Dictionary size
#4
by SeniorGrandMaster - opened
Thank you for this model and your efforts. I want learn, why you didn't expanded tokenizer.json ? 150+ language but same dictionay size with orginal model. Very interesting decision.
Good question! Since we continued pretraining from the Qwen3.5 base model, we kept its original vocabulary to avoid introducing unforeseen issues. We plan to explore expanding the vocabulary for low-resource languages in future versions.