Upload app.py
Browse files
app.py
CHANGED
@@ -2,7 +2,7 @@ from transformers import VisionEncoderDecoderModel, ViTFeatureExtractor, AutoTok
|
|
2 |
import torch
|
3 |
from PIL import Image
|
4 |
|
5 |
-
import gradio as gr
|
6 |
|
7 |
from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
|
8 |
from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
|
@@ -12,13 +12,16 @@ feature_extractor = ViTFeatureExtractor.from_pretrained("nlpconnect/vit-gpt2-ima
|
|
12 |
tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
|
13 |
|
14 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
15 |
-
model.to(device)
|
16 |
|
17 |
models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
|
18 |
"facebook/fastspeech2-en-ljspeech",
|
19 |
-
arg_overrides={"vocoder": "hifigan", "fp16":
|
20 |
)
|
|
|
21 |
model1 = models[0]
|
|
|
|
|
22 |
TTSHubInterface.update_cfg_with_data_cfg(cfg, task.data_cfg)
|
23 |
generator = task.build_generator(models, cfg)
|
24 |
|
@@ -27,32 +30,27 @@ num_beams = 4
|
|
27 |
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
|
28 |
|
29 |
|
30 |
-
def
|
31 |
-
images = []
|
32 |
-
text = ""
|
33 |
|
34 |
-
for image_path in image_paths:
|
35 |
-
|
36 |
-
|
37 |
i_image = i_image.convert(mode="RGB")
|
38 |
-
print(image_path)
|
39 |
|
40 |
-
|
41 |
-
print(images)
|
42 |
-
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
|
43 |
pixel_values = pixel_values.to(device)
|
44 |
|
45 |
output_ids = model.generate(pixel_values, **gen_kwargs)
|
46 |
|
47 |
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
|
48 |
preds = [pred.strip() for pred in preds]
|
49 |
-
|
50 |
-
|
51 |
-
sample = TTSHubInterface.get_model_input(task,
|
|
|
52 |
wav, rate = TTSHubInterface.get_prediction(task, model1, generator, sample)
|
53 |
-
return wav
|
54 |
-
#return ipd.Audio(wav, rate=rate)
|
55 |
|
56 |
|
57 |
-
interface = gr.Interface(
|
58 |
interface.launch()
|
|
|
2 |
import torch
|
3 |
from PIL import Image
|
4 |
|
5 |
+
import gradio as gr
|
6 |
|
7 |
from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
|
8 |
from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
|
|
|
12 |
tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
|
13 |
|
14 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
15 |
+
model = model.to(device)
|
16 |
|
17 |
models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
|
18 |
"facebook/fastspeech2-en-ljspeech",
|
19 |
+
arg_overrides={"vocoder": "hifigan", "fp16": True}
|
20 |
)
|
21 |
+
|
22 |
model1 = models[0]
|
23 |
+
model1 = model1.to(device)
|
24 |
+
|
25 |
TTSHubInterface.update_cfg_with_data_cfg(cfg, task.data_cfg)
|
26 |
generator = task.build_generator(models, cfg)
|
27 |
|
|
|
30 |
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
|
31 |
|
32 |
|
33 |
+
def inference(image_paths):
|
|
|
|
|
34 |
|
35 |
+
#for image_path in image_paths:
|
36 |
+
i_image = Image.fromarray(image_paths)
|
37 |
+
if i_image.mode != "RGB":
|
38 |
i_image = i_image.convert(mode="RGB")
|
|
|
39 |
|
40 |
+
pixel_values = feature_extractor(images=i_image, return_tensors="pt").pixel_values
|
|
|
|
|
41 |
pixel_values = pixel_values.to(device)
|
42 |
|
43 |
output_ids = model.generate(pixel_values, **gen_kwargs)
|
44 |
|
45 |
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
|
46 |
preds = [pred.strip() for pred in preds]
|
47 |
+
|
48 |
+
|
49 |
+
sample = TTSHubInterface.get_model_input(task, preds)
|
50 |
+
|
51 |
wav, rate = TTSHubInterface.get_prediction(task, model1, generator, sample)
|
52 |
+
return wav
|
|
|
53 |
|
54 |
|
55 |
+
interface = gr.Interface(inference, gr.Image(), "audio")
|
56 |
interface.launch()
|