Skip to content

Use all gpus available for training #293

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

Draft
wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

Use all gpus available for training #293

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
2 changes: 1 addition & 1 deletion finetune/adapter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@
save_interval = 1000
eval_iters = 100
log_interval = 1
devices = 1
devices = torch.cuda.device_count()

# Hyperparameters
learning_rate = 9e-3
Expand Down
2 changes: 1 addition & 1 deletion finetune/full.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@
save_interval = 1000
eval_iters = 100
log_interval = 100
devices = 4
devices = torch.cuda.device_count()
Copy link
Contributor

Choose a reason for hiding this comment

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

Suggested change
devices = torch.cuda.device_count()
devices = "auto"


# Hyperparameters
learning_rate = 3e-5
Expand Down
3 changes: 2 additions & 1 deletion finetune/lora.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,7 +50,8 @@ def main(
out_dir: str = "out/lora/alpaca",
):

fabric = L.Fabric(accelerator="cuda", devices=1, precision="bf16-true")
devices = torch.cuda.device_count()
fabric = L.Fabric(accelerator="cuda", devices=devices, precision="bf16-true")
Comment on lines +53 to +54
Copy link
Contributor

Choose a reason for hiding this comment

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

Suggested change
devices = torch.cuda.device_count()
fabric = L.Fabric(accelerator="cuda", devices=devices, precision="bf16-true")
fabric = L.Fabric(accelerator="cuda", devices="auto", precision="bf16-true")

Copy link
Contributor

Choose a reason for hiding this comment

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

@Abecid Did multi-gpu training work with this script and how many did you use? Is the loss convergence comparable to single gpu training?

Copy link

Choose a reason for hiding this comment

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

i ran this with two NVIDIA Tesla T4 cards with devices="auto" and got this error:

RuntimeError: [1] is setting up NCCL communicator and retrieving ncclUniqueId from [0] via c10d key-value store by key '0', but store->get('0') got error: Broken pipe. This may indicate a possible application crash on rank 0 or a network set up issue.

fabric.launch()
fabric.seed_everything(1337 + fabric.global_rank)

Expand Down