Repository navigation

Issues

New issue

🚨FAQs | 常见问题🚨
#4614 · hiyouga opened on Jun 28, 2024
关于npu训练模型总结以及疑问
#4388 · sweetning0809 opened on Jun 20, 2024
35
顺利在 Apple silicon M3 上运行 README 中 Llama3-8B 相关示例工作流的小波折
#4341 · mapix opened on Jun 17, 2024
7

有关大模型分类，像jev这种
hiyouga/LlamaFactory#10890 · opened on Oct 14, 2026
Transparent images become black: _preprocess_image drops alpha with convert("RGB")
hiyouga/LlamaFactory#10887 · opened on Oct 12, 2026
更新微信群二维码
hiyouga/LlamaFactory#10881 · opened on Oct 9, 2026
训练Qwen3.8-27B时使用Deepspeed错误
hiyouga/LlamaFactory#10880 · opened on Oct 9, 2026
[Feature discussion] Automatically load common dataset files and schemas, with dataset_info.json as a fallback
hiyouga/LlamaFactory#10879 · opened on Oct 9, 2026
TypeError: unhashable type: 'set' when loading LoRA adapter on Llama-3.1-8B
hiyouga/LlamaFactory#10859 · opened on Sep 30, 2026
[Bug] SGLang async chat blocks the event loop because to_thread only constructs the generator
hiyouga/LlamaFactory#10858 · opened on Sep 30, 2026
EvalPort: portable interchange for ComputeAccuracy/ComputeSimilarity results
hiyouga/LlamaFactory#10850 · opened on Sep 27, 2026
新手提问：微调 Qwen2.5-0.5B 最低需要什么显卡配置？
hiyouga/LlamaFactory#10841 · opened on Sep 17, 2026
3
[First-time contributor] Looking for beginner-friendly issues to contribute
hiyouga/LlamaFactory#10821 · opened on Sep 8, 2026
[BUG] KTO with mrope models (Qwen3.5 family) crashes at step 0: "Multimodal data was passed ... but mm_token_type_ids is missing"
hiyouga/LlamaFactory#10817 · opened on Sep 5, 2026
[Bug] transformers >= 5.15 removes the GDN kernel attributes: NPU patch becomes a silent no-op, GDN context parallel raises
hiyouga/LlamaFactory#10788 · opened on Aug 25, 2026
4
是否支持LiquidAI/LFM2.5-2.6B
hiyouga/LlamaFactory#10781 · opened on Aug 24, 2026
[Updated] transformers>=4.55.0 floor still allows a broken 4.x version (extra_special_tokens list crash)
hiyouga/LlamaFactory#10766 · opened on Aug 21, 2026
ppo报错
hiyouga/LlamaFactory#10757 · opened on Aug 17, 2026
3
Prevent Margin Leaks & Surprise LLM Bills: Add a Hybrid Spend Firewall
hiyouga/LlamaFactory#10756 · opened on Aug 16, 2026
[Bug] DPO training crashes with Qwen3.5-35B-A3B (MoE VLM): TypeError: expected Tensor as element 0 in argument 0, but got list in collator
hiyouga/LlamaFactory#10755 · opened on Aug 14, 2026
1
about ppo unwrap_model
hiyouga/LlamaFactory#10740 · opened on Aug 10, 2026
3
Python 3.14 support addition
hiyouga/LlamaFactory#10732 · opened on Aug 7, 2026
2
RFC: Expert Parallel Training in LlamaFactory V1 with FSDPTurbo
hiyouga/LlamaFactory#10721 · opened on Aug 4, 2026
top_p=0.0 is silently rewritten to 1.0 before reaching vLLM, inverting the request
hiyouga/LlamaFactory#10709 · opened on Aug 1, 2026
2
单机多卡，开启swanlab日志输出，报错：[rank0]: raise RuntimeError("No active Run. Call swanlab.init() first.")
hiyouga/LlamaFactory#10704 · opened on Jul 30, 2026
1
_calculate_timestamps参数使用
hiyouga/LlamaFactory#10675 · opened on Jul 23, 2026
3
Feature Request: Support semantic-aware sequence packing for improved training efficiency
hiyouga/LlamaFactory#10673 · opened on Jul 23, 2026
1
Feature Request: Support per-epoch resampling for interleave dataset mixing strategy
hiyouga/LlamaFactory#10672 · opened on Jul 23, 2026

Navigation Menu

Repository navigation

Issues

Issues

有关大模型分类，像jev这种

Transparent images become black: `_preprocess_image` drops alpha with `convert("RGB")`

更新微信群二维码

训练Qwen3.8-27B时使用Deepspeed错误

[Feature discussion] Automatically load common dataset files and schemas, with dataset_info.json as a fallback

TypeError: unhashable type: 'set' when loading LoRA adapter on Llama-3.1-8B

[Bug] SGLang async chat blocks the event loop because to_thread only constructs the generator

EvalPort: portable interchange for ComputeAccuracy/ComputeSimilarity results

新手提问：微调 Qwen2.5-0.5B 最低需要什么显卡配置？

[First-time contributor] Looking for beginner-friendly issues to contribute

[BUG] KTO with mrope models (Qwen3.5 family) crashes at step 0: "Multimodal data was passed ... but mm_token_type_ids is missing"

[Bug] transformers >= 5.15 removes the GDN kernel attributes: NPU patch becomes a silent no-op, GDN context parallel raises

是否支持LiquidAI/LFM2.5-2.6B

[Updated] transformers>=4.55.0 floor still allows a broken 4.x version (extra_special_tokens list crash)

ppo报错

Prevent Margin Leaks & Surprise LLM Bills: Add a Hybrid Spend Firewall

[Bug] DPO training crashes with Qwen3.5-35B-A3B (MoE VLM): TypeError: expected Tensor as element 0 in argument 0, but got list in collator

about ppo unwrap_model

Python 3.14 support addition

RFC: Expert Parallel Training in LlamaFactory V1 with FSDPTurbo

top_p=0.0 is silently rewritten to 1.0 before reaching vLLM, inverting the request

单机多卡，开启swanlab日志输出，报错：[rank0]: raise RuntimeError("No active Run. Call swanlab.init() first.")

_calculate_timestamps参数使用

Feature Request: Support semantic-aware sequence packing for improved training efficiency

Feature Request: Support per-epoch resampling for interleave dataset mixing strategy