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

set logging_step to 5 with baichuan && qwen benchmark #8928

Merged
merged 4 commits into from
Aug 14, 2024
Merged

set logging_step to 5 with baichuan && qwen benchmark #8928

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
45d6f31
set logging_step to 5 with baichuan && qwen benchmark
deepllz Aug 14, 2024
cde3764
Merge remote-tracking branch 'upstream/develop' into develop
deepllz Aug 14, 2024
e4deb07
polish
deepllz Aug 14, 2024
a34ea60
polish
deepllz Aug 14, 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
4 changes: 2 additions & 2 deletions ...ph/hybrid_parallelism/baichun2/auto_config_baichun2_13b/pretrain-baichun2_13b-config.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,12 +26,12 @@
"scale_loss": 1024,
"learning_rate": 1e-05,
"min_learning_rate": 5e-06,
"max_steps": 100,
"max_steps": 200,
"save_steps": 5000,
"weight_decay": 0.01,
"warmup_ratio": 0.01,
"max_grad_norm": 1.0,
"logging_steps": 1,
"logging_steps": 5,
Copy link
Contributor

Choose a reason for hiding this comment

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

这个max_steps是100，这里logging_steps设置成 5，打印的ips不够用

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Done

"dataloader_num_workers": 1,
"eval_steps": 1000,
"disable_tqdm": true,
Expand Down
2 changes: 1 addition & 1 deletion ...raph/hybrid_parallelism/baichun2/auto_config_baichun2_7b/pretrain-baichun2_7b-config.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@
"weight_decay": 0.01,
"warmup_ratio": 0.01,
"max_grad_norm": 1.0,
"logging_steps": 1,
"logging_steps": 5,
Copy link
Contributor

Choose a reason for hiding this comment

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

这个max_steps是100，这里logging_steps设置成 5，打印的ips不够用

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Done

"dataloader_num_workers": 1,
"eval_steps": 1000,
"disable_tqdm": true,
Expand Down
3 changes: 2 additions & 1 deletion ...sm/qwen/N4C32/qwen-qwen-14b_seqlen4096_pretrain_bs32_bf16_MP2-PP1-sharding16-mbs1-acc2.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,8 @@ param+="run_mode=MP2-PP1-sharding16-mbs1-acc2 "
param+="device_num=N4C32 "
param+="global_batch_size=32 "
param+="model_item=qwen-qwen-14b_seqlen4096_pretrain "
param+="max_steps=100 "
param+="max_steps=200 "
param+="logging_steps=5 "
Copy link
Contributor

Choose a reason for hiding this comment

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

这个max_steps是100，这里logging_steps设置成 5，打印的ips不够用

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Done

param+="gradient_accumulation_steps=2 "
param+="pp_recompute_interval=1 "
param+="tensor_parallel_config=enable_mp_async_allreduce,enable_mp_skip_c_identity,enable_mp_fused_linear_param_grad_add, "
Expand Down
3 changes: 2 additions & 1 deletion ...ism/qwen/N4C32/qwen-qwen-7b_seqlen4096_pretrain_bs32_bf16_MP1-PP1-sharding32-mbs1-acc1.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,8 @@ param+="run_mode=MP1-PP1-sharding32-mbs1-acc1 "
param+="device_num=N4C32 "
param+="global_batch_size=32 "
param+="model_item=qwen-qwen-7b_seqlen4096_pretrain "
param+="max_steps=100 "
param+="max_steps=200 "
param+="logging_steps=5 "
Copy link
Contributor

Choose a reason for hiding this comment

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

这个max_steps是100，这里logging_steps设置成 5，打印的ips不够用

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Done

param+="gradient_accumulation_steps=1 "
param+="pp_recompute_interval=1 "
param+="tensor_parallel_config=enable_mp_async_allreduce,enable_mp_skip_c_identity,enable_mp_fused_linear_param_grad_add "
Expand Down
3 changes: 2 additions & 1 deletion tests/test_tipc/dygraph/hybrid_parallelism/qwen/benchmark_common/run_benchmark.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,7 @@ function _set_params(){
global_batch_size=${global_batch_size:-16}
model_item=${model_item:-"qwen-qwen-7b_seqlen2048_pretrain"}
max_steps=${max_steps:-150}
logging_steps=${logging_steps:-1}
gradient_accumulation_steps=${gradient_accumulation_steps:-8}
pp_recompute_interval=${pp_recompute_interval:-1}
tensor_parallel_config=${tensor_parallel_config:-"enable_mp_async_allreduce,enable_mp_skip_c_identity,enable_mp_fused_linear_param_grad_add"}
Expand Down Expand Up @@ -162,7 +163,7 @@ function _train(){
--weight_decay 0.01 \
--warmup_ratio 0.01 \
--max_grad_norm 1.0 \
--logging_steps 1 \
--logging_steps ${logging_steps} \
--dataloader_num_workers 1 \
--eval_steps ${eval_steps} \
--sharding ${sharding} \
Expand Down
Loading