Loading README.md +2 −4 Original line number Diff line number Diff line Loading @@ -169,14 +169,12 @@ Each run takes ~4 hours to reach 150 steps on 4 H100s. You can decrease testing # Project Roadmap ## Short-term Goals - 🗓️ Mar 25, 2025: We release VAGEN, a multi-turn reinforcement learning framework for training VLM Agents! - [ ] Merge to RAGEN for better package mangement - [ ] Address training stability issues - [ ] Lora Support - [ ] Implement improved early stopping and gradient clip strategies - [ ] Reward normalization (see RAGEN) - [ ] Implement improved early stopping and gradient clipping strategies - [ ] Reward normalization (refer to RAGEN) - [ ] Expand evaluation framework to more diverse visual environments - [ ] Scaling to larger models and applying TRICO to text-only tasks Loading Loading
README.md +2 −4 Original line number Diff line number Diff line Loading @@ -169,14 +169,12 @@ Each run takes ~4 hours to reach 150 steps on 4 H100s. You can decrease testing # Project Roadmap ## Short-term Goals - 🗓️ Mar 25, 2025: We release VAGEN, a multi-turn reinforcement learning framework for training VLM Agents! - [ ] Merge to RAGEN for better package mangement - [ ] Address training stability issues - [ ] Lora Support - [ ] Implement improved early stopping and gradient clip strategies - [ ] Reward normalization (see RAGEN) - [ ] Implement improved early stopping and gradient clipping strategies - [ ] Reward normalization (refer to RAGEN) - [ ] Expand evaluation framework to more diverse visual environments - [ ] Scaling to larger models and applying TRICO to text-only tasks Loading
