# Google Colab Ready Quick Inference
# 1. Install dependencies:
!pip install huggingface_hub omnilingual-asr torchaudio

# 2. Download the model checkpoint directly from Hugging Face:
import torch
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="andrewbawitlung/omni-asr-ctc-300m-v2-mizonal3-E2-lus-v2026.06",
    filename="sdp_00.pt"
)

# 3. Load the base model pipeline and inject fine-tuned weights:
from omnilingual_asr.models.inference.pipeline import ASRInferencePipeline

print("Loading base model pipeline...")
pipeline = ASRInferencePipeline(model_card="omniASR_CTC_300M_v2")

print("Loading fine-tuned checkpoint weights...")
checkpoint = torch.load(model_path, map_location="cpu")
state_dict = checkpoint['model'] if 'model' in checkpoint else checkpoint
pipeline.model.load_state_dict(state_dict, strict=False)

# 4. Run transcription (make sure audio files are present in the specified paths):
audio_files = ["/content/segment_003.wav", "/content/segment_108.wav"]
lang = ["lus_Latn", "lus_Latn"]
transcriptions = pipeline.transcribe(audio_files, lang=lang, batch_size=2)

print("Transcription 0:", transcriptions[0])
print("Transcription 1:", transcriptions[1])

Model description

Experiment Configurations

This repository is part of a series of experiments. The different configurations are:

E1 (Baseline): Standard training configuration.
E2 (Noise): Training with background noise augmentation.
E3 (Speed): Training with speed perturbation augmentation.
E4 (SpecAug): Training with SpecAugment (time and frequency masking).
E5 (Combined): Training with a combination of all augmentations.

All Models in this Family

Experiment	Hugging Face Repository
E1 (Baseline)	andrewbawitlung/omni-asr-ctc-300m-v2-mizonal3-E1-lus-v2026.06
E2 (Noise)	andrewbawitlung/omni-asr-ctc-300m-v2-mizonal3-E2-lus-v2026.06
E3 (Speed)	andrewbawitlung/omni-asr-ctc-300m-v2-mizonal3-E3-lus-v2026.06
E4 (SpecAug)	andrewbawitlung/omni-asr-ctc-300m-v2-mizonal3-E4-lus-v2026.06
E5 (Combined)	andrewbawitlung/omni-asr-ctc-300m-v2-mizonal3-E5-lus-v2026.06

Training hyperparameters

Training hyperparameters not available in bin format.

Training results

step	epoch	train_loss	eval_loss	eval_wer	eval_cer	grad_norm
100	4.0000	73.2138				312.7692
200	8.0000	44.8471				111.1043
300	12.0000	33.1846				96.7360
400	15.0000	27.3954				83.3923
500	19.0000	22.3309	14.6167	24.7654	5.0040	81.9629
600	23.0000	18.7707				75.9650
700	26.0000	16.4997				65.1045
800	30.0000	14.4537				66.4703
900	34.0000	12.5522				46.2519
1000	38.0000	10.9942	12.1225	18.9015	3.7322	48.9006
1100	41.0000	10.3972				50.6749
1200	45.0000	8.9089				45.9600
1300	49.0000	7.6246				37.5720
1400	52.0000	6.9561				42.0430
1500	56.0000	6.2497	15.0347	16.9664	3.2585	34.8496
1600	60.0000	5.7502				30.8823
1700	63.0000	5.0200				33.8705
1800	67.0000	4.2765				25.7846
1900	71.0000	4.0509				31.5831
2000	75.0000	4.0045	16.1042	16.8686	3.1840	27.1806
2100	78.0000	3.6620				29.1253
2200	82.0000	2.6542				22.9630
2300	86.0000	2.5402				20.2393
2400	89.0000	2.5665				22.1200
2500	93.0000	2.1627	19.5373	15.7154	3.0013	19.3820
2600	97.0000	2.1236				31.4884
2700	100.0000	1.9041				18.4045
2800	104.0000	2.1353				18.8731
2900	108.0000	1.8568				17.3822
3000	112.0000	1.6859	20.4668	15.7056	3.0173	18.1024
3100	115.0000	1.5612				16.1802
3200	119.0000	1.7075				18.1627
3300	123.0000	1.3753				15.0050
3400	126.0000	1.5098				16.3814
3500	130.0000	1.3662	21.3131	15.5688	2.9765	14.8677
3600	134.0000	1.4248				14.7347
3700	138.0000	1.2052				14.7049
3800	141.0000	1.5432				19.3035
3900	145.0000	1.4243				17.0426
4000	149.0000	1.4820	21.7087	15.5786	2.9712	16.0188