Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Perf megablox #694

Draft
wants to merge 17 commits into
base: main
Choose a base branch
from
Draft

Perf megablox #694

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
17 commits
Select commit Hold shift + click to select a range
662d778
Add vanilla megablox to MoE
RissyRan Jun 5, 2024
8098e13
improve perf
RissyRan Jun 9, 2024
cc43d18
Update gates sharding
RissyRan Jun 10, 2024
f51a781
test on v5p
RissyRan Jun 10, 2024
43c06bd
Merge branch 'main' into perf_megablox
RissyRan Jun 11, 2024
115b399
perf_debug
RissyRan Jun 12, 2024
5c1d1bb
check unit test and end-to-end
RissyRan Jun 12, 2024
120ec39
06_12
RissyRan Jun 12, 2024
7bf2f4a
06_12
RissyRan Jun 13, 2024
0ceeba2
clean_up
RissyRan Jun 14, 2024
d9db136
scale up
RissyRan Jun 17, 2024
5823a14
MantaRay
RissyRan Jun 18, 2024
990684a
1t scale up
RissyRan Jun 21, 2024
d520ced
Add num_exp config
RissyRan Jun 22, 2024
9dd62c6
global scale
RissyRan Jun 24, 2024
d26005f
xprof
RissyRan Jun 26, 2024
d483ad6
arithmetic
RissyRan Jul 1, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
MantaRay
  • Loading branch information
@RissyRan
RissyRan committed Jun 19, 2024
commit 5823a14a4617ec6c2c1342acae42da39c9b82a4b
4 changes: 2 additions & 2 deletions MaxText/layers/linears.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -375,8 +375,8 @@ def unpermute(self, intermediate, sorted_selected_experts, weights):

def call_gmm(self, inputs, gate_logits, config, w0_kernel, w1_kernel, wo_kernel):
# TODO(ranran): update the static default tile_size
tile_size = (512, 512, 512)
# tile_size = None
# tile_size = (512, 512, 512)
tile_size = None
# replicated_sharding = jax.sharding.NamedSharding(self.mesh, PartitionSpec(None))

def gmm(inputs, kernel, group_sizes):
Expand Down
2 changes: 2 additions & 0 deletions requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,3 +31,5 @@ transformers
git+https://github.com/mlperf/logging.git
google-jetstream
jsonlines
google-cloud-aiplatform==1.50.0
shapely==1.8.5.post1
Loading