dela
584f7e9e73
Standalone tree split from LLMRL/projects/kda. Includes Triton dt_bias backward fix, train_k3 --preset 0.5b, SFT, Docker runtime, and tests.
8 lines
245 B
Python
8 lines
245 B
Python
"""Vendored NVIDIA-Triton KDA path from flash-linear-attention (MIT).
|
|
|
|
This package is imported as ``kda._fla``, never as the upstream ``fla``
|
|
distribution. Context-parallel, Ascend, and TileLang backends are omitted.
|
|
"""
|
|
|
|
__version__ = "0.5.2"
|