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

Deepspeed benchmarking #878

Draft
wants to merge 24 commits into
base: main
Choose a base branch
from
Draft

Deepspeed benchmarking #878

Changes from 1 commit
Commits
Show all changes
24 commits
Select commit Hold shift + click to select a range
f4706e0
add flash_attn_kvpacked
satpalsr Mar 29, 2023
f4a9106
Changed is_pipe_parallel setting to fix pipeline-parallel inference
curt-tigges Mar 31, 2023
83a7b9a
Update NeoXArgs docs automatically
invalid-email-address Mar 31, 2023
45d7052
fix formatting
satpalsr Apr 11, 2023
857c556
gpt benchmark script
cr458 Apr 3, 2023
1ab5bf3
remove duplicate argparse
cr458 Apr 4, 2023
afb6b29
HF inference
cr458 Apr 4, 2023
3f7d605
benchmarking configs + script changes
cr458 Apr 11, 2023
d99d2ce
plot directly, runs deepspeed and hf for single benchmark
cr458 Apr 12, 2023
b0e9745
remove plotting comments
cr458 Apr 12, 2023
9c645dd
accept changes from main & resolve conflicts
satpalsr Apr 15, 2023
ee99945
Merge branch 'main' into flash_attn_infer
satpalsr Apr 15, 2023
9b1733e
tmp changes
cr458 Apr 17, 2023
22cac56
Merge remote-tracking branch 'satpalsr/flash_attn_infer' into deepspe…
cr458 Apr 17, 2023
466749b
merge conflict git hash
cr458 Apr 17, 2023
b10739f
separate scripts for Deepspeed/HF and neox
cr458 Apr 18, 2023
4990f9b
debugging: works when world size > 1 but not otherwise
cr458 Apr 18, 2023
88981b2
working ( but not serially)
cr458 Apr 19, 2023
5e3ca7f
working ish gpt-neox just need to figure out how to get dataframe back
cr458 Apr 20, 2023
3ee9d3b
get dataframe output from stdout
cr458 Apr 20, 2023
2a6e8cd
remove gpt neox inference from script
cr458 May 21, 2023
7ea22d9
remove lines
cr458 May 21, 2023
ef4fdd4
device error
cr458 May 21, 2023
d8184f3
Add DS inference
satpalsr May 22, 2023
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
Changed is_pipe_parallel setting to fix pipeline-parallel inference
  • Loading branch information
@curt-tigges
curt-tigges committed Mar 31, 2023
commit f4a9106160f64701780144dba81ea2a472720c41
2 changes: 1 addition & 1 deletion megatron/neox_arguments/arguments.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -989,7 +989,7 @@ def calculate_derived(self):
# Update 'is pipe parallel' flag
# if we set pipe_parallel_size to 0 or 1, GPT2ModelPipe.to_sequential() is called, and we run training with
# the sequential model without the PipelineModule wrapper to avoid the overhead it incurs
self.update_value("is_pipe_parallel", self.pipe_parallel_size >= 2)
self.update_value("is_pipe_parallel", self.pipe_parallel_size >= 1)

# Attention config
if self.attention_config is None:
Expand Down