forum
Full-Fine tuning over unsloth
Problems installing Unsloth
Gemma3 4B vision in react native
Audio/Speech Support
Problem merging Lora with based model.
General Advice for a Low-Resource Language
Open
Fine tuning generates repetitive output
Evaluation & custom compute_metrics can't produce coherent output
Open
Multi-GPU / Unsloth Pro
Phi_4_(14B)_GRPO.ipynb
GRPO+VIsion Models
DeepSeek-V3-0324-UD coding errors in output for Heptagon Test
AttributeError: 'Gemma3Config' object has no attribute 'vocab_size'
Running dynamic 4-bit models with VLLM
Effects of multiple GPUs on training of LLMs
Gemma support with pytorch 2.4.0
AttributeError: 'Gemma3Processor' object has no attribute 'tokenize'
Open
Gemma 3 Finetuning Issues
Solved
Unsloth Qwen2.5-1.5B Finetuning Issue
Open
Getting errors after Tokenization with Gemma 3 only
Looking for tips
Solved
`vllm serve` an unsloth bnb Dynamic 4-bit model with tool support
Searching for Good Audio Tokenizer
SmolVLM2: Unexpected Indentation Error
Open
Gemma 3 Colab Notebook not training
KeyError: '' During SFTTrainer Creation When Fine Tuning Llama-3.2-11B-Vision-Instruct
Solved
vLLM Server Error
Unsloth support for Transfer Learning
Trying to run Unsloth models using vLLM with Kubernetes.
How to enable embed_tokens training in Gemma 3 with Unsloth?
Facing Error during inference: Cache only has 0 layers, attempted to access layer with index 0
Test Bug fixes without gpus
Looking for Advice: Building a Water Management Chatbot with a Fine-Tuned LLM
Open
_unsloth_get_batch_samples error when running demo locally
Open
issue in inferencing output
Open
Should it be this slow?
Solved
Failed to load quantized Gemma model from local directory after saving with save_pretrained
Open
got an unexpected keyword argument 'num_items_in_batch'
Open
Issue on with Docker image as Unsloth not using GPU drivers
Gemma3 1b: expected mat1 and mat2 to have the same dtype, but got: float != struct c10::BFloat16
New Mistral 3.1 24b model finetune
gemma3 loses multimodal support in Ollama when finetuned with Unsloth
GRPO training when you have reasoning tokens
PeftConfig error: Can't find 'adapter_config.json' at 'unsloth/gemma-3-1b-it-unsloth-bnb-4bit'
Last update causing gemma oom
Unable to Save Model; mmap can only be used with files
Solved
cant save my model in pytorch format therefore cant load it
Continued Pre-Training is broken
Open
My nightly job 20250319 failed with error VLLM_USE_V1=1 is not supported with --quantization
Fast Model Loads 4-bit Variant Even When load_in_4_bit=False