Search results

Add VAS to TRL

✨ enhancementNew feature or requestNew feature or request

Status: Open (in progress).

huggingface/trlnumber 2195

#2195 In huggingface/trl;

· idanshen opened on Oct 7, 2024

[CLI] Extend training support to all trainers

✨ enhancementNew feature or requestNew feature or request

👶 good first issueGood for newcomersGood for newcomers

Status: Open.

#2101 In huggingface/trl;

· lewtun opened on Sep 23, 2024

DataCollatorForCompletionOnlyLM not working with MPS

✨ enhancementNew feature or requestNew feature or request

🏋 SFTRelated to SFTRelated to SFT

🐛 bugSomething isn't workingSomething isn't working

Status: Open.

#1913 In huggingface/trl;

· giuliabaldini opened on Aug 9, 2024

Training on Teacher model logits

✨ enhancementNew feature or requestNew feature or request

Status: Open.

#1975 In huggingface/trl;

· Theodotus1243 opened on Aug 26, 2024

Correct masking when the same roles are present in adjacent messages in DataCollatorForCompletionOnlyLM

✨ enhancementNew feature or requestNew feature or request

🗃️ dataRelated to dataRelated to data

Status: Open.

#1994 In huggingface/trl;

· lsy641 opened on Aug 30, 2024

[Data] Implement dataset mixer for combining datasets in training

✨ enhancementNew feature or requestNew feature or request

🗃️ dataRelated to dataRelated to data

Status: Open.

#2112 In huggingface/trl;

· lewtun opened on Sep 24, 2024

PPOv2Trainer throws AttributeError: 'NoneType' object has no attribute 'modules' because value_model's default is None

✨ enhancementNew feature or requestNew feature or request

🏋 PPORelated to PPORelated to PPO

🙋 help from community wantedOpen invitation for community members to contributeOpen invitation for community members to contribute

🧒 good second issueGood for contributors with basic project familiarityGood for contributors with basic project familiarity

Status: Open.

#1976 In huggingface/trl;

· RylanSchaeffer opened on Aug 26, 2024

Always allow ref_model=None

✨ enhancementNew feature or requestNew feature or request

🏋 DPORelated to DPORelated to DPO

🙋 help from community wantedOpen invitation for community members to contributeOpen invitation for community members to contribute

🧒 good second issueGood for contributors with basic project familiarityGood for contributors with basic project familiarity

Status: Open.

#2047 In huggingface/trl;

· qgallouedec opened on Sep 10, 2024

Packing without cross contamination

✨ enhancementNew feature or requestNew feature or request

🙋 help from community wantedOpen invitation for community members to contributeOpen invitation for community members to contribute

🧒 good second issueGood for contributors with basic project familiarityGood for contributors with basic project familiarity

Status: Open.

#1230 In huggingface/trl;

· nivibilla opened on Jan 15, 2024

RLOOTrainer & PPOv2Trainer - Modify Name for W&B Logged Table

✨ enhancementNew feature or requestNew feature or request

🏋 PPORelated to PPORelated to PPO

🏋 RLOORelated to RLOORelated to RLOO

🙋 help from community wantedOpen invitation for community members to contributeOpen invitation for community members to contribute

Status: Open.

#2045 In huggingface/trl;

· RylanSchaeffer opened on Sep 9, 2024

PPOv2Trainer and RLOOTrainer: Remove the implicit assumption that the reward model & policy model share the same tokenizer

✨ enhancementNew feature or requestNew feature or request

🏋 Online DPORelated to Online DPORelated to Online DPO

🏋 PPORelated to PPORelated to PPO

🙋 help from community wantedOpen invitation for community members to contributeOpen invitation for community members to contribute

Status: Open.

#1979 In huggingface/trl;

· RylanSchaeffer opened on Aug 26, 2024

Add support for ShareGPT-formatted datasets

✨ enhancementNew feature or requestNew feature or request

🗃️ dataRelated to dataRelated to data

Status: Open.

#2083 In huggingface/trl;

· lewtun opened on Sep 19, 2024