Instructions to use HuggingAnalist/mms-1b-asr-lin with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use HuggingAnalist/mms-1b-asr-lin with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="HuggingAnalist/mms-1b-asr-lin")# Load model directly from transformers import AutoProcessor, AutoModelForCTC processor = AutoProcessor.from_pretrained("HuggingAnalist/mms-1b-asr-lin") model = AutoModelForCTC.from_pretrained("HuggingAnalist/mms-1b-asr-lin", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: lin
|
| 3 |
+
library_name: transformers
|
| 4 |
+
pipeline_tag: automatic-speech-recognition
|
| 5 |
+
tags:
|
| 6 |
+
- mms
|
| 7 |
+
- wav2vec2
|
| 8 |
+
- ctc
|
| 9 |
+
- asr
|
| 10 |
+
- waxal
|
| 11 |
+
- lingala
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# MMS-300M — Lingala ASR
|
| 15 |
+
|
| 16 |
+
`facebook/mms-300m` fine-tuned (character CTC head) on WaxalNLP `lin_asr` for the Waxal ASR challenge. Trained in bf16/fp32 with `ctc_zero_infinity=True`.
|
| 17 |
+
|
| 18 |
+
`language_model/lm_lin.arpa` is a KenLM 4-gram built from the training transcriptions, for beam-search decoding with pyctcdecode (tune alpha/beta on validation).
|