File size: 4,921 Bytes
22379f7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
from typing import Optional

import gradio as gr
import spaces
import torch
from dotenv import load_dotenv
from huggingface_hub import login
from peft import PeftModel
from transformers import WhisperForConditionalGeneration, WhisperProcessor, pipeline
from transformers.pipelines.base import Pipeline

load_dotenv()


def ensure_hf_login() -> None:
    token = os.getenv("HF_TOKEN")
    if not token:
        print("HF_TOKEN not set; skipping Hugging Face login.")
        return

    try:
        login(token=token)
    except Exception as exc:
        print(f"Failed to login to Hugging Face Hub: {exc}")


ensure_hf_login()


LANGUAGE = "Arabic"
BATCH_SIZE = 1
DEVICE = 0 if torch.cuda.is_available() else "cpu"
BASE_MODEL_PATH = "openai/whisper-large-v3-turbo"
LORA_PATH = "anaszil/whisper-large-v3-turbo-darija"
PIPELINE: Optional[Pipeline] = None


@spaces.GPU
def _build_pipeline() -> Pipeline:
    torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

    base_model = WhisperForConditionalGeneration.from_pretrained(
        BASE_MODEL_PATH,
        torch_dtype=torch_dtype,
    )
    model = PeftModel.from_pretrained(base_model, LORA_PATH)
    processor = WhisperProcessor.from_pretrained(
        BASE_MODEL_PATH,
        language=LANGUAGE,
        task="transcribe",
    )

    model.generation_config.language = LANGUAGE
    model.generation_config.task = "transcribe"
    model.generation_config.forced_decoder_ids = None
    model.eval()

    return pipeline(
        task="automatic-speech-recognition",
        model=model,
        tokenizer=processor.tokenizer,
        feature_extractor=processor.feature_extractor,
        chunk_length_s=30,
        device=DEVICE,
    )


def get_pipeline() -> Pipeline:
    global PIPELINE

    if PIPELINE is None:
        print("Loading Darija LoRA model...")
        PIPELINE = _build_pipeline()

    return PIPELINE


def format_timestamp(
    seconds: Optional[float],
    always_include_hours: bool = False,
    decimal_marker: str = ".",
) -> Optional[str]:
    if seconds is None:
        return seconds

    milliseconds = round(seconds * 1000.0)

    hours = milliseconds // 3_600_000
    milliseconds -= hours * 3_600_000

    minutes = milliseconds // 60_000
    milliseconds -= minutes * 60_000

    whole_seconds = milliseconds // 1_000
    milliseconds -= whole_seconds * 1_000

    hours_marker = f"{hours:02d}:" if always_include_hours or hours > 0 else ""

    return (
        f"{hours_marker}{minutes:02d}:{whole_seconds:02d}"
        f"{decimal_marker}{milliseconds:03d}"
    )


@spaces.GPU
def _run_transcription(audio_input, return_timestamps: bool) -> str:
    asr_pipeline = get_pipeline()

    outputs = asr_pipeline(
        audio_input,
        batch_size=BATCH_SIZE,
        generate_kwargs={"task": "transcribe", "language": LANGUAGE},
        return_timestamps=return_timestamps,
    )

    text = outputs["text"]
    if return_timestamps:
        chunks = outputs.get("chunks") or []
        text = "\n".join(
            f"[{format_timestamp(chunk['timestamp'][0])} -> "
            f"{format_timestamp(chunk['timestamp'][1])}] {chunk['text']}"
            for chunk in chunks
        )

    return text


def transcribe(audio_input, return_timestamps: bool):
    if audio_input is None:
        return "Please provide audio input either via microphone or file upload."

    return _run_transcription(audio_input, return_timestamps)


def process_audio(audio_input, timestamps):
    if audio_input is None:
        return "Please provide audio input.", "No audio input detected."

    try:
        transcription = transcribe(audio_input, timestamps)
        return transcription, "Transcription completed with the Darija LoRA model."
    except Exception as exc:
        return f"Error: {exc}", f"Transcription failed: {exc}"


with gr.Blocks(title="Darija Speech Transcription") as demo:
    gr.Markdown("# Darija Speech Transcription Demo")
    gr.Markdown("Transcribe Darija audio with the fine-tuned Whisper LoRA model.")

    with gr.Row():
        with gr.Column(scale=1):
            timestamps_checkbox = gr.Checkbox(
                label="Return timestamps",
                value=False,
            )

            audio_component = gr.Audio(
                sources=["microphone", "upload"],
                type="filepath",
                label="Record or Upload Audio",
            )

            transcribe_button = gr.Button("Transcribe", variant="primary")

        with gr.Column(scale=1):
            output_text = gr.Textbox(
                label="Transcription Output",
                lines=10,
                show_copy_button=True,
            )

            status_message = gr.Markdown("")

    transcribe_button.click(
        fn=process_audio,
        inputs=[audio_component, timestamps_checkbox],
        outputs=[output_text, status_message],
    )


demo.launch()