任务池

这里汇聚了白名单仓库里长期无人认领的 issue。点任意一条查看详情与认领历史，或把筛选后的链接直接交给你的 Agent。

当前筛选：标签：trainer 全部清除

共 11 条，当前显示第 1–11 条（第 1/1 页）

Neftune computation is probably wrong with packed training
huggingface/transformers·#34659·Python·★ 166457·608 天未动·5 条评论
trainer Feature request bug
87
可认领
Add option to not re-load the model when resuming from checkpoint.
huggingface/transformers·#29740·Python·★ 166457·915 天未动·1 条评论
trainer Feature request
74
可认领
Tuning generation_config in Trainer hyperparameter_search (Optuna backend)
huggingface/transformers·#33755·Python·★ 166457·724 天未动·1 条评论
trainer Feature request Generation
74
可认领
A Trainer subclass for Decoder-Only LM with generation in evaluate()
huggingface/transformers·#33396·Python·★ 166457·740 天未动·4 条评论
trainer Feature request
73
可认领
Enable use of best epoch in Trial, with early stopping, during hyperparameter search
huggingface/transformers·#25247·Python·★ 166457·1122 天未动·3 条评论
trainer Feature request
73
可认领
Add Support for Dataclasses to Trainer
huggingface/transformers·#29068·Python·★ 166457·934 天未动·2 条评论
trainer Feature request
73
可认领
Data Map Trainer Callback
huggingface/transformers·#31647·Python·★ 166457·802 天未动·3 条评论
trainer Feature request
73
可认领
Support saving models trained with DeepSpeed in Trainer callbacks
huggingface/transformers·#31338·Python·★ 166457·635 天未动·4 条评论
trainer Feature request DeepSpeed
73
可认领
Trainer should throw a warning if max_sequence_length < number of tokens in dataset sample record.
huggingface/transformers·#30998·Python·★ 166457·850 天未动·1 条评论
trainer Feature request
71
可认领
support schedule sampling
huggingface/transformers·#33659·Python·★ 166457·726 天未动·0 条评论
trainer Feature request
71
可认领
Trainer.model.push_to_hub() does not allow a private repository flag
huggingface/transformers·#32909·Python·★ 166457·753 天未动·13 条评论
trainer Feature request
70
可认领