Skip to content

Optimizing flux2klein #458

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

Open
amepas wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Optimizing flux2klein #458

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bb40f15
feat(flux2klein): Onboard FLUX.2-Klein 4B and 9B model family with fa…
amepas Aug 10, 2026
d78a39e
perf(flux2klein): FLUX.2-Klein 9B end-to-end performance optimization…
amepas Aug 11, 2026
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
20 changes: 16 additions & 4 deletions src/maxdiffusion/configs/base_flux2klein.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,6 +75,17 @@ mask_padding_tokens: True
attention_sharding_uniform: True

flash_block_sizes: {}
block_q: 0

@eltsai eltsai Aug 11, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe set these variable using some runnable values (like the optimal ones you found)

block_kv: 0
block_kv_compute: 0
ulysses_shards: -1
ulysses_attention_chunks: 1
text_encoder_attention: 'dot_product'
text_encoder_flash_block_sizes: {}
text_encoder_block_q: 0
text_encoder_block_kv: 0
text_encoder_block_kv_compute: 0
text_encoder_max_layer: 27
# GroupNorm groups
norm_num_groups: 32

Expand Down Expand Up @@ -154,12 +165,12 @@ data_sharding: [['data', 'fsdp', 'context', 'tensor']]
# value to auto-shard based on available slices and devices.
# By default, product of the DCN axes should equal number of slices
# and product of the ICI axes should equal number of devices per slice.
dcn_data_parallelism: 1 # recommended DCN axis to be auto-sharded
dcn_fsdp_parallelism: -1
dcn_data_parallelism: 1
dcn_fsdp_parallelism: 1
dcn_context_parallelism: 1
dcn_tensor_parallelism: 1
ici_data_parallelism: 1
ici_fsdp_parallelism: -1
ici_fsdp_parallelism: 1
ici_context_parallelism: 1
ici_tensor_parallelism: 1

Expand Down Expand Up @@ -203,7 +214,7 @@ num_train_epochs: 1
seed: 0
output_dir: 'output/'
output_name: "flux2klein_generated_image.png"
per_device_batch_size: 1
per_device_batch_size: 1.0

warmup_steps_fraction: 0.1
learning_rate_schedule_steps: -1 # By default the length of the schedule is set to the number of steps.
Expand Down Expand Up @@ -231,6 +242,7 @@ do_classifier_free_guidance: True
guidance_scale: 4.0
guidance_rescale: 0.0
num_inference_steps: 4
num_reps: 1
save_final_checkpoint: False

# SDXL Lightning parameters
Expand Down
20 changes: 16 additions & 4 deletions src/maxdiffusion/configs/base_flux2klein_9B.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,6 +75,17 @@ mask_padding_tokens: True
attention_sharding_uniform: True

flash_block_sizes: {}
block_q: 0
block_kv: 0
block_kv_compute: 0
ulysses_shards: -1
ulysses_attention_chunks: 1
text_encoder_attention: 'dot_product'
text_encoder_flash_block_sizes: {}
text_encoder_block_q: 0
text_encoder_block_kv: 0
text_encoder_block_kv_compute: 0
text_encoder_max_layer: 27
# GroupNorm groups
norm_num_groups: 32

Expand Down Expand Up @@ -154,12 +165,12 @@ data_sharding: [['data', 'fsdp', 'context', 'tensor']]
# value to auto-shard based on available slices and devices.
# By default, product of the DCN axes should equal number of slices
# and product of the ICI axes should equal number of devices per slice.
dcn_data_parallelism: 1 # recommended DCN axis to be auto-sharded
dcn_fsdp_parallelism: -1
dcn_data_parallelism: 1
dcn_fsdp_parallelism: 1
dcn_context_parallelism: 1
dcn_tensor_parallelism: 1
ici_data_parallelism: 1
ici_fsdp_parallelism: -1 # recommended ICI axis to be auto-sharded
ici_fsdp_parallelism: 1
ici_context_parallelism: 1
ici_tensor_parallelism: 1

Expand Down Expand Up @@ -203,7 +214,7 @@ num_train_epochs: 1
seed: 0
output_dir: 'output/'
output_name: "flux2klein_generated_image.png"
per_device_batch_size: 1
per_device_batch_size: 1.0

warmup_steps_fraction: 0.1
learning_rate_schedule_steps: -1 # By default the length of the schedule is set to the number of steps.
Expand Down Expand Up @@ -231,6 +242,7 @@ do_classifier_free_guidance: True
guidance_scale: 4.0
guidance_rescale: 0.0
num_inference_steps: 4
num_reps: 1
save_final_checkpoint: False

# SDXL Lightning parameters
Expand Down
Loading
Loading