forum

1495 threads · Page 1 of 30

Is the dataset passed to `UnslothTrainer` shuffled by default? 3 messages
Open
GRPO fails in terms of accuracy in datasets different than gsm8k 2 messages
Any plans for Unsloth Dynamic 2.0 GGUFs for DeepSeek V4 Flash? 2 messages
Open
Unusual loss curves 3 messages
Env mask for custom rollout_func 8 messages
Solved
Speech to text models training without other language performance degradation 6 messages
how to install unsloth in offline environment 7 messages
trying to find something that can create a bot for tetris 2 messages
Alternatives to qlora finetuning on qwen3.6 27B with rtx5090? 4 messages
No cuda support on last update unsloth 2026.5.2 is up to date 7 messages
Open
Unsloth loading model weights at full precision even though I have manually set it to load at 4 bit. 242 messages
Solved
Trying to find simple GPRO/RLHF example (no reasoning) 36 messages
Open
Qwen 3.6-27B only text-to-text in llama.cpp 3 messages
Qwen3.5-122B-A10B-GGUF Reasoning/Tool call issues 3 messages
Windows AMD ROCm issue 4 messages
Bug in Unsloth Docker Upload Failed 41 messages
Open
UnslothGRPOTrainer switches on gradient checkpointing when it shouldnt 68 messages
Strix Halo Unsloth fails to detect unified memory (0.35GB VRAM) 86 messages
Open
Speculative decoding: `--spec-ngram-size-n` argument has been removed 2 messages
LoRA for TokenClassification 6 messages
Unsloth Studio + remote Training 3 messages
Open
Unsloth Studio + vLLM 2 messages
Open
Help me to build a RIG 2 messages
Open
Silent hang on Vision DPO path with Gemma 4 E4B 7 messages
Solved
weird mismatch error in grpo multi image training 3 messages
How do I work around OOM on LoRA merge? 5 messages
Solved
Merging lora weights from checkpoint into base model 5 messages
Why does unsloth-zoo depend on torch < 2.11.0? 3 messages
Open
Where are settings for splitting models across multiple gpus/CPU ram + Vram? 7 messages
Agents Support not working in Unsloth Studio 3 messages
Open
self and mat2 must have the same dtype, but got Half and BFloat16 - GRPO Lora Training 5 messages
Solved
torch._dynamo.exc.Unsupported: HigherOrderOperator 3 messages
Solved
Installation Error [Windows] 135 messages
Solved
How to load custom dataset into Unsloth Studio 2 messages
Solved
Linux install Unsloth Problem 5 messages
Qwen Coder with Unsloth Studio API Error 405 2 messages
Open
Confirm that native long context extension built into Unsloth 3 messages
Open
Opinion: I want to fine-tune a model for PII/PHI detection. Qwen3.5 2B (via unsloth) vs BERT-based? 8 messages
Open
Does Unsloth really support video finetune? 2 messages
i want to use the unsloth studio in kaggle, how can i do that? 10 messages
Open
Path Handling in Windows 43 messages
Unable to run Unsloth's DeepSeek OCR on Google Colab 4 messages
qwen 3.6 mlx quants dont support reasoning on lmstudio 3 messages
Trying to run unsloth/gemma-4-E4B-it-GGUF, quant: UD-Q4_K_XL 2 messages
Colab Qwen3 phone deployment 9 messages
Open Solved
Error Importing Unsloth 2 messages
Latest Ernie text to image model GGUF inside unsloth studio 10 messages
Open
colab - Unsloth studio 13 messages
Open
Fine-tuning multiple datasets of different domains and different size (no of rows) 4 messages
Gemma e4b error on Colab[A100] FlashAttention forward only supports head dimension at most 256 22 messages