forum
issue in inferencing using streamlit
Open
How to load safetensor model weights after saving to hugging face?
Open
save_pretrained_merged ruins my model
Open
Dora seems to be using twice the memory?
Solved
chat template or instruction?
llama 3.2v 11B Instruct fine tuning
Issue with Inferencing fine tuned Mistral Nemo Instruct model
Open
Building memory for chatbot
Open
I want to re-use LoRA weight for continuing training.
Model stuck in a repeat loop when asking for longer responses
Open
Memory usage is higher on WSL? Any idea why?
Open
Suffering from massive drops in model accuracy after conversion to .gguf
Optimal long-context finetuning
Solved
mistral endless generation
TypeError: patch_torch_compile() got an unexpected keyword argument 'ignore_errors'
Solved
Error in fp16
Install Issues
Installing in windows, currently impossible ?
Solved
Am I preparing my dataset correctly?
multi fine tunning
train_on_responses_only setting all tokens to -100 ( instruction_part / response_part set wrong?)
Unsloth insanely slow
Open
Make strict dataset checking optional
Open
ModuleNotFoundError: No module named 'unsloth' on T4
Solved
Tensor shape error after Llama3.1 8B Instruct template fine-tuning
can't load my data dataset to train on it
Open
Memorization problem
'LlamaForCausalLM' object has no attribute 'max_seq_length'
Solved
Problems with fine tune with RAG
ValueError: The model did not return a loss from the inputs,
Natural text to Sql
Newline after EOS token from tokenizer.apply_chat_template
Solved
Crash while train half steps
Solved
Quantized models of llama 3.2 vision instruct
Unsloth: Quantization failed! for (q5_k_m, q4_k_m)
New system message not applicated with ollama after finetuning
from unsloth import FastLanguageModel
Inference UI notebook doesn't work
Open
Version of transformers and 'LlamaTokenizerFast' object has no attribute '_ollama_modelfile'
to_sharegpt()
Dataset without hugging face?
Someone help me create a valid dataset
Weird answer from llama3.1 model
Open
Solved :ModuleNotFoundError: No module named 'huggingface_hub.utils._token'
Fine-tuning question | speed of unsloth trainer
Finetune Locally
Open
Loss calculation on eval_dataset to avoid overfitting.
Training works, Validation Fails OOM (With Reproduction Notebook)
Open
Model answers differently after importing it into Ollama
Open
A100 40GB runs OOM seemingly very easily