You can deploy this model and start earning money today!

ZhengyanWan/
dFlowGRPO-ScienceQA

Name: dFlowGRPO-ScienceQA
Availability: InStock

Views

No views yet

ZhengyanWan/dFlowGRPO-ScienceQA

LoRA adapter for FUDOKI, fine-tuned with dFlowGRPO on the ScienceQA reward. Training step: 1500.

See WanZhengyan/dFlowGRPO for training & evaluation code.

bash

huggingface-cli download ZhengyanWan/dFlowGRPO-ScienceQA \
    --local-dir output_grpo_u_scienceqa_kl0.01/checkpoint_0_step1500

Then run the matching evaluator from discrete_flow_grpo/.