forum
#help DPO finetuning qwen models
A possible way to automatically detect more modules to be optimized that make Finetune faster
Open
Saving to gguf - Downgrade the protobuf package to 3.20.x or lower.
does not appear to have a file named config.json
Loss stuck on 0.6931 on DPO training
support for encoder-only and encoder-decoder models
Can't load tokenizer
Custom Eval metric in training
Open
Tokenizer load error
HFValidationError
Tips on improving quality of lower parameter model's generation
Open
The attention mask is not set and cannot be inferred from input because pad token is same as eos tok
UndefinedError: 'str object' has no attribute 'role' when mapping dataset.
Open
Solved
llama 3.1 orpo / dpo kaggle
Origanl saved model doesn't work and gguf in ollama either
Fastest way to do inference on trained model
Open
The model did not return a loss from the inputs
Solved
Using Unsloth + trl Online DPO
Finetuning Llama 3.2 1B
Open
Solved
Phi 3 (unsloth/Phi-3-mini-4k-instruct) giving random responses
Open
creating and using custom dataset
Open
Does it make sense to load and finetune a model in q4 quantization and later on export it as 8 bit?
Remove <|eot_id|> after fine-tuning Llama3.2-1B-Instruct
Solved
Using Unsloth fine tuned model in transformers.js
Gemma 2 issues on raw pretraining
Open
Fine-Tuning Llama3.1 Issues
Open
Zero train loss
Open
Llama runner process has terminated: error loading model: error loading vocabulary
Cannot adapt 4bit finetuning of llama 3.1 to 8bit, with lora. Help?
How to get thousands of documents/books into prompt-completion pairs for instruction fine-tuning ?
GGUF / llama.cpp Conversion Error
Exception: data did not match any variant of untagged enum ModelWrapper at line 757452 column 3
Qwen2.5 14B Instruct giving padding error
Unrecognized processing class in unsloth/Llama-3.2-11B-Vision
How I can use Speculative Decoding with Unsloth
Unable to Run Models: Llama-3.2-3B and Llama-3.2-3B-Instruct
Solved
KTO Without Reference Model
Open
Finetuning on Llama 3.1 instruct version throws Untrained tokens found error
Open
KTO Trainer not functioning.
Issue when resuming from checkpoint
Solved
Im new with AI and Unsloth, I have some question.
Open
Alpaca Chat Template Issue with vLLM API
finetune Qwen2.5 coder instruct
Llama 8B model for inference giving drastic results when loaded from locally saved model.
Issue loading model: unsloth/Meta-Llama-3.1-8B-bnb-4bit does not appear to have a file named config.
Continuous pre-training on instruct model?
How to convert unsloth model to standard HF model e.g., to lead to vLLM?
0.5B model usage is more than 100G GPU RAM?!
Open
worse results when loading with transformers compared to unsloth.fastlangmodel
No Module Named Unsloth