Instructions to use trysem/patched-conformer-ml with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- NeMo
How to use trysem/patched-conformer-ml with NeMo:
# tag did not correspond to a valid NeMo domain.
- Notebooks
- Google Colab
- Kaggle
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
# Patched Malayalam Conformer-ML
|
| 3 |
+
|
| 4 |
+
This is a patched version of `trysem/conformer-ml` modified for compatibility with NeMo ONNX export.
|
| 5 |
+
|
| 6 |
+
### Changes Made:
|
| 7 |
+
1. **Experimental Keys Removed**: `multisoftmax`, `multilingual`, and `language_keys` were removed from config to prevent instantiation errors.
|
| 8 |
+
2. **Tokenizer Paths Fixed**: Tokenizer `model_path` and `vocab_path` were converted to relative `nemo:` paths.
|
| 9 |
+
3. **Vocab Placeholder**: Ensured `vocab.txt` exists within the archive.
|
| 10 |
+
4. **BPE Enforced**: Tokenizer type explicitly set to `bpe`.
|
| 11 |
+
5. **Structure**: Re-bundled for `restore_from` compatibility.
|