Daniel Han

icon location San Francisco icon location Unsloth - 2x faster 70% less VRAM finetuning Llama-3.1, Mistral, Gemma-2, Phi-3

Results 1105 comments of Daniel Han
trafficstars

Are there any guidelines for loading a CPT (continued pre-training) model and retraining it on a different data set?

@daegonYu Yes that should work (I think) - The continued pretraining notebook does train on the same LoRA adapters twice - https://colab.research.google.com/drive/1tEd1FrOXWMnCU9UIvdYhs61tkxdMuKZu?usp=sharing so it should function (hopefully)

Are there any guidelines for loading a CPT (continued pre-training) model and retraining it on a different data set?

@daegonYu You might be interested in our conversational notebook which masks out the instruction - https://colab.research.google.com/drive/1T5-zKWM_5OD21QHwXHiV9ixTRR7k3iB9?usp=sharing Also see https://github.com/unslothai/unsloth/wiki#train-on-completions--responses-only-do-not-train-on-inputs

Are there any guidelines for loading a CPT (continued pre-training) model and retraining it on a different data set?

@daegonYu Sorry on the delay! Yes they're equivalent EXCEPT if you're doing more than 1 conversation. HF's one does not support it, whilst Unsloth does.

Are there any guidelines for loading a CPT (continued pre-training) model and retraining it on a different data set?

@Candice1995 Apologies on the delay - DPO has a prompt, then 2 other fields - the accepted or rejected answer to the prompt. These fields have varying lengths, and so...

Add tool calling demo notebook to README.md

Thanks! Is there a way to actually make `calculate_loan_emi` executable? Maybe via exec / eval?

`functools.partial` object has no attribute `apply_chat_template` [FIXED]

Apologies just fixed - please do `pip install --force-reinstall --no-deps --no-cache-dir unsloth unsloth_zoo`

Bug in `offload_input_embeddings` with `Pytorch` 2.6.0

Will fix this asap!

After LoRa training (or loading the checkpoint) consecutive inference gives different results even if do_sample is False

Is this single batched inference?

After LoRa training (or loading the checkpoint) consecutive inference gives different results even if do_sample is False

@ziemowit-s Maybe I might have solved it, but unsure with yesterday's patch

Can I Fine-Tune a Model on CPU Using Unsloth?

Currently not, but we plan to provide that in the future - currently please use HuggingFace