Search results

自定义模型如何创建kt_optimize_rule

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9629 In hiyouga/LLaMA-Factory;

· Lorenz5622 opened on Dec 18, 2025

Training hangs during backward pass with MoE models when some experts are not activated

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9628 In hiyouga/LLaMA-Factory;

· jiaqiw09 opened on Dec 18, 2025

Ascend deepspeed zero3-offload 全参微调 Qwen3-VL-30B-A3B速度明显慢于 Qwen3-VL-32B模型

bugSomething isn't workingSomething isn't working

npuThis problem is related to NPU devicesThis problem is related to NPU devices

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

Status: Open.

#9625 In hiyouga/LLaMA-Factory;

· ZhengChang467 opened on Dec 17, 2025

PPO LoRA training with Qwen-14B on Ascend NPU: past_key_values NoneType error in generate (v0.9.4.dev0 + DeepSpeed)

bugSomething isn't workingSomething isn't working

npuThis problem is related to NPU devicesThis problem is related to NPU devices

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

Status: Open.

#9623 In hiyouga/LLaMA-Factory;

· shanghaiyangming opened on Dec 17, 2025

Qwen3_VL模型VQA任务微调，video_fps失效

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9620 In hiyouga/LLaMA-Factory;

· BianYuang opened on Dec 16, 2025

sft时 instruction 和 input是在哪个代码里拼起来的，没找到

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9614 In hiyouga/LLaMA-Factory;

· TZYSJTU opened on Dec 15, 2025

卡死不动， 多卡多机

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9603 In hiyouga/LLaMA-Factory;

· Codingfarmer-hkl opened on Dec 11, 2025

[RFC] Upgrade trl Dependency to Latest Version to Resolve Compatibility

enhancementNew feature or requestNew feature or request

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

Status: Open.

#9601 In hiyouga/LLaMA-Factory;

· jiaqiw09 opened on Dec 11, 2025

Invalid condition in "Dropped invalid example"

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9600 In hiyouga/LLaMA-Factory;

· ngminhthanh12a3 opened on Dec 10, 2025

lora 微调后，使用vllm_infer.py推理问题

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9598 In hiyouga/LLaMA-Factory;

· houliangxue opened on Dec 10, 2025

预训练Streaming 加载 fineweb 本地数据集

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9596 In hiyouga/LLaMA-Factory;

· windows10207 opened on Dec 10, 2025

RoPE scaling configuration not applied when using mcore_adapter for training

bugSomething isn't workingSomething isn't working

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

Status: Open.

#9589 In hiyouga/LLaMA-Factory;

· sijyy opened on Dec 9, 2025