Search results

[Bug] Gradient explosion (inf/NaN) training Qwen 3.5 at >65536 sequence length

bug

Status: Open.

#4906 In unslothai/unsloth;

· djkazic opened on Apr 7, 2026

[Bug] Developer Installation fails on Windows 11 due to possibly hardcoded .venv path

bug

Status: Open.

#4905 In unslothai/unsloth;

· omkar-4 opened on Apr 7, 2026

[Bug] Windows: User Names with spaces cause uv/pip to fail on install

bug

Status: Open.

#4904 In unslothai/unsloth;

· Ambyli opened on Apr 7, 2026

[Feature] Unsloth studio: Set dataset streaming mode

feature requestFeature request pending on roadmapFeature request pending on roadmap

Status: Open.

#4903 In unslothai/unsloth;

· fablevi opened on Apr 7, 2026

[Bug] audio dataset preview and finetuning doesn't work due to torchcodec version mismatch

bug

Status: Open.

#4900 In unslothai/unsloth;

· sleepydirt opened on Apr 7, 2026

环境太严格了

Status: Open.

#4898 In unslothai/unsloth;

· 86523553 opened on Apr 7, 2026

[Feature] add custom tag or name to model training

feature requestFeature request pending on roadmapFeature request pending on roadmap

Status: Open.

#4895 In unslothai/unsloth;

· darkworks opened on Apr 7, 2026

[Bug] Context length is reduced to 4096 after failed to load a model until last successfully loaded model is unloaded manually

bug

Status: Open.

#4893 In unslothai/unsloth;

· Ackerka opened on Apr 7, 2026

[Bug] RuntimeError: self and mat2 must have the same dtype (Half and BFloat16) in matmul_lora during GRPO training with 4-bit quantization

bug

Status: Open.

#4891 In unslothai/unsloth;

· anandn1 opened on Apr 7, 2026

How to chat with finetuned model.

Status: Open.

#4888 In unslothai/unsloth;

· KunalSin9h opened on Apr 7, 2026

[Bug] Incorrect args for gradient_checkpoint_enable

bug

Status: Open.

#4886 In unslothai/unsloth;

· jonahsamost opened on Apr 7, 2026

[Bug] GRPO: grpo_compute_loss_slow called with wrong positional args

bug

Status: Open.

#4885 In unslothai/unsloth;

· jonahsamost opened on Apr 7, 2026