forum
How does max_seq_length affect a models current context window?
Question about conversational datasets
Pretraining then merging Gemma 3 doesn't work
Open
mistral:7.0-Instruct
Gemma3 `save_pretrained_merged()` doesn't work.
Open
load_lora for Gemma3ForCausalLM
Solved
Custom Trainer
Solved
Question about dynamic quants
Request for unsloth version of canopylabs/3b-hi-ft-research_release
Open
Help in understanding train_on_responses_only
Open
Fine tuning Qwen 3 - need tips to improve final loss
Solved
Tokenizing issue after installing unsloth
Open
undefined symbol: _ZN5torch3
Open
Qwen3-8B-Q4_K_M GGUF Looping forever
Open
RuntimeError: PassManager::run failed
Open
Difference between FastLanguageModel and FastModel.
Solved
Help in finetuning Qwen2.5 for SQL Generation
Open
Saving Gemma 3 LoRA finetune failed
Open
Model acting differently in colab than ollama.
Segmentation fault (Core dumped)
Running inference on saved model behaves weirdly
Unsloth GRPO trainer evaluates entire eval set every eval_step
Help on Choosing a Model for a Prompt-Processing.
Open
Merging 4-bit Checkpoint into Phi-4 Base Model – Model Inference Inconsistent
Open
Gemma3 Finetune Save_pretrained_merged error
Vision finetune gemma3 raised "TypeError: 'int' object is not iterable."
how can i use prompt-completion dataset and completion_only_loss paramater in SFTTrainer？
AttributeError: 'LLMEngine' object has no attribute 'model_executor' - fix in this thread
Open
Does this snippet only accepts GGUF models?
QLORA's specific features
Open
How to use custom optimiser ?
Gemma3 gguf conversion doesn't work
Open
Error creating PEFT model with Qwen3MoE
Open
modernbert issue
Confused about reasoning + non-reasoning data mix in Qwen3 (14B) notebook
Open
BLEU compute metrics causes trainer to crash
Open
Help, unsloth compiled cash
GRPO custom parameters
Phi-4-reasoning not separating thought in llama.cpp
NotImplementedError: No operator found for `memory_efficient_attention_backward`
Solved
Need some help to understand the results I got from custom dataset fientuning
unsloth/Qwen3-32B-unsloth-bnb-4bit
Questions regarding multi-turn conversation finetuning
Open
Thank you Unsloth bros
Cannot load local checkpoints
Qwen3 doesn't support vLLM
Open
So i tried to tune qwen using a grpo and eventually i got some problem.
Open
Unsloth: Failed to patch SmolVLMForConditionalGeneration forward function.
Fine-tuning Gemma3
Solved
Support for GLM?
Open