Search results

合并lora微调后模型+vllm部署与直接使用llamafactory分别加载基座模型与checkpoint建立的接口推理效果不一致

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10596 In hiyouga/LlamaFactory;

· liyf-llm opened on Jun 22, 2026

Unsloth vs. LlamaFactory - incompatible torch

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10593 In hiyouga/LlamaFactory;

· mykeehu opened on Jun 19, 2026

Qwen3.5-122B-A10B lora SFT，仅把lora_target从all改为q_proj, v_proj，加载模型出现Cuda out of memory

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10591 In hiyouga/LlamaFactory;

· HawkFaust opened on Jun 18, 2026

How to set a random seed in the weibuUI interface to control randomness

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10581 In hiyouga/LlamaFactory;

· starsh0ne opened on Jun 16, 2026

训练Qwen3.5-4B时报错了

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10580 In hiyouga/LlamaFactory;

· lbl1120 opened on Jun 16, 2026

Feature Proposal: Add FunASR ASR Support for Multimodal Training Data

Status: Open.

#10568 In hiyouga/LlamaFactory;

· LauraGPT opened on Jun 12, 2026

[Doc] Misleading _nothink suffix warning for reasoning-only model families (e.g. Qwen3.6)

Status: Open.

#10566 In hiyouga/LlamaFactory;

· Melanie-czm opened on Jun 11, 2026

scripts/megatron_merge.py,转换报错,qwen3.5_27b

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10565 In hiyouga/LlamaFactory;

· yc0815024 opened on Jun 11, 2026

通过rancher镜像部署，配置volcano虚拟化GPU无法正常启动

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10564 In hiyouga/LlamaFactory;

· CCzzzzzzz opened on Jun 11, 2026

使用lamafactory微调时模型config文件指定了tie_word_embedding=true，但是训练结束后还是保留了lm.head权重

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10560 In hiyouga/LlamaFactory;

· Xinhao33 opened on Jun 10, 2026

Will Gemma4 12b fine-tuning be supported?

enhancementNew feature or requestNew feature or request

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10548 In hiyouga/LlamaFactory;

· lwang10-atlassian opened on Jun 3, 2026

Qwen Omni 2.5 Inference with vllm_infer.py stuck

bugSomething isn't workingSomething isn't working

pendingThis problem is yet to be addressedThis problem is yet to be addressed

Status: Open.

#10546 In hiyouga/LlamaFactory;

· AnnikaIgl opened on Jun 2, 2026