Spaces:
Running
on
Zero
Running
on
Zero
Initial commit app.py
Browse files
app.py
ADDED
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from dataclasses import dataclass, field
|
2 |
+
import logging
|
3 |
+
|
4 |
+
import sys
|
5 |
+
sys.path.append("/home/user/app/src/sonicverse")
|
6 |
+
|
7 |
+
from huggingface_hub import login
|
8 |
+
import os
|
9 |
+
|
10 |
+
hf_token = os.getenv("HUGGINGFACE_HUB_TOKEN")
|
11 |
+
if not hf_token:
|
12 |
+
raise ValueError("Missing HUGGINGFACE_HUB_TOKEN. Set it as a secret in your Space.")
|
13 |
+
|
14 |
+
login(token=hf_token)
|
15 |
+
|
16 |
+
import gradio as gr
|
17 |
+
import torch
|
18 |
+
import transformers
|
19 |
+
import torchaudio
|
20 |
+
|
21 |
+
from multi_token.model_utils import MultiTaskType
|
22 |
+
from multi_token.training import ModelArguments
|
23 |
+
from multi_token.inference import load_trained_lora_model
|
24 |
+
from multi_token.data_tools import encode_chat
|
25 |
+
|
26 |
+
|
27 |
+
@dataclass
|
28 |
+
class ServeArguments(ModelArguments):
|
29 |
+
load_bits: int = field(default=16)
|
30 |
+
max_new_tokens: int = field(default=128)
|
31 |
+
temperature: float = field(default=0.01)
|
32 |
+
|
33 |
+
|
34 |
+
# Load arguments and model
|
35 |
+
logging.getLogger().setLevel(logging.INFO)
|
36 |
+
|
37 |
+
parser = transformers.HfArgumentParser((ServeArguments,))
|
38 |
+
serve_args, _ = parser.parse_args_into_dataclasses(return_remaining_strings=True)
|
39 |
+
|
40 |
+
model, tokenizer = load_trained_lora_model(
|
41 |
+
model_name_or_path=serve_args.model_name_or_path,
|
42 |
+
model_lora_path=serve_args.model_lora_path,
|
43 |
+
load_bits=serve_args.load_bits,
|
44 |
+
use_multi_task=MultiTaskType(serve_args.use_multi_task),
|
45 |
+
tasks_config=serve_args.tasks_config
|
46 |
+
)
|
47 |
+
|
48 |
+
|
49 |
+
def generate_caption(audio_file):
|
50 |
+
# waveform, sample_rate = torchaudio.load(audio_file)
|
51 |
+
|
52 |
+
req_json = {
|
53 |
+
"messages": [
|
54 |
+
{"role": "user", "content": "Describe the music. <sound>"}
|
55 |
+
],
|
56 |
+
"sounds": [audio_file]
|
57 |
+
}
|
58 |
+
|
59 |
+
encoded_dict = encode_chat(req_json, tokenizer, model.modalities)
|
60 |
+
|
61 |
+
with torch.inference_mode():
|
62 |
+
output_ids = model.generate(
|
63 |
+
input_ids=encoded_dict["input_ids"].unsqueeze(0).to(model.device),
|
64 |
+
max_new_tokens=serve_args.max_new_tokens,
|
65 |
+
use_cache=True,
|
66 |
+
do_sample=True,
|
67 |
+
temperature=serve_args.temperature,
|
68 |
+
modality_inputs={
|
69 |
+
m.name: [encoded_dict[m.name]] for m in model.modalities
|
70 |
+
},
|
71 |
+
)
|
72 |
+
|
73 |
+
outputs = tokenizer.decode(
|
74 |
+
output_ids[0, encoded_dict["input_ids"].shape[0]:],
|
75 |
+
skip_special_tokens=True
|
76 |
+
).strip()
|
77 |
+
|
78 |
+
return outputs
|
79 |
+
|
80 |
+
|
81 |
+
demo = gr.Interface(
|
82 |
+
fn=generate_caption,
|
83 |
+
inputs=gr.Audio(type="filepath", label="Upload an audio file"),
|
84 |
+
outputs=gr.Textbox(label="Generated Caption"),
|
85 |
+
title="SonicVerse",
|
86 |
+
description="Upload an audio file to generate a caption using SonicVerse"
|
87 |
+
)
|
88 |
+
|
89 |
+
if __name__ == "__main__":
|
90 |
+
demo.launch()
|