Repository Issue Activity (beta)

#227 smollm3 DPO recipe fails when using liger_kernal

#214 Unable to train SmoLlm model

#220 A little question: why num examples is much less than the total amount of my training dataset?

#218 Will you release SmolLM 3 recipe?

#37 Question about "ModuleNotFoundError: No module named 'alignment'"

#78 wierd conversation with zephyr-7b-dpo-lora

#207 run_sft.py SFTrainer: got unexpected keyword model_init_kwargs

#215 Use alignment-handbook on Apple Silicon

#170 Question about sft with deepspeed

#119 ImportError: Flash Attention 2 is not available

#213 Warning: Could not find `flash_attn`

#144 Can we please add the option to work with a tokenized dataset, escpailly for the CPT task.

#211 SmolLM2 minor issues

#210 Problem with multi-epoch training

#209 Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:1! (when checking argument for argument index in method wrapper_CUDA__index_select)

#38 DPO loss

#203 SFTTrainer.create_model_card() got an unexpected keyword argument 'finetuned_from'

#204 Cross Contamination in SFT Trainer

#9 How to finetune or lora on custom dataset

#169 Cannot flatten integer dtype tensors

#147 Multi-GPU Training with DPO Full Parameter Stucks

#201 Full parameter fine-tuning keeps consuming system RAM and lead to crash

#200 HumanEval loaded on import