Spaces:
Running
Running
Commit
•
4eb36d2
1
Parent(s):
f513635
Update app.py (#3)
Browse files- Update app.py (fe7399d2bb60ee8104736952121646297025506e)
- Update requirements.txt (f65e4a52d4d5d2173891b195bee96c3b4462dbdb)
Co-authored-by: Yoach Lacombe <ylacombe@users.noreply.huggingface.co>
- app.py +55 -84
- requirements.txt +1 -1
app.py
CHANGED
@@ -9,22 +9,26 @@ import numpy as np
|
|
9 |
import spaces
|
10 |
import gradio as gr
|
11 |
import torch
|
|
|
|
|
12 |
|
13 |
from parler_tts import ParlerTTSForConditionalGeneration
|
14 |
from pydub import AudioSegment
|
15 |
from transformers import AutoTokenizer, AutoFeatureExtractor, set_seed
|
16 |
|
17 |
-
|
|
|
|
|
18 |
torch_dtype = torch.bfloat16 if device != "cpu" else torch.float32
|
19 |
|
20 |
repo_id = "ai4bharat/indic-parler-tts-pretrained"
|
21 |
-
|
22 |
|
23 |
model = ParlerTTSForConditionalGeneration.from_pretrained(
|
24 |
repo_id, attn_implementation="eager", torch_dtype=torch_dtype,
|
25 |
).to(device)
|
26 |
-
|
27 |
-
|
28 |
).to(device)
|
29 |
|
30 |
tokenizer = AutoTokenizer.from_pretrained(repo_id)
|
@@ -89,7 +93,7 @@ examples = [
|
|
89 |
]
|
90 |
|
91 |
|
92 |
-
|
93 |
[
|
94 |
"मुले बागेत खेळत आहेत आणि पक्षी किलबिलाट करत आहेत.",
|
95 |
"Sunita speaks slowly in a calm, moderate-pitched voice, delivering the news with a neutral tone. The recording is very high quality with no background noise.",
|
@@ -171,44 +175,30 @@ def numpy_to_mp3(audio_array, sampling_rate):
|
|
171 |
sampling_rate = model.audio_encoder.config.sampling_rate
|
172 |
frame_rate = model.audio_encoder.config.frame_rate
|
173 |
|
174 |
-
# @spaces.GPU
|
175 |
-
# def generate_base(text, description, play_steps_in_s=2.0):
|
176 |
-
# play_steps = int(frame_rate * play_steps_in_s)
|
177 |
-
# streamer = ParlerTTSStreamer(model, device=device, play_steps=play_steps)
|
178 |
-
|
179 |
-
# inputs = description_tokenizer(description, return_tensors="pt").to(device)
|
180 |
-
# prompt = tokenizer(text, return_tensors="pt").to(device)
|
181 |
-
|
182 |
-
# generation_kwargs = dict(
|
183 |
-
# input_ids=inputs.input_ids,
|
184 |
-
# prompt_input_ids=prompt.input_ids,
|
185 |
-
# streamer=streamer,
|
186 |
-
# do_sample=True,
|
187 |
-
# temperature=1.0,
|
188 |
-
# min_new_tokens=10,
|
189 |
-
# )
|
190 |
-
|
191 |
-
# set_seed(SEED)
|
192 |
-
# thread = Thread(target=model.generate, kwargs=generation_kwargs)
|
193 |
-
# thread.start()
|
194 |
-
|
195 |
-
# for new_audio in streamer:
|
196 |
-
# print(f"Sample of length: {round(new_audio.shape[0] / sampling_rate, 2)} seconds")
|
197 |
-
# yield numpy_to_mp3(new_audio, sampling_rate=sampling_rate)
|
198 |
-
|
199 |
@spaces.GPU
|
200 |
-
def generate_base(text, description,
|
201 |
# Initialize variables
|
202 |
-
|
203 |
-
chunk_size = 15 # Process 10 words at a time
|
204 |
|
205 |
# Tokenize the full text and description
|
206 |
inputs = description_tokenizer(description, return_tensors="pt").to(device)
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
chunks = [
|
211 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
212 |
all_audio = []
|
213 |
|
214 |
# Process each chunk
|
@@ -223,8 +213,6 @@ def generate_base(text, description, play_steps_in_s=2.0):
|
|
223 |
prompt_input_ids=prompt.input_ids,
|
224 |
prompt_attention_mask=prompt.attention_mask,
|
225 |
do_sample=True,
|
226 |
-
# temperature=1.0,
|
227 |
-
# min_new_tokens=10,
|
228 |
return_dict_in_generate=True
|
229 |
)
|
230 |
|
@@ -243,43 +231,30 @@ def generate_base(text, description, play_steps_in_s=2.0):
|
|
243 |
print(f"Sample of length: {round(combined_audio.shape[0] / sampling_rate, 2)} seconds")
|
244 |
yield numpy_to_mp3(combined_audio, sampling_rate=sampling_rate)
|
245 |
|
246 |
-
# @spaces.GPU
|
247 |
-
# def generate_jenny(text, description, play_steps_in_s=2.0):
|
248 |
-
# play_steps = int(frame_rate * play_steps_in_s)
|
249 |
-
# streamer = ParlerTTSStreamer(jenny_model, device=device, play_steps=play_steps)
|
250 |
-
|
251 |
-
# inputs = description_tokenizer(description, return_tensors="pt").to(device)
|
252 |
-
# prompt = tokenizer(text, return_tensors="pt").to(device)
|
253 |
-
|
254 |
-
# generation_kwargs = dict(
|
255 |
-
# input_ids=inputs.input_ids,
|
256 |
-
# prompt_input_ids=prompt.input_ids,
|
257 |
-
# streamer=streamer,
|
258 |
-
# do_sample=True,
|
259 |
-
# temperature=1.0,
|
260 |
-
# min_new_tokens=10,
|
261 |
-
# )
|
262 |
-
|
263 |
-
# set_seed(SEED)
|
264 |
-
# thread = Thread(target=jenny_model.generate, kwargs=generation_kwargs)
|
265 |
-
# thread.start()
|
266 |
-
|
267 |
-
# for new_audio in streamer:
|
268 |
-
# print(f"Sample of length: {round(new_audio.shape[0] / sampling_rate, 2)} seconds")
|
269 |
-
# yield sampling_rate, new_audio
|
270 |
|
271 |
@spaces.GPU
|
272 |
-
def
|
273 |
# Initialize variables
|
274 |
-
|
275 |
-
chunk_size = 15 # Process 10 words at a time
|
276 |
|
277 |
# Tokenize the full text and description
|
278 |
inputs = description_tokenizer(description, return_tensors="pt").to(device)
|
279 |
-
|
280 |
-
|
281 |
-
|
282 |
-
chunks = [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
283 |
|
284 |
all_audio = []
|
285 |
|
@@ -289,14 +264,12 @@ def generate_jenny(text, description, play_steps_in_s=2.0):
|
|
289 |
prompt = tokenizer(chunk, return_tensors="pt").to(device)
|
290 |
|
291 |
# Generate audio for the chunk
|
292 |
-
generation =
|
293 |
input_ids=inputs.input_ids,
|
294 |
attention_mask=inputs.attention_mask,
|
295 |
prompt_input_ids=prompt.input_ids,
|
296 |
prompt_attention_mask=prompt.attention_mask,
|
297 |
do_sample=True,
|
298 |
-
# temperature=1.0,
|
299 |
-
# min_new_tokens=10,
|
300 |
return_dict_in_generate=True
|
301 |
)
|
302 |
|
@@ -387,29 +360,27 @@ with gr.Blocks(css=css) as block:
|
|
387 |
with gr.Tab("Finetuned"):
|
388 |
with gr.Row():
|
389 |
with gr.Column():
|
390 |
-
input_text = gr.Textbox(label="Input Text", lines=2, value=
|
391 |
-
description = gr.Textbox(label="Description", lines=2, value=
|
392 |
-
play_seconds = gr.Slider(3.0, 7.0, value=jenny_examples[0][2], step=2, label="Streaming interval in seconds", info="Lower = shorter chunks, lower latency, more codec steps")
|
393 |
run_button = gr.Button("Generate Audio", variant="primary")
|
394 |
with gr.Column():
|
395 |
-
audio_out = gr.Audio(label="Parler-TTS generation", format="mp3", elem_id="audio_out",
|
396 |
|
397 |
-
inputs = [input_text, description
|
398 |
outputs = [audio_out]
|
399 |
-
gr.Examples(examples=
|
400 |
-
run_button.click(fn=
|
401 |
|
402 |
with gr.Tab("Pretrained"):
|
403 |
with gr.Row():
|
404 |
with gr.Column():
|
405 |
input_text = gr.Textbox(label="Input Text", lines=2, value=default_text, elem_id="input_text")
|
406 |
description = gr.Textbox(label="Description", lines=2, value="", elem_id="input_description")
|
407 |
-
play_seconds = gr.Slider(3.0, 7.0, value=3.0, step=2, label="Streaming interval in seconds", info="Lower = shorter chunks, lower latency, more codec steps")
|
408 |
run_button = gr.Button("Generate Audio", variant="primary")
|
409 |
with gr.Column():
|
410 |
-
audio_out = gr.Audio(label="Parler-TTS generation", format="mp3", elem_id="audio_out",
|
411 |
|
412 |
-
inputs = [input_text, description
|
413 |
outputs = [audio_out]
|
414 |
gr.Examples(examples=examples, fn=generate_base, inputs=inputs, outputs=outputs, cache_examples=False)
|
415 |
run_button.click(fn=generate_base, inputs=inputs, outputs=outputs, queue=True)
|
|
|
9 |
import spaces
|
10 |
import gradio as gr
|
11 |
import torch
|
12 |
+
import nltk
|
13 |
+
|
14 |
|
15 |
from parler_tts import ParlerTTSForConditionalGeneration
|
16 |
from pydub import AudioSegment
|
17 |
from transformers import AutoTokenizer, AutoFeatureExtractor, set_seed
|
18 |
|
19 |
+
nltk.download('punkt_tab')
|
20 |
+
|
21 |
+
device = "cuda:0" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
|
22 |
torch_dtype = torch.bfloat16 if device != "cpu" else torch.float32
|
23 |
|
24 |
repo_id = "ai4bharat/indic-parler-tts-pretrained"
|
25 |
+
finetuned_repo_id = "ai4bharat/indic-parler-tts"
|
26 |
|
27 |
model = ParlerTTSForConditionalGeneration.from_pretrained(
|
28 |
repo_id, attn_implementation="eager", torch_dtype=torch_dtype,
|
29 |
).to(device)
|
30 |
+
finetuned_model = ParlerTTSForConditionalGeneration.from_pretrained(
|
31 |
+
finetuned_repo_id, attn_implementation="eager", torch_dtype=torch_dtype,
|
32 |
).to(device)
|
33 |
|
34 |
tokenizer = AutoTokenizer.from_pretrained(repo_id)
|
|
|
93 |
]
|
94 |
|
95 |
|
96 |
+
finetuned_examples = [
|
97 |
[
|
98 |
"मुले बागेत खेळत आहेत आणि पक्षी किलबिलाट करत आहेत.",
|
99 |
"Sunita speaks slowly in a calm, moderate-pitched voice, delivering the news with a neutral tone. The recording is very high quality with no background noise.",
|
|
|
175 |
sampling_rate = model.audio_encoder.config.sampling_rate
|
176 |
frame_rate = model.audio_encoder.config.frame_rate
|
177 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
@spaces.GPU
|
179 |
+
def generate_base(text, description,):
|
180 |
# Initialize variables
|
181 |
+
chunk_size = 25 # Process max 25 words or a sentence at a time
|
|
|
182 |
|
183 |
# Tokenize the full text and description
|
184 |
inputs = description_tokenizer(description, return_tensors="pt").to(device)
|
185 |
+
|
186 |
+
sentences_text = nltk.sent_tokenize(text) # this gives us a list of sentences
|
187 |
+
curr_sentence = ""
|
188 |
+
chunks = []
|
189 |
+
for sentence in sentences_text:
|
190 |
+
candidate = " ".join([curr_sentence, sentence])
|
191 |
+
if len(candidate.split()) >= chunk_size:
|
192 |
+
chunks.append(curr_sentence)
|
193 |
+
curr_sentence = sentence
|
194 |
+
else:
|
195 |
+
curr_sentence = candidate
|
196 |
+
|
197 |
+
if curr_sentence != "":
|
198 |
+
chunks.append(curr_sentence)
|
199 |
+
|
200 |
+
print(chunks)
|
201 |
+
|
202 |
all_audio = []
|
203 |
|
204 |
# Process each chunk
|
|
|
213 |
prompt_input_ids=prompt.input_ids,
|
214 |
prompt_attention_mask=prompt.attention_mask,
|
215 |
do_sample=True,
|
|
|
|
|
216 |
return_dict_in_generate=True
|
217 |
)
|
218 |
|
|
|
231 |
print(f"Sample of length: {round(combined_audio.shape[0] / sampling_rate, 2)} seconds")
|
232 |
yield numpy_to_mp3(combined_audio, sampling_rate=sampling_rate)
|
233 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
234 |
|
235 |
@spaces.GPU
|
236 |
+
def generate_finetuned(text, description):
|
237 |
# Initialize variables
|
238 |
+
chunk_size = 25 # Process max 25 words or a sentence at a time
|
|
|
239 |
|
240 |
# Tokenize the full text and description
|
241 |
inputs = description_tokenizer(description, return_tensors="pt").to(device)
|
242 |
+
|
243 |
+
sentences_text = nltk.sent_tokenize(text) # this gives us a list of sentences
|
244 |
+
curr_sentence = ""
|
245 |
+
chunks = []
|
246 |
+
for sentence in sentences_text:
|
247 |
+
candidate = " ".join([curr_sentence, sentence])
|
248 |
+
if len(candidate.split()) >= chunk_size:
|
249 |
+
chunks.append(curr_sentence)
|
250 |
+
curr_sentence = sentence
|
251 |
+
else:
|
252 |
+
curr_sentence = candidate
|
253 |
+
|
254 |
+
if curr_sentence != "":
|
255 |
+
chunks.append(curr_sentence)
|
256 |
+
|
257 |
+
print(chunks)
|
258 |
|
259 |
all_audio = []
|
260 |
|
|
|
264 |
prompt = tokenizer(chunk, return_tensors="pt").to(device)
|
265 |
|
266 |
# Generate audio for the chunk
|
267 |
+
generation = finetuned_model.generate(
|
268 |
input_ids=inputs.input_ids,
|
269 |
attention_mask=inputs.attention_mask,
|
270 |
prompt_input_ids=prompt.input_ids,
|
271 |
prompt_attention_mask=prompt.attention_mask,
|
272 |
do_sample=True,
|
|
|
|
|
273 |
return_dict_in_generate=True
|
274 |
)
|
275 |
|
|
|
360 |
with gr.Tab("Finetuned"):
|
361 |
with gr.Row():
|
362 |
with gr.Column():
|
363 |
+
input_text = gr.Textbox(label="Input Text", lines=2, value=finetuned_examples[0][0], elem_id="input_text")
|
364 |
+
description = gr.Textbox(label="Description", lines=2, value=finetuned_examples[0][1], elem_id="input_description")
|
|
|
365 |
run_button = gr.Button("Generate Audio", variant="primary")
|
366 |
with gr.Column():
|
367 |
+
audio_out = gr.Audio(label="Parler-TTS generation", format="mp3", elem_id="audio_out", autoplay=True)
|
368 |
|
369 |
+
inputs = [input_text, description]
|
370 |
outputs = [audio_out]
|
371 |
+
gr.Examples(examples=finetuned_examples, fn=generate_finetuned, inputs=inputs, outputs=outputs, cache_examples=False)
|
372 |
+
run_button.click(fn=generate_finetuned, inputs=inputs, outputs=outputs, queue=True)
|
373 |
|
374 |
with gr.Tab("Pretrained"):
|
375 |
with gr.Row():
|
376 |
with gr.Column():
|
377 |
input_text = gr.Textbox(label="Input Text", lines=2, value=default_text, elem_id="input_text")
|
378 |
description = gr.Textbox(label="Description", lines=2, value="", elem_id="input_description")
|
|
|
379 |
run_button = gr.Button("Generate Audio", variant="primary")
|
380 |
with gr.Column():
|
381 |
+
audio_out = gr.Audio(label="Parler-TTS generation", format="mp3", elem_id="audio_out", autoplay=True)
|
382 |
|
383 |
+
inputs = [input_text, description]
|
384 |
outputs = [audio_out]
|
385 |
gr.Examples(examples=examples, fn=generate_base, inputs=inputs, outputs=outputs, cache_examples=False)
|
386 |
run_button.click(fn=generate_base, inputs=inputs, outputs=outputs, queue=True)
|
requirements.txt
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
torch
|
2 |
spaces
|
3 |
git+https://github.com/huggingface/parler-tts.git
|
4 |
-
|
|
|
1 |
torch
|
2 |
spaces
|
3 |
git+https://github.com/huggingface/parler-tts.git
|
4 |
+
nltk
|