Release Notes – Release 1.6¶

Key Features and Enhancements¶

[pyTorch] Added a new make_graphed_callables API call for NVIDIA^® CUDA^® graph capture, including FP8 support.
[pyTorch] Added beta support for two boolean arguments in the DelayedScaling FP8 recipe (fp8_dpa and fp8_mha) to support FP8 attention. Note that the API exposure of this feature may change in future releases.

[pyTorch] Fixed a numerical issue with storing weights in FP8 via the fp8_model_init API call.
[pyTorch] Fixed a bug that caused PyTorch modules to use excessive memory when training with frozen weights by storing unnecessary activations for the backward pass.
[JAX] Fixed a bug that caused an incorrect shape to be passed for layernorm gradient.

These issues are unchanged from the previous release.

FlashAttention v2, which is a dependency of this release of Transformer Engine, has a known issue with excessive memory usage during installation (https://github.com/Dao-AILab/flash-attention/issues/358). You can work around this issue by setting the environment variable MAX_JOBS=1 during Transformer Engine installation.
[pyTorch] FlashAttention v2.1 changed the behavior of the causal mask when performing cross-attention (see https://github.com/Dao-AILab/flash-attention#21-change-behavior-of-causal-flag for reference). In order for Transformer Engine to keep consistent behavior between versions and backends, FlashAttention is disabled for this use case (cross attention with casual masking) when 2.1+ version of FlashAttention is installed.

There are no breaking changes in this release.

There are no deprecated features in this release.