [] }
#gradios set_state/get_state does not persist session for now using global
#global history
if history["character"] != character:
#switching character
history = {"character": character, "message_history" : [] }
response = get_chat_response(character,history=history["message_history"],input_txt=message)
history["message_history"].append((message, response))
#emotion = get_emotion(response)
html = "<div class='chatbot'>"
for user_msg, resp_msg in history["message_history"]:
html += f"<div class='user_msg'>You: {user_msg}</div>"
html += f"<div class='resp_msg'>{character}: {resp_msg}</div>"
html += "</div>"
return html,history
personality_choices = ["Gandalf", "Riddick", "Macleod", "Morpheus", "Neo","Spock","Vader","Indy"]
examples= ["Gandalf", "What is your name?"]
css="""
.chatbox {display:flex;flex-direction:column}
.user_msg, .resp_msg {padding:4px;margin-bottom:4px;border-radius:4px;width:80%}
.user_msg {background-color:cornflowerblue;color:white;align-self:start}
.resp_msg {background-color:lightgray;align-self:self-end}
"""
#some selected ones are in for demo use
personality_choices = ["Gandalf", "Riddick", "Macleod", "Morpheus", "Neo","Spock","Vader","Indy", "Ig-11","Threepio","Tony Stark","Batman","Vizzini"]
title = "Movie Chatbot with Coqui YourTTS"
description = "Chat with your favorite movie characters, making characters voice like you. See Coqui Space for more TTS models https://huggingface.co/spaces/coqui/CoquiTTS"
article = "<p style='text-align: center'><a href='https://www.linkedin.com/pulse/ai-goes-job-interview-g%C3%B6rkem-g%C3%B6knar/' target='_blank'>AI Goes to Job Interview</a> | <a href='https://www.metayazar.com/' target='_blank'>Metayazar AI Writer</a> |<a href='https://www.linkedin.com/in/goknar/' target='_blank'>Görkem Göknar</a></p>"
#History not implemented in this demo, use metayazar.com/chatbot for a movie and character dropdown chat interface
##interface = gr.Interface(fn=greet, inputs=[gr.inputs.Dropdown(personality_choices) ,"text"], title=title, description=description, outputs="text")
examples=[['Gandalf','dragon.wav','Who are you sir?',{}]]
history = {"character": "None", "message_history" : [] }
interface_mic= gr.Interface(fn=greet,
inputs=[gr.inputs.Dropdown(personality_choices),
gr.inputs.Audio(source="microphone", type="filepath") ,
"text",
"state"],
outputs=["html","state",gr.outputs.Audio(type="file")],
css=css, title="Chat with Your Voice", description=description,article=article )
interface_text = gr.Interface(fn=greet,
inputs=[gr.inputs.Dropdown(personality_choices),
"text",
"state"],
outputs=["html","state"],
css=css, title="Chat Text Only", description=description,article=article)
interface_mic= gr.Interface(fn=greet,
inputs=[gr.inputs.Dropdown(personality_choices),
gr.inputs.Audio(source="microphone", type="filepath") ,
"text",
"state"],
outputs=["html","state",gr.outputs.Audio(type="file")],
css=css, title=title, description=description,article=article )
interface_file= gr.Interface.load(greet_textonly,
inputs=[gr.inputs.Dropdown(personality_choices),
gr.inputs.Audio(type="filepath"),
"text",
"state"],
outputs=["html","state",gr.outputs.Audio(type="file")],
css=css, title="Chat with Uploaded file", description=description,article=article )
appinterface = gr.TabbedInterface([interface_mic,interface_file, interface_text], ["Chat with Mic Record","Chat with Audio Upload" , "Chat Text only"])
appinterface.launch() |