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

fix bugs in allgather #314

Merged
merged 18 commits into from
Oct 18, 2024
Merged

fix bugs in allgather #314

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
18 commits
Select commit Hold shift + click to select a range
1bd7a69
feat: Add github workflow
xibosun Sep 13, 2024
99f09df
feat: add a docker for github action
xibosun Sep 14, 2024
211535a
feat: pre-install dependencies in docker
xibosun Sep 15, 2024
6c3ae75
fix: fix typos and rename containers
xibosun Sep 16, 2024
040bbda
Merge branch 'xdit-project:main' into 0916
xibosun Sep 18, 2024
9792700
fix: change a runner machine
xibosun Sep 18, 2024
356b6dd
Merge branch 'xdit-project:main' into cogvideo
xibosun Sep 20, 2024
e2a1f6c
fix: use xFuserJointLongContextAttention as the underlying SP method …
xibosun Sep 20, 2024
4e28310
fix: update the comment format and update github actions
xibosun Sep 23, 2024
b59564d
Merge branch 'main' into cogvideo
xibosun Sep 23, 2024
afea6de
fix: patch embedding for CogVideo
xibosun Sep 23, 2024
62a3ed3
feat: add cfg support to CogVideo
xibosun Sep 23, 2024
80f4e53
merge to main
xibosun Oct 14, 2024
4ab5d36
support VAE tiling/slicing and sequential/model offloading in CogVideoX
xibosun Oct 14, 2024
a298175
fix xFuserArgs members
xibosun Oct 14, 2024
24ac8df
Merge remote-tracking branch 'upstream/main' into cogvideo
xibosun Oct 14, 2024
8dd6f17
Merge remote-tracking branch 'upstream/main' into cogvideo
xibosun Oct 18, 2024
70fdd63
fix bugs in allgather
xibosun Oct 18, 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
6 changes: 3 additions & 3 deletions examples/run_cogvideo.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,9 +14,9 @@ mkdir -p ./results
TASK_ARGS="--height 480 --width 720 --num_frames 9"

# CogVideoX parallel configuration
N_GPUS=4
PARALLEL_ARGS="--ulysses_degree 2 --ring_degree 1"
CFG_ARGS="--use_cfg_parallel"
N_GPUS=6
PARALLEL_ARGS="--ulysses_degree 2 --ring_degree 3"
#CFG_ARGS="--use_cfg_parallel"

# Uncomment and modify these as needed
# PIPEFUSION_ARGS="--num_pipeline_patch 8"
Expand Down
17 changes: 1 addition & 16 deletions xfuser/model_executor/pipelines/pipeline_cogvideox.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,22 +343,7 @@ def __call__(
progress_bar.update()

if get_sequence_parallel_world_size() > 1:
sp_degree = get_sequence_parallel_world_size()
sp_latents_list = get_sp_group().all_gather(latents, separate_tensors=True)
latents_list = []
for pp_patch_idx in range(get_runtime_state().num_pipeline_patch):
latents_list += [
sp_latents_list[sp_patch_idx][
:,
:,
get_runtime_state()
.pp_patches_start_idx_local[pp_patch_idx] : get_runtime_state()
.pp_patches_start_idx_local[pp_patch_idx + 1],
:,
]
for sp_patch_idx in range(sp_degree)
]
latents = torch.cat(latents_list, dim=-2)
latents = get_sp_group().all_gather(latents, dim=-2)

if is_dp_last_group():
if not (output_type == "latents" or output_type == "latent"):
Expand Down