Fine-tuned omniASRLLM1Bv2 for the Google WAXAL ASR Challenge on Lingala (linLatn), Luganda (lugLatn), and Shona (snaLatn). See runconfig.json for the exact reproducible configuration and seed.
Modellquelle
Quellenbeschreibung
Fine-tuned omniASR_LLM_1B_v2 for the Google WAXAL ASR Challenge on Lingala (lin_Latn), Luganda (lug_Latn), and Shona (sna_Latn). See run_config.json for the exact reproducible configuration and seed.
Quellen
1 QuelleVerifiziert 30. Aug.
Modellartefakte
2 Artefaktews_1.2a8bfda1/checkpoints/step_1000/data_reader/dp_00.pt
pt · 1,52 MB · SHA-256 193a832f75ba…3f06 · Hugging Face
Herunterladenws_1.2a8bfda1/checkpoints/step_1000/model/pp_00/tp_00/sdp_00.pt
pt · 8,49 GB · SHA-256 ccb86d045153…c5e3 · Hugging Face
HerunterladenQuellenauszüge
2 Auszüge--- library_name: fairseq2 license: apache-2.0 language: - ln - lg - sn tags: - automatic-speech-recognition - omnilingual-asr --- # waxal_llm_1b_v2_200step Fine-tuned `omniASR_LLM_1B_v2` for the Google WAXAL ASR Challenge on Lingala (`lin_Latn`), Luganda (`lug_Latn`), and Shona (`sna_Latn`). See `run_config.json` for the exact reproducible configuration and seed.
Source context: 0 downloads · 1 likes · Pipeline automatic-speech-recognition · Library fairseq2 · Repo mlai-dante/waxal-omniASR-LLM-1B-v2