Search results

基于qwen3-14B-base模型进行sft训练，解码出现大量重复解码问题

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8416 In hiyouga/LLaMA-Factory;

· LemonDrinkTea opened on Jun 20, 2025

layer-wise GaLore仅支持单卡，对于200B+参数量的模型应该怎么完成finetune

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8415 In hiyouga/LLaMA-Factory;

· PhzCode opened on Jun 20, 2025

如何在开启deepspeed zero3的情况下将保存的optimizer状态合并？

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8413 In hiyouga/LLaMA-Factory;

· zbnlala opened on Jun 19, 2025

importlib.metadata.PackageNotFoundError: No package metadata was found for The 'vllm' distribution was not found and is required by this application. To fix: run <code>pip install vllm</code>.

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8411 In hiyouga/LLaMA-Factory;

· EASTERNTIGER opened on Jun 19, 2025

The shape of attention_mask doesn't match router loss calculation.

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8405 In hiyouga/LLaMA-Factory;

· Vermouth7 opened on Jun 18, 2025

为什么sft训练了0.1几个epoch突然断开了？

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8401 In hiyouga/LLaMA-Factory;

· ZackLiang opened on Jun 18, 2025

无法访问首页的问题

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8400 In hiyouga/LLaMA-Factory;

· ytfrdfiw opened on Jun 18, 2025

webui界面访问后端的ip:port与实际不符导致界面渲染异常

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8398 In hiyouga/LLaMA-Factory;

· Asonxw opened on Jun 17, 2025

Solution for TypeError: argument of type 'NoneType' is not iterable at if v not in ALL_PARALLEL_STYLES

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8397 In hiyouga/LLaMA-Factory;

· hiyouga opened on Jun 17, 2025

reward model推理结果不一致

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8391 In hiyouga/LLaMA-Factory;

· snowyrain opened on Jun 17, 2025

load_best_model_at_end 设置为 True，accelerate launch 分布式训练，trainer.train() 结束后 0卡显存泄露

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8384 In hiyouga/LLaMA-Factory;

· TpBlair opened on Jun 16, 2025

How can tensor parallelism be implemented under this framework when pretraining a 14B Qwen model on A40 GPUs?

bugSomething isn't workingSomething isn't working

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

Status: Open.

#8376 In hiyouga/LLaMA-Factory;

· Pixel-anter opened on Jun 14, 2025