flash-linear-attention and flame

Flash-linear-attention provides the optimized model implementations that flame is purpose-built to train at scale, making them complements that work together in a stack rather than alternatives.

flash-linear-attention
89
Verified
flame
52
Established
Maintenance 23/25
Adoption 21/25
Maturity 25/25
Community 20/25
Maintenance 6/25
Adoption 10/25
Maturity 16/25
Community 20/25
Stars: 4,549
Forks: 431
Downloads: 438,484
Commits (30d): 30
Language: Python
License: MIT
Stars: 355
Forks: 58
Downloads: —
Commits (30d): 0
Language: Python
License: MIT
No risk flags
No Package No Dependents

About flash-linear-attention

fla-org/flash-linear-attention

🚀 Efficient implementations of state-of-the-art linear attention models

Provides PyTorch and Triton kernels for linear attention variants (RetNet, GLA, Mamba, RWKV, DeltaNet, and 20+ emerging architectures), optimized for CPU and GPU across NVIDIA, AMD, and Intel platforms. Includes fused operators, hybrid model support, and variable-length sequence handling to reduce memory overhead during training. Integrates with Hugging Face model hub and the companion `flame` training framework for distributed model development.

About flame

fla-org/flame

🔥 A minimal training framework for scaling FLA models

Related comparisons

flash-linear-attention and SageAttention flash-linear-attention and Star-Attention flash-linear-attention and Flash-Sparse-Attention flash-linear-attention and flash_attention_inference

Scores updated daily from GitHub, PyPI, and npm data. How scores work