fla-org/flash-linear-attention

News

Models

Installation

# CUDA
pip install flash-linear-attention[cuda]

# ROCm
pip install --index-url https://download.pytorch.org/whl/rocm7.2 torch
pip install flash-linear-attention[rocm]

Note

Choose a backend extra to install the appropriate torch and triton dependencies. Bare pip install flash-linear-attention does not install them.

Usage

Token Mixing

>>> import torch
>>> from fla.layers import MultiScaleRetention
>>> batch_size, num_heads, seq_len, hidden_size = 32, 4, 2048, 1024
>>> device, dtype = 'cuda:0', torch.bfloat16
>>> retnet = MultiScaleRetention(hidden_size=hidden_size, num_heads=num_heads).to(device=device, dtype=dtype)
>>> x = torch.randn(batch_size, seq_len, hidden_size).to(device=device, dtype=dtype)
>>> y, *_ = retnet(x)
>>> y.shape
torch.Size([32, 2048, 1024])

>>> from fla.models import GLAConfig
>>> from transformers import AutoModelForCausalLM
>>> config = GLAConfig()
>>> model = AutoModelForCausalLM.from_config(config)

Fused Modules

Important

You can set fuse_linear_cross_entropy in the model configuration to enable or disable the fused linear cross entropy loss.

This fused implementation is more memory-efficient but may reduce numerical precision. Due to this trade-off, it is disabled by default. If you enable this feature and encounter training instability (e.g., loss divergence), we recommend disabling it to see if the issue is resolved.

Generation

>>> import fla
>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> name = 'fla-hub/gla-1.3B-100B'
>>> tokenizer = AutoTokenizer.from_pretrained(name)
>>> model = AutoModelForCausalLM.from_pretrained(name).cuda()
>>> input_prompt = "Power goes with permanence. Impermanence is impotence. And rotation is castration."
>>> input_ids = tokenizer(input_prompt, return_tensors="pt").input_ids.cuda()
>>> outputs = model.generate(input_ids, max_length=64)
>>> tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]

$ python -m benchmarks.benchmark_generation \
  --path 'fla-hub/gla-1.3B-100B' \
  --repetition_penalty 2. \
  --prompt="Hello everyone, I'm Songlin Yang"

Prompt:
Hello everyone, I'm Songlin Yang
Generated:
Hello everyone, I'm Songlin Yang.
I am a 20 year old girl from China who is currently studying in the United States of America for my Master degree and also working as an English teacher at school here on campus since last summer (1st semester). My main goal to be able do well with this course so that we can have

Prompt length: 10, generation length: 64
Total prompt processing + decoding time: 4593ms

>>> from huggingface_hub import list_models
>>> for model in list_models(author='fla-hub'): print(model.id)

Hybrid Models

>>> from fla.models import SambaConfig
>>> from transformers import AutoModelForCausalLM
>>> config = SambaConfig(num_hidden_layers=2)
>>> config.attn = {
  'layers': [1],
  'num_heads': 18,
  'num_kv_heads': 18,
  'qkv_bias': False,
  'rope_theta': 10000.,
  'window_size': 2048
}
>>> model = AutoModelForCausalLM.from_config(config)

>>> config = SambaConfig(
...   num_hidden_layers=6,
...   attn=[
...     {
...       'layers': [1, 3],
...       'num_heads': 18,
...       'num_kv_heads': 18,
...       'qkv_bias': False,
...       'rope_theta': 10000.,
...       'window_size': 2048,
...     },
...     {
...       'layers': [5],
...       'num_heads': 18,
...       'num_kv_heads': 18,
...       'qkv_bias': False,
...       'rope_theta': 10000.,
...       'window_size': None,
...     },
...   ],
... )
>>> model = AutoModelForCausalLM.from_config(config)

Training

Evaluation

Benchmarks

# you might have to first install `fla` via `pip install -e .` to enable its import
$ python -m benchmarks.ops.run --op chunk_retention chunk_gla chunk_gdn flash_attn
=================================================================================
  Machine: NVIDIA GB200 | CUDA 12.9 | PyTorch 2.9.0+cu129.msh
=================================================================================
  fwd        B      T    H    D  op                            main[0a484709](ms)
          -----------------------------------------------------------------------
             1   8192   96  128  chunk_retention                            0.787
                                 chunk_gla                                  1.765
                                 chunk_gdn                                  1.265
                                 flash_attn                                 3.753
          -----------------------------------------------------------------------
             2  16384   16  128  chunk_retention                            0.792
                                 chunk_gla                                  1.445
                                 chunk_gdn                                  1.029
                                 flash_attn                                 5.035
          -----------------------------------------------------------------------
             4   2048   16  128  chunk_retention                            0.559
                                 chunk_gla                                  0.514
                                 chunk_gdn                                  0.753
                                 flash_attn                                 0.346
          -----------------------------------------------------------------------
             4   4096   64  128  chunk_retention                            0.997
                                 chunk_gla                                  2.251
                                 chunk_gdn                                  1.581
                                 flash_attn                                 2.560
          -----------------------------------------------------------------------
             8   1024    8   64  chunk_retention                            0.425
                                 chunk_gla                                  0.358
                                 chunk_gdn                                  0.631
                                 flash_attn                                 0.157
          -----------------------------------------------------------------------
             8   2048   32  256  chunk_retention                            1.174
                                 chunk_gla                                  2.897
                                 chunk_gdn                                  1.831
                                 flash_attn                                 1.408
=================================================================================
  fwdbwd     B      T    H    D  op                            main[0a484709](ms)
          -----------------------------------------------------------------------
             1   8192   96  128  chunk_retention                            2.618
                                 chunk_gla                                  7.670
                                 chunk_gdn                                  4.738
                                 flash_attn                                15.371
          -----------------------------------------------------------------------
             2  16384   16  128  chunk_retention                            2.122
                                 chunk_gla                                  5.984
                                 chunk_gdn                                  3.616
                                 flash_attn                                19.960
          -----------------------------------------------------------------------
             4   2048   16  128  chunk_retention                            1.047
                                 chunk_gla                                  1.434
                                 chunk_gdn                                  2.085
                                 flash_attn                                 0.902
          -----------------------------------------------------------------------
             4   4096   64  128  chunk_retention                            3.459
                                 chunk_gla                                 10.216
                                 chunk_gdn                                  5.964
                                 flash_attn                                10.815
          -----------------------------------------------------------------------
             8   1024    8   64  chunk_retention                            0.898
                                 chunk_gla                                  1.707
                                 chunk_gdn                                  1.974
                                 flash_attn                                 0.477
          -----------------------------------------------------------------------
             8   2048   32  256  chunk_retention                           51.103
                                 chunk_gla                                 13.797
                                 chunk_gdn                                  8.644
                                 flash_attn                                 6.748
=================================================================================

Citation

@software{yang2024fla,
  title  = {FLA: A Triton-Based Library for Hardware-Efficient Implementations of Linear Attention Mechanism},
  author = {Yang, Songlin and Zhang, Yu},
  url    = {https://github.com/fla-org/flash-linear-attention},
  month  = jan,
  year   = {2024}
}

@misc{chen2026attnres,
  title         = {Attention Residuals},
  author        = {Chen, Guangyu  and Zhang, Yu  and Su, Jianlin  and Xu, Weixin  and Pan, Siyuan  and Wang, Yaoyu  and Wang, Yucheng  and Chen, Guanduo  and Yin, Bohong  and Chen, Yutian  and Yan, Junjie  and Wei, Ming  and Zhang, Y.  and Meng, Fanqing  and Hong, Chao  and Xie, Xiaotong  and Liu, Shaowei  and Lu, Enzhe  and Tai, Yunpeng  and Chen, Yanru  and Men, Xin  and Guo, Haiqing  and Charles, Y.  and Lu, Haoyu  and Sui, Lin  and Zhu, Jinguo  and Zhou, Zaida  and He, Weiran  and Huang, Weixiao  and Xu, Xinran  and Wang, Yuzhi  and Lai, Guokun  and Du, Yulun  and Wu, Yuxin  and Yang, Zhilin  and Zhou, Xinyu},
  year          = {2026},
  eprint        = {2603.15031},
  archiveprefix = {arXiv},
  primaryclass  = {cs.CL}
}

@misc{zhang2025kda,
  title         = {Kimi Linear: An Expressive, Efficient Attention Architecture},
  author        = {Zhang, Yu  and Lin, Zongyu  and Yao, Xingcheng  and Hu, Jiaxi  and Meng, Fanqing  and Liu, Chengyin  and Men, Xin  and Yang, Songlin  and Li, Zhiyuan  and Li, Wentao  and Lu, Enzhe  and Liu, Weizhou  and Chen, Yanru  and Xu, Weixin  and Yu, Longhui  and Wang, Yejie  and Fan, Yu  and Zhong, Longguang  and Yuan, Enming  and Zhang, Dehao  and Zhang, Yizhi  and T. Liu, Y.  and Wang, Haiming  and Fang, Shengjun  and He, Weiran  and Liu, Shaowei  and Li, Yiwei  and Su, Jianlin  and Qiu, Jiezhong  and Pang, Bo  and Yan, Junjie  and Jiang, Zhejun  and Huang, Weixiao  and Yin, Bohong  and You, Jiacheng  and Wei, Chu  and Wang, Zhengtao  and Hong, Chao  and Chen, Yutian  and Chen, Guanduo  and Wang, Yucheng  and Zheng, Huabin  and Wang, Feng  and Liu, Yibo  and Dong, Mengnan  and Zhang, Zheng  and Pan, Siyuan  and Wu, Wenhao  and Wu, Yuhao  and Guan, Longyu  and Tao, Jiawen  and Fu, Guohong  and Xu, Xinran  and Wang, Yuzhi  and Lai, Guokun  and Wu, Yuxin  and Zhou, Xinyu  and Yang, Zhilin  and Du, Yulun},
  year          = {2025},
  eprint        = {2510.26692},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL}
}

@inproceedings{yang2025path,
  title     = {PaTH Attention: Position Encoding via Accumulating Householder Transformations},
  author    = {Yang, Songlin  and Shen, Yikang and Wen, Kaiyue and Tan, Shawn  and Mishra, Mayank  and Ren, Liliang  and Panda, Rameswar  and Kim, Yoon},
  booktitle = {Proceedings of NeurIPS},
  year      = {2025}
}

@inproceedings{yang2024gdn,
  title     = {Gated Delta Networks: Improving Mamba2 with Delta Rule},
  author    = {Yang, Songlin  and Kautz, Jan  and Hatamizadeh, Ali},
  booktitle = {Proceedings of ICLR},
  year      = {2025}
}

@inproceedings{yang2024deltanet,
  title     = {Parallelizing Linear Transformers with the Delta Rule over Sequence Length},
  author    = {Yang, Songlin and Wang, Bailin and Zhang, Yu and Shen, Yikang and Kim, Yoon},
  booktitle = {Proceedings of NeurIPS},
  year      = {2024}
}

@inproceedings{zhang2024gsa,
  title     = {Gated Slot Attention for Efficient Linear-Time Sequence Modeling},
  author    = {Zhang, Yu and Yang, Songlin and Zhu, Ruijie and Zhang, Yue and Cui, Leyang and Wang, Yiqiao and Wang, Bolun and Shi, Freda and Wang, Bailin and Bi, Wei and Zhou, Peng and Fu, Guohong},
  booktitle = {Proceedings of NeurIPS},
  year      = {2024}
}

@inproceedings{qin2024hgrn2,
  title     = {HGRN2: Gated Linear RNNs with State Expansion},
  author    = {Qin, Zhen and Yang, Songlin and Sun, Weixuan and Shen, Xuyang and Li, Dong and Sun, Weigao and Zhong, Yiran},
  booktitle = {Proceedings of COLM},
  year      = {2024}
}

@inproceedings{yang2024gla,
  title     = {Gated Linear Attention Transformers with Hardware-Efficient Training},
  author    = {Yang, Songlin and Wang, Bailin and Shen, Yikang and Panda, Rameswar and Kim, Yoon},
  booktitle = {Proceedings of ICML},
  year      = {2024}
}

Acknowledgements