Modelpublic

Saraswathy/vlm-mix-resume-capmix-perception50-visual50-step95

sourceHugging Faceupdated 14d agoView on Hugging Face

0likes

Model Card

Perception 50 / visual reasoning 50: step-95 resume archive

Public archive of a complete EasyR1 training-resume checkpoint at global step

1.It includes FSDP model and optimizer state, extra state, dataloader state, and the evaluation-ready LoRA adapter under actor/lora_adapter/. It is not a merged standalone model; load the adapter on Qwen/Qwen3-VL-4B-Instruct. Verify files using SHA256SUMS.json.