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

stage first axes mesh #886

Closed
wants to merge 1 commit into from
Closed

stage first axes mesh #886

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
4 changes: 2 additions & 2 deletions MaxText/configs/base.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -181,7 +181,7 @@ jax_cache_dir: "~/jax_cache"
hardware: 'tpu' # Supported hardware types are 'tpu', 'gpu', 'gpu_multiprocess' and 'cpu'

# Parallelism
mesh_axes: ['data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'tensor', 'expert', 'autoregressive']
mesh_axes: ['stage', 'data', 'fsdp', 'fsdp_transpose', 'sequence', 'tensor', 'expert', 'autoregressive']
logical_axis_rules: [
['activation_batch', ['data', 'fsdp', 'fsdp_transpose', 'expert']],
['activation_batch_no_exp', ['data', 'fsdp', 'fsdp_transpose']],
Expand Down Expand Up @@ -221,7 +221,7 @@ logical_axis_rules: [
['exp', 'expert'],
]
# Axes used for DCN must be earlier in this list than ICI, see (b/339009148) for details
data_sharding: [['data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'tensor', 'expert', 'autoregressive']]
data_sharding: [['stage', 'data', 'fsdp', 'fsdp_transpose', 'sequence', 'tensor', 'expert', 'autoregressive']]

# One axis for each parallelism type may hold a placeholder (-1)
# value to auto-shard based on available slices and devices.
Expand Down
4 changes: 2 additions & 2 deletions MaxText/max_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -385,8 +385,8 @@ def create_device_mesh(config, devices=None):
multi_slice_env = num_slices > 1

dcn_parallelism = [
config.dcn_data_parallelism,
config.dcn_pipeline_parallelism,
config.dcn_data_parallelism,
config.dcn_fsdp_parallelism,
config.dcn_fsdp_transpose_parallelism,
config.dcn_sequence_parallelism,
Expand All @@ -395,8 +395,8 @@ def create_device_mesh(config, devices=None):
config.dcn_autoregressive_parallelism,
]
ici_parallelism = [
config.ici_data_parallelism,
config.ici_pipeline_parallelism,
config.ici_data_parallelism,
config.ici_fsdp_parallelism,
config.ici_fsdp_transpose_parallelism,
config.ici_sequence_parallelism,
Expand Down
Loading