Search results

Is there an example of training LLM using praxis?

Status: Open.

#87 In google/praxis;

· knightXun opened on Dec 17, 2024

Support for loading fp8 checkpoint

Status: Open.

#68 In google/praxis;

· wenscarl opened on May 21, 2024

Add activation checkpoint offloading

Status: Open.

#63 In google/praxis;

· jon-chuang opened on May 6, 2024

Cross-layer attention weight sharing fails in different scopes

Status: Open.

#52 In google/praxis;

· mqyqlx opened on Mar 19, 2024

Incorrect conversion from tf dtype to jax dtype

Status: Open.

#49 In google/praxis;

· backpropper opened on Feb 2, 2024

[Feature Request] Need Matmul Attention layer instead of Einsum to support GPU running

Status: Open.

#46 In google/praxis;

· MoFHeka opened on Jan 27, 2024

[Feature Request] Need ZeRo-1/2 to cooperate with PP+TP+DP. Which may more faster than FSDP sometimes.

Status: Open.

#45 In google/praxis;

· MoFHeka opened on Jan 26, 2024

gpu_fast_attention not passing segment_ids to jax pallas attention mha

Status: Open.

#30 In google/praxis;

· Cjkkkk opened on Oct 2, 2023

Any publicly available document?

Status: Open.

#7 In google/praxis;

· Lekja00160612 opened on Mar 15, 2023