This repo contains fast Triton-based implementation (maybe CUTLASS/CUTE in the future) of causal linear attention (i.e., RNNs with 2D hidden states), with a specific focus on modern decoder-only language models. Join discord if you are interested in this project!
Order by my expected implementation time. If you are not using the Triton nightly release version, please avoid using the FusedChunk implementation (see issue)
| Date | Title | Paper | Code | Support |
|---|---|---|---|---|
| 2023-07 | 🔥🔥🔥[RetNet] Retentive network: a successor to transformer for large language models(@MRSA@THU) | [arxiv] | [official]  [RetNet]  |
Parallel✅ FusedRecurrent✅ FusedChunk✅ ParallelChunk ✅ |
| 2023-12 | 🔥🔥[GLA] Gated Linear Attention Transformers with Hardware-Efficient Training (@MIT@IBM) | [arxiv] | [official]  |
FusedRecurrent✅ ParallelChunk✅ FusedChunk✅ |
| 2023-12 | 🔥🔥[Based] An Educational and Effective Sequence Mixer (@Stanford Hazyresearch) | [blog] | [official]  |
Parallel✅ FusedChunk✅ |
| 2023-07 | 🔥🔥[TransnormerLLM] A Faster and Better Large Language Model with Improved TransNormer (@Shanghai AI Lab) | openreview arxiv | [official]  [Lightning2]  |
TODO |
| 2023-05 | 🔥🔥🔥[RWKV-v6] Reinventing RNNs for the Transformer Era (@BlinkDL) | arxiv | [official]  |
TODO |
| 2023-10 | 🔥[GateLoop]Fully Data-Controlled Linear Recurrence for Sequence Modeling | openreview arxiv | [jax]  |
TODO |
| 2021-10 | [ABC] Attention with Bounded-memory Control (@UW) | arxiv | - | TODO |
| 2023-09 | 🔥[VQ-transformer] Linear-Time Transformers via Vector Quantization | arxiv | [official]  |
TODO |
The following requirements should be satisfied
pip install -U --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/Triton-Nightly/pypi/simple/ triton-nightly
Currently fla is actively developed, and no released packages are provided at this time.
If you do need to use fla modules/ops and contemplate further explorations, an alternative way is to manage fla with submodules
git submodule add https://github.com/sustcsonglin/flash-linear-attention.git 3rdparty/flash-linear-attention
ln -s 3rdparty/flash-linear-attention/fla flaWe provide "token mixing" linear attention layers in fla.layers for you to use. You can replace the standard multihead attention layer in your transformer with the other linear attention layers. Example usage is as follows:
from fla.layers import MultiScaleRetention, GatedLinearAttention, BasedLinearAttention
d_model = 1024
num_head = 4
device = "cuda:0"
dtype = torch.bfloat16
retnet = MultiScaleRetention(d_model=d_model, num_head=num_head).to(device).to(dtype)
gla = GatedLinearAttention(d_model=d_model, num_head=num_head).to(device).to(dtype)
based = BasedLinearAttention(d_model=d_model, num_head=num_head).to(device).to(dtype)
bsz, seq_len, d_model = 32, 2048, 1024
x = torch.randn(bsz, seq_len, d_model).to(device).to(dtype)
y1 = retnet(x)
y2 = gla(x)
y3 = based(x)
asssert y1.shape == y2.shape == y3.shape == x.shapeWe compared our Triton-based RetNet implementation with CUDA-based FlashAttention2, using a batch size of 8, 32 heads, and a head dimension of 128, across different sequence lengths. These tests were conducted on a single A100 80GB GPU, as illustrated in the following graph
➜ benchmarks git:(main) ✗ python benchmark_retention.py
Performance:
seq_len fused_chunk_fwd chunk_fwd parallel_fwd fused_chunk_fwdbwd chunk_fwdbwd parallel_fwdbwd flash_fwd flash_fwdbwd
0 128.0 0.093184 0.185344 0.067584 1.009664 1.591296 1.044480 0.041984 0.282624
1 256.0 0.165888 0.219136 0.126976 1.024000 1.596928 1.073152 0.074752 0.413696
2 512.0 0.308224 0.397312 0.265216 1.550336 1.603584 1.301504 0.156672 0.883712
3 1024.0 0.603136 0.747520 0.706560 3.044864 3.089408 3.529728 0.467968 2.342912
4 2048.0 1.191424 1.403904 2.141184 6.010880 6.059008 11.009024 1.612800 7.135232
5 4096.0 2.377728 2.755072 7.392256 11.932672 11.938816 37.792770 5.997568 24.435200
6 8192.0 4.750336 5.491712 26.402817 23.759359 23.952385 141.014023 22.682114 90.619904
7 16384.0 9.591296 10.870784 101.262337 47.666176 48.745472 539.853821 91.346947 346.318848
Please refer to Sectiton 2.3 of GLA paper for hardware considerations of different forms of linear attention.
-
Parallel: Self-attention-styled computation in
$O(L^2)$ time with sequence parallelism. -
FusedRecurrent: Recurrent computation in
$O(L)$ time. Hidden states are computed on-the-fly in shared memory without any materialization to global memory (see Algorithm1 of this paper for more details!). This saves a lot of I/O cost and should be a strong baseline for speed comparison. -
FusedChunk: Chunkwise computation in
$O(LC)$ time where$C$ is the chunk size. Hidden states are computed on-the-fly without any materialization to global memory likewise FusedRecurrent. This version is usually better than FusedReuccurent because tensor cores can be used for sequence level "reduction", whilst FusedRecurrent cannot use tensor cores at all. Note that there is no sequence level parallelism in this implementation. More memory efficient than ParallelChunk though. -
ParallelChunk: Chunkwise computation with sequence parallelism. Need to materialize hidden states to global memory for each chunk.
$C$ is needed to set properly to achieve good performance because when$C$ is small there are too many hidden states to load/store to global memory; and when$C$ is too large the FLOPs are high. Recommened$C$ is 128 or 256. This function is generally faster than FusedChunk but less memory efficient.
If you find this repo useful, please consider citing our work:
@article{yang2023gated,
title={Gated Linear Attention Transformers with Hardware-Efficient Training},
author={Yang, Songlin and Wang, Bailin and Shen, Yikang and Panda, Rameswar and Kim, Yoon},
journal={arXiv preprint arXiv:2312.06635},
year={2023}
}