Search results

Prompttuning/ptuning not working with SFTTrainer due to token accuracy

⚡ PEFTRelated to PEFTRelated to PEFT

🏋 SFTRelated to SFTRelated to SFT

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#3812 In huggingface/trl;

· DominikHil opened on Jul 30, 2025

GSPO-token implementation

✨ enhancementNew feature or requestNew feature or request

Status: Open.

#3811 In huggingface/trl;

· BounharAbdelaziz opened on Jul 30, 2025

Move vllm_serve FastAPI app out of main method

✨ enhancementNew feature or requestNew feature or request

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

Status: Open.

#3809 In huggingface/trl;

· nialldevlin1 opened on Jul 30, 2025

Failed to broadcast barrier_id Bug in <code>_move_model_to_vllm()</code> of grpo_trainer when training lora model

🏋 GRPORelated to GRPORelated to GRPO

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#3808 In huggingface/trl;

· shiwanghua opened on Jul 30, 2025

🐛 Bug Report: RuntimeError: element 0 of tensors does not require grad during .train() using SFTTrainer (latest trl release)

🏋 SFTRelated to SFTRelated to SFT

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#3807 In huggingface/trl;

· dt-cs opened on Jul 30, 2025

How to view the training parameters after training is completed

❓ questionSeeking clarification or more informationSeeking clarification or more information

🏋 GRPORelated to GRPORelated to GRPO

Status: Open.

#3797 In huggingface/trl;

· Tuziking opened on Jul 29, 2025

Import error with the last release v0.20.0

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#3795 In huggingface/trl;

· khalil-Hennara opened on Jul 29, 2025

GRPO max_num_batched_tokens hardcoded

🏋 GRPORelated to GRPORelated to GRPO

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#3788 In huggingface/trl;

· dgiofre opened on Jul 28, 2025

Sequence Shuffling Error with GSPO

🏋 GRPORelated to GRPORelated to GRPO

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#3787 In huggingface/trl;

· Pclanglais opened on Jul 28, 2025

OnlineDPO missing vLLM server mode support (like GRPO has)

✨ enhancementNew feature or requestNew feature or request

🏋 Online DPORelated to Online DPORelated to Online DPO

Status: Open.

#3782 In huggingface/trl;

· vaelev opened on Jul 27, 2025

Potential oversight or incompatibility in <code>SFTTrainer</code>: <code>assistant_only_loss=True</code> silently ignored when <code>use_liger_kernel=True</code>

🏋 SFTRelated to SFTRelated to SFT

🐛 bugSomething isn't workingSomething isn't working

🚀 deepspeedRelated to deepspeedRelated to deepspeed

Status: Open.

#3781 In huggingface/trl;

· andylvua opened on Jul 27, 2025

[Feature Request] Add support for padding-free reward modeling training in TRL

✨ enhancementNew feature or requestNew feature or request

🏋 RewardRelated to Reward modellingRelated to Reward modelling

Status: Open.

#3780 In huggingface/trl;

· zjjMaiMai opened on Jul 27, 2025