Initial commit

.gitignore

@@ -0,0 +1,165 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+.idea
+ASR.pth
+Test Arabic.mp3
+test_main.http
+# C extensions
+*.so
+*.pth
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/

Inference.py

@@ -0,0 +1,58 @@
+import numpy as np
+from utils.Audio_Processing import preprocess_audio
+from utils.Constants import *
+from utils.MMS import get_device, MMS, greedyDecoder
+from utils.NLP import preprocess_vocab
+import torch
+
+############################################################################################
+
+
+model_path = "./ASR_2_1_220.pth"
+
+
+############################################################################################
+
+
+def predict(audio_file):
+    device = get_device()
+
+    processed_audios = []
+    mel_spec, duration = preprocess_audio(audio_file)
+    processed_audios.append(mel_spec)
+    padded_audios = [
+        (
+            mel_spec.shape[-1],
+            np.pad(
+                mel_spec,
+                ((0, 0), (0, N_FRAMES - mel_spec.shape[-1])),
+                mode="constant",
+            ),
+        )
+        for mel_spec in processed_audios
+    ]
+
+    char2idx, idx2char, vocab_size = preprocess_vocab()
+
+    # load model
+
+    model = MMS(
+        vocab_size=vocab_size,
+        max_encoder_seq_len=math.ceil(N_FRAMES / 2),
+        max_decoder_seq_len=MAX_SEQ_LEN,
+        num_encoder_layers=2,
+        num_decoder_layers=1,
+        d_model=512,
+        nhead=8,
+        dim_feedforward=2048,
+    )
+
+    model.load_state_dict(torch.load(model_path, weights_only=True))
+    model.to(device)
+    model.eval()
+
+    result = greedyDecoder(
+        model, padded_audios[0][1], padded_audios[0][0], char2idx, idx2char, device
+    )
+
+    return result

main.py

@@ -0,0 +1,97 @@
+import sys
+import io
+
+from fastapi import (
+    FastAPI,
+    File,
+    UploadFile,
+)
+from openai import OpenAI
+import dotenv
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel
+
+from Inference import predict
+
+import tempfile
+import os
+from utils.Translation import get_translate
+
+dotenv.load_dotenv()
+app = FastAPI()
+client = OpenAI()
+
+sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8")
+
+# Add CORS middleware
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],  # Allow all origins
+    allow_credentials=True,
+    allow_methods=["*"],  # Allow all methods (GET, POST, etc.)
+    allow_headers=["*"],  # Allow all headers
+)
+
+
+@app.get("/")
+async def root():
+    return {"message": "Hello World"}
+
+
+@app.post("/translate-any")
+async def translate(txt: str, language: str):
+    completion = client.chat.completions.create(
+        model="gpt-4o-mini",
+        messages=[
+            {
+                "role": "system",
+                "content": f"Translate this {txt} into {language} directly without any other words",
+            }
+        ],
+    )
+    print(completion.choices[0].message.content)
+    return {"response": completion.choices[0].message.content}
+
+
+class TranslationRequest(BaseModel):
+    text: str
+
+
+@app.post("/translate/auto")
+async def translate(request: TranslationRequest):
+    response = get_translate(request.text)
+    return {"translation": response}
+
+
+@app.post("/translate/en")
+async def translate(request: TranslationRequest):
+    # response = get_translate(request.text)
+    return {"translation": "ليس بعد"}
+
+
+@app.post("/audio2text")
+async def upload_audio(file: UploadFile = File(...)):
+    # Read the uploaded audio file into memory
+    contents = await file.read()
+
+    # Get the current working directory
+    current_dir = os.getcwd()
+    print(current_dir, flush=True)
+
+    # Create a temporary file in the current working directory
+    with tempfile.NamedTemporaryFile(
+            dir=current_dir, delete=False, suffix=".wav"
+    ) as tmp_file:
+        tmp_file.write(contents)
+        tmp_file_path = tmp_file.name  # Get the path of the temp file
+
+    try:
+        # Pass the path of the saved file to the predict function
+        print(f"Temporary file created at: {tmp_file_path}", flush=True)
+        result = predict(tmp_file_path)
+    finally:
+        # Clean up the temporary file after prediction
+        os.remove(tmp_file_path)
+        print(f"Temporary file deleted: {tmp_file_path}", flush=True)
+
+    return {"text": result}

requirements.txt

@@ -0,0 +1,10 @@
+fastapi~=0.115.0
+openai~=1.50.2
+python-dotenv~=1.0.1
+numpy~=2.0.2
+torch~=2.4.1
+uvicorn~=0.31.0
+python-multipart~=0.0.12
+pydantic~=2.8.2
+librosa~=0.10.2.post1
+requests~=2.32.3

utils/Audio_Processing.py

@@ -0,0 +1,38 @@
+import librosa
+import numpy as np
+from utils.Constants import *
+
+
+def pad_or_trim(array, length=N_SAMPLES, axis=-1, padding=True):
+    if array.shape[axis] > length:
+        array = array.take(indices=range(length), axis=axis)
+
+    if padding & (array.shape[axis] < length):
+        pad_widths = [(0, 0)] * array.ndim
+        pad_widths[axis] = (0, length - array.shape[axis])
+        array = np.pad(array, pad_widths)
+
+    return array
+
+
+# Function to load and preprocess audio
+def preprocess_audio(file_path):
+    audio_data, _ = librosa.load(file_path, sr=SAMPLE_RATE)
+
+    duration = librosa.get_duration(y=audio_data, sr=SAMPLE_RATE)
+
+    modified_audio = pad_or_trim(audio_data, padding=False)
+
+    sgram = librosa.stft(y=modified_audio, n_fft=N_FFT, hop_length=HOP_LENGTH)
+
+    sgram_mag, _ = librosa.magphase(sgram)
+
+    mel_scale_sgram = librosa.feature.melspectrogram(
+        S=sgram_mag, sr=SAMPLE_RATE, n_fft=N_FFT, hop_length=HOP_LENGTH, n_mels=N_MELS
+    )
+
+    mel_sgram = librosa.amplitude_to_db(mel_scale_sgram, ref=np.min)
+
+    del audio_data, modified_audio, sgram, mel_scale_sgram
+
+    return mel_sgram, duration

utils/Constants.py

@@ -0,0 +1,22 @@
+import math
+
+N_ROWS = 50000  # NUMBER OF ROWS TAKEN FROM THE DATA
+MAX_TEXT_LEN = 70
+MAX_SEQ_LEN = 70
+N_MELS = 128
+SAMPLE_RATE = 16000  # NUMBER OF SAMPLES PER SECOND
+HOP_LENGTH = (
+    512  # THE STEP LENGTH OF THE SLIDING WINDOW, Commonly set to one-fourth of N_FFT
+)
+N_FFT = 2048  # NUMBER OF FFT POINTS (WINDOW SIZE) THIS CONTROLS THE RESOLUTION OF THE FREQUENCY DOMAIN ANALYSIS
+CHUNK_LENGTH = 15  # 15 SECOND CHUNK
+N_SAMPLES = SAMPLE_RATE * CHUNK_LENGTH
+N_FRAMES = math.ceil(N_SAMPLES / HOP_LENGTH)
+N_SAMPLES_PER_TOKEN = 2 * HOP_LENGTH
+FRAMES_PER_SECOND = SAMPLE_RATE // HOP_LENGTH  # OR N_FRAMES // CHUNK_LENGTH
+TOKENS_PER_SECOND = SAMPLE_RATE // N_SAMPLES_PER_TOKEN
+NEG_INFTY = -1e9
+special_tokens = ["<PAD>", "<UNK>", "<SOS>", "<EOS>"]
+
+
+############################################################################################