Search results

smollm3 DPO recipe fails when using liger_kernal #227

🏋 DPORelated to DPORelated to DPO

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#4544 In huggingface/trl;

· shamanez opened on Nov 19, 2025

[OpenEnv] multi step training issue in GRPO trainer

🏋 GRPORelated to GRPORelated to GRPO

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#4543 In huggingface/trl;

· akshaygunokia opened on Nov 18, 2025

Is attn_implementation=sdpa not supported when using SFTTrainer with mllama?

Status: Open.

#4541 In huggingface/trl;

· osaidr opened on Nov 18, 2025

Significant Performance Difference between DeepSpeed's <code>zero_stage=1</code> and <code>zero_stage=2</code>

Status: Open.

#4540 In huggingface/trl;

· liyc-ai opened on Nov 18, 2025

TCPStore Issues in GRPOTrainer+vLLM

🏋 GRPORelated to GRPORelated to GRPO

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#4538 In huggingface/trl;

· DU1999 opened on Nov 18, 2025

GRPOTrainer fails to train when using Accelerate with DeepSpeed ZeRO-3.

⚡accelerateRelated to accelerateRelated to accelerate

🏋 GRPORelated to GRPORelated to GRPO

🐛 bugSomething isn't workingSomething isn't working

🚀 deepspeedRelated to deepspeedRelated to deepspeed

Status: Open.

#4535 In huggingface/trl;

· 1luik opened on Nov 17, 2025

Adaptive KL in GRPO

✨ enhancementNew feature or requestNew feature or request

🏋 GRPORelated to GRPORelated to GRPO

Status: Open.

#4532 In huggingface/trl;

· hakbari14 opened on Nov 17, 2025

GRPO training: Incorrect RepeatSampling when num_iterations &gt;1

🏋 GRPORelated to GRPORelated to GRPO

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#4531 In huggingface/trl;

· oapandit opened on Nov 17, 2025

How to modify the advantage computation in GRPOTrainer

❓ questionSeeking clarification or more informationSeeking clarification or more information

🏋 GRPORelated to GRPORelated to GRPO

Status: Open.

#4525 In huggingface/trl;

· Tuziking opened on Nov 14, 2025

Replace torch image by vllm image in Docker

🏋 GRPORelated to GRPORelated to GRPO

📱 cliRelated to the Command-line interfaceRelated to the Command-line interface

Status: Open.

#4523 In huggingface/trl;

· qgallouedec opened on Nov 13, 2025

Usage of gradient_accumulation_steps in GRPO

⚡accelerateRelated to accelerateRelated to accelerate

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#4511 In huggingface/trl;

· Fyeward opened on Nov 12, 2025

Can a multimodal model like Gemma be trained in the same way as a text-only model like Qwen, but with the goal of improving only its text capabilities?

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#4507 In huggingface/trl;

· Tuziking opened on Nov 11, 2025