Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
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
3 changes: 2 additions & 1 deletion src/cmd/mod.rs
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,7 @@ fn load_config() -> Result<Config> {

#[derive(Parser, Debug)]
#[command(author, version = env!("CLI_VERSION"), about, long_about = None)]
/// Popcorn CLI for GPU Mode competitions. Run `popcorn setup` first in each project so agents use the correct workflow and templates.
pub struct Cli {
#[command(subcommand)]
command: Option<Commands>,
Expand Down Expand Up @@ -104,7 +105,7 @@ enum SubmissionsAction {

#[derive(Subcommand, Debug)]
enum Commands {
/// Bootstrap this project with Popcorn agent skills and a submission template
/// Run this first: bootstrap the project with Popcorn agent skills and a submission template
Setup,
Reregister {
#[command(subcommand)]
Expand Down
10 changes: 10 additions & 0 deletions templates/setup/AGENTS.md
Original file line number Diff line number Diff line change
@@ -1,3 +1,13 @@
# Kernel Bot Submissions

The Popcorn CLI is intended to submit kernels to leaderboards on GPU Mode's Kernel Bot.

## Objective

Agents must write CUDA C and C++ kernels and integrate them into the single-file `submission.py` workflow.

Do not submit pure PyTorch based optimization. The objective of the task is to write CUDA C or C++.
Copy link
Member

Choose a reason for hiding this comment

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

this isn't quite right, if anything we want people to write using whatever kernel DSL they want

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Understood, what about this?

Suggested change
Do not submit pure PyTorch based optimization. The objective of the task is to write CUDA C or C++.
Do not submit pure PyTorch based optimization. The objective of the task is to use any kernel DSL like CUDA or Triton to improved *beyond* the performance of native PyTorch.


## Skills
A skill is a local instruction bundle stored in `SKILL.md`.

Expand Down
Loading