Spaces:
Runtime error
Runtime error
File size: 6,253 Bytes
96e9536 |
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 |
# Copyright 2023 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
import numpy as np
from transformers import (
MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING,
AutoProcessor,
TextToAudioPipeline,
pipeline,
)
from transformers.testing_utils import (
is_pipeline_test,
require_torch,
require_torch_gpu,
require_torch_or_tf,
slow,
)
from .test_pipelines_common import ANY
@is_pipeline_test
@require_torch_or_tf
class TextToAudioPipelineTests(unittest.TestCase):
model_mapping = MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING
# for now only text_to_waveform and not text_to_spectrogram
@slow
@require_torch
def test_small_model_pt(self):
speech_generator = pipeline(task="text-to-audio", model="facebook/musicgen-small", framework="pt")
forward_params = {
"do_sample": False,
"max_new_tokens": 250,
}
outputs = speech_generator("This is a test", forward_params=forward_params)
# musicgen sampling_rate is not straightforward to get
self.assertIsNone(outputs["sampling_rate"])
audio = outputs["audio"]
self.assertEqual(ANY(np.ndarray), audio)
# test two examples side-by-side
outputs = speech_generator(["This is a test", "This is a second test"], forward_params=forward_params)
audio = [output["audio"] for output in outputs]
self.assertEqual([ANY(np.ndarray), ANY(np.ndarray)], audio)
# test batching
outputs = speech_generator(
["This is a test", "This is a second test"], forward_params=forward_params, batch_size=2
)
self.assertEqual(ANY(np.ndarray), outputs[0]["audio"])
@slow
@require_torch
def test_large_model_pt(self):
speech_generator = pipeline(task="text-to-audio", model="suno/bark-small", framework="pt")
# test text-to-speech
forward_params = {
# Using `do_sample=False` to force deterministic output
"do_sample": False,
"semantic_max_new_tokens": 100,
}
outputs = speech_generator("This is a test", forward_params=forward_params)
self.assertEqual(
{"audio": ANY(np.ndarray), "sampling_rate": 24000},
outputs,
)
# test two examples side-by-side
outputs = speech_generator(
["This is a test", "This is a second test"],
forward_params=forward_params,
)
audio = [output["audio"] for output in outputs]
self.assertEqual([ANY(np.ndarray), ANY(np.ndarray)], audio)
# test other generation strategy
forward_params = {
"do_sample": True,
"semantic_max_new_tokens": 100,
"semantic_num_return_sequences": 2,
}
outputs = speech_generator("This is a test", forward_params=forward_params)
audio = outputs["audio"]
self.assertEqual(ANY(np.ndarray), audio)
# test using a speaker embedding
processor = AutoProcessor.from_pretrained("suno/bark-small")
temp_inp = processor("hey, how are you?", voice_preset="v2/en_speaker_5")
history_prompt = temp_inp["history_prompt"]
forward_params["history_prompt"] = history_prompt
outputs = speech_generator(
["This is a test", "This is a second test"],
forward_params=forward_params,
batch_size=2,
)
audio = [output["audio"] for output in outputs]
self.assertEqual([ANY(np.ndarray), ANY(np.ndarray)], audio)
@slow
@require_torch_gpu
def test_conversion_additional_tensor(self):
speech_generator = pipeline(task="text-to-audio", model="suno/bark-small", framework="pt", device=0)
processor = AutoProcessor.from_pretrained("suno/bark-small")
forward_params = {
"do_sample": True,
"semantic_max_new_tokens": 100,
}
# atm, must do to stay coherent with BarkProcessor
preprocess_params = {
"max_length": 256,
"add_special_tokens": False,
"return_attention_mask": True,
"return_token_type_ids": False,
"padding": "max_length",
}
outputs = speech_generator(
"This is a test",
forward_params=forward_params,
preprocess_params=preprocess_params,
)
temp_inp = processor("hey, how are you?", voice_preset="v2/en_speaker_5")
history_prompt = temp_inp["history_prompt"]
forward_params["history_prompt"] = history_prompt
# history_prompt is a torch.Tensor passed as a forward_param
# if generation is successfull, it means that it was passed to the right device
outputs = speech_generator(
"This is a test", forward_params=forward_params, preprocess_params=preprocess_params
)
self.assertEqual(
{"audio": ANY(np.ndarray), "sampling_rate": 24000},
outputs,
)
def get_test_pipeline(self, model, tokenizer, processor):
speech_generator = TextToAudioPipeline(model=model, tokenizer=tokenizer)
return speech_generator, ["This is a test", "Another test"]
def run_pipeline_test(self, speech_generator, _):
outputs = speech_generator("This is a test")
self.assertEqual(ANY(np.ndarray), outputs["audio"])
forward_params = {"num_return_sequences": 2, "do_sample": True}
outputs = speech_generator(["This is great !", "Something else"], forward_params=forward_params)
audio = [output["audio"] for output in outputs]
self.assertEqual([ANY(np.ndarray), ANY(np.ndarray)], audio)
|