Search results

LogSoftmaxBackward0 returns NaN during training on Kaggle versioned runs (Granite 2B 4bit)

bug

Status: Open.

#3002 In unslothai/unsloth;

· bx0-0 opened on Jul 18, 2025

[Bug] Qwen3Moe(ForCausalLM) does not work with LoRA at int4 because some data is float32

bug

Status: Open.

#3001 In unslothai/unsloth;

· Killusions opened on Jul 18, 2025

[Bug] Qwen3Moe(ForCausalLM) does not respect UNSLOTH_RETURN_HIDDEN_STATES when loss and labels are given.

bug

Status: Open.

#3000 In unslothai/unsloth;

· Killusions opened on Jul 18, 2025

[Bug] Gemma 2 Notebook: 'Can't find 'adapter_config.json'

bug

Status: Open.

#2997 In unslothai/unsloth;

· greenirvavril opened on Jul 18, 2025

Support for finetuning intern vl 2.5 - 1B

feature requestFeature request pending on roadmapFeature request pending on roadmap

Status: Open.

#2996 In unslothai/unsloth;

· azimb-170 opened on Jul 18, 2025

[Feature] QAT support

feature requestFeature request pending on roadmapFeature request pending on roadmap

Status: Open.

#2993 In unslothai/unsloth;

· FilippoBoni1921 opened on Jul 18, 2025

[Bug] RuntimeError: *** Unsloth: Failed compiling llama.cpp with WARNING:hf-to-gguf:************************************************************************************** . Please report this ASAP!

bug

Status: Open.

#2984 In unslothai/unsloth;

· C-rypto opened on Jul 17, 2025

[Bug] ModuleNotFoundError: No module named 'vllm.model_executor.model_loader.bitsandbytes_loader'

bug

Status: Open.

#2983 In unslothai/unsloth;

· hellcastter opened on Jul 17, 2025

Already found peft_config warning without any model/adapter merging

Status: Open.

#2981 In unslothai/unsloth;

· nerner94 opened on Jul 17, 2025

[Bug] Installation doesn't seem to get compatible vllm version

unsure bug?I'm unsureI'm unsure

Status: Open.

#2977 In unslothai/unsloth;

· csanadpoda opened on Jul 16, 2025

[Bug] unwrap_model_for_generation switches model mode from eval to train

unsure bug?I'm unsureI'm unsure

Status: Open.

#2973 In unslothai/unsloth;

· tilaks opened on Jul 15, 2025

RuntimeError: Trying to backward through the graph a second time (or directly access saved tensors after they have already been freed). Saved intermediate values of the graph are freed when you call .backward() or autograd.grad(). Specify retain_graph=True if you need to backward through the graph a second time or if you need to access saved tensors after calling backward.

unsure bug?I'm unsureI'm unsure

Status: Open.

#2972 In unslothai/unsloth;

· adi-lb-phoenix opened on Jul 15, 2025