Full Parameter Finetuning 0.5B 8192 context length Qwen 2 on Malaysian text 12.86B tokens

Continue pretraining https://huggingface.co/Qwen/Qwen2-0.5B on 12.86B tokens Malaysian dataset using 8192 context length.

WanDB at https://wandb.ai/huseinzol05/Qwen2-0.5B-fpf/workspace?nw=nwuserhuseinzol05

Downloads last month
68
Safetensors
Model size
494M params
Tensor type
BF16
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Collection including mesolitica/Qwen2-0.5B-8192-fpf