#finetune

Custom LogitsProcessor for HuggingFace Transformers: Fixing JSON Repetition Hallucinations in Qwen

Step-by-step guide to building a custom LogitsProcessor in HuggingFace Transformers. Applies no-repeat-ngram control selectively inside JSON Content fields to suppress repetition hallucinations during fine-tuned Qwen inference.

How I Finetuned IBM Granite Speech 1B on Japanese Audio and Improved CER from 0.37 to 0.14

I finetuned IBM Granite Speech (`granite-4.0-1b-speech`) on 100 hours of Japanese speech data and reduced CER from 0.37 to 0.14. The official script's Projector+LoRA-only training has a ceiling on accuracy gains. The key breakthrough was additionally training `lm_head` and the last 8 layers of the Language Model. The result matches Qwen3-ASR-1.7B (CER 0.14) with only 1B parameters.