Commits

History for

Commits on May 14, 2024

Remove cases where the deprecated --xla_gpu_simplify_all_fp_conversions is set to its default value.

dimitar-asenov
authored and
maxtext authors
committedMay 14, 2024

Commits on May 10, 2024

Add forward pass logit check test for Llama2-7b
A9isha
committedMay 10, 2024
Merge pull request #641 from google:prefill-optimize
maxtext authors
committedMay 10, 2024
Merge pull request #644 from google:moe_param
maxtext authors
committedMay 10, 2024
Update llama2_7b_gpu.yml
maxtext authors
committedMay 10, 2024

Commits on May 9, 2024

fix sharding on generate cache in prefill results.
jwyang-google
committedMay 9, 2024
Update tflops calculation
RissyRan
committedMay 9, 2024

Commits on May 7, 2024

Merge pull request #635 from google:mattdavidow-numpy-prod
maxtext authors
committedMay 7, 2024
Reverts e8b53e5

golechwierowicz
authored and
maxtext authors
committedMay 7, 2024

Commits on May 6, 2024

Make some AQT dataclasses to use keyword-only fields (1/N)
maxtext authors
committedMay 6, 2024

Commits on May 3, 2024

Internal change only.
maxtext authors
committedMay 3, 2024

Commits on May 2, 2024

fix norm sharding
ZhiyuLi-goog
committedMay 2, 2024

Commits on May 1, 2024

Replace deprecated np.product with np.prod
gobbleturk
committedMay 1, 2024

Commits on Apr 30, 2024

Merge pull request #631 from google:roshanin_entropy_2
maxtext authors
committedApr 30, 2024
Enable entropy on multihost CPUs.
RoshaniN
committedApr 30, 2024
loosen tolerance in assert_params_sufficiently_sharded
ZhiyuLi-goog
committedApr 30, 2024
Fix test_tokenize
khatwanimohit
committedApr 30, 2024

Commits on Apr 26, 2024

Change l2norm to use jnp.sqrt
raymondzouu
committedApr 26, 2024
change norm sharding
ZhiyuLi-goog
committedApr 26, 2024

Commits on Apr 25, 2024

Update jax.tree_map to jax.tree_util.tree_map
RissyRan
committedApr 25, 2024

Commits on Apr 24, 2024

[MaxText] Rename llama2_7b_single_host_gpu.yml to make it clear that it can be used for any number of host.

bixia1
authored and
maxtext authors
committedApr 24, 2024
Merge pull request #602 from google:mohit/codespell
maxtext authors
committedApr 24, 2024
Update 128B config on v5e to use qkv_proj_offloaded remat_policy
raymondzouu
committedApr 24, 2024

Commits on Apr 23, 2024

Pre-commit config
khatwanimohit
committedApr 23, 2024
Explicitly set AQT Freezer mode in MaxText.

cdh4696
authored and
maxtext authors
committedApr 23, 2024

Commits on Apr 20, 2024

inference_microbenchmark:
morgandu
committedApr 20, 2024

Commits on Apr 19, 2024

Merge pull request #600 from google:patemotter_attn_calc_fix
maxtext authors
committedApr 19, 2024
Merge branch 'main' into patemotter_attn_calc_fix
patemotter
committedApr 19, 2024
inference microbenchmark
morgandu
committedApr 19, 2024
Fix subset of hosts dataloading
khatwanimohit
committedApr 19, 2024

Commits on Apr 18, 2024

Fixes
patemotter
committedApr 18, 2024
Revert "Fixes for inf testing."
patemotter
committedApr 18, 2024
Fixes for inf testing.
patemotter
committedApr 18, 2024
Removes batch size from prefill attention calculation.
patemotter
committedApr 18, 2024

Commits on Apr 16, 2024

adding script to fix the style and adding modified/fixed files with line length 125
ssusie
committedApr 16, 2024