Subnet 1 Current Competitions

Registry of competitions currently active on SN1 APEX.

Matrix Compression

Battleship

1. Matrix Compression

The first competition - Matrix Compression - explores how small neural activations - both forward and backward - can be compressed while still retaining all their original information. Reducing activation size enables faster data transfer across the internet, a crucial step toward making distributed training more efficient, as it’s often constrained by network bandwidth. The top-performing algorithms from this competition will be integrated to enhance training on subnet 9 IOTA.

Competition Dashboard

Evaluation

Miners aim to optimize the following:

• Compression Ratio - How small the compressed solution is on disk versus the starting matrix.

• Time - How fast the compression/decompression algorithm runs.

To surpass the current winner of the competition, a miner must earn a score of at least 1% higher than the current top score. If there is no current winner, then a miner must earn a score of at least 1% higher than the baseline score.

• The score_to_beat is displayed in the Apex CLI dashboard, under competition information.

Score is calculated by:

score = np.clip((1 - compression) * (1 - task_time / (1 + 0.012)), 0.0, 1.0)

• Where task_time includes both compression and decompression.

• Compression is calculated by compressed_file_size / original_file_size.

• If compression is not lossless, the submission will receive a score of 0.

• Only top evaluated submission is re-evaluated the next round.

• Previous top score is not counting towards the neww top score.

Matrix Compression Settings

• The submitted code remains hidden for 24 hours to everyone but the submission owner, after which it becomes accessible to all miners.

• Round length: 2 days

• Burn rate: 90%

  • The top scorer receives the remaining incentive pool (10%), decreasing linearly over a 10 day period, provided no other submissions surpass the current top score.

    • If a new submission exceeds the top score, the incentive decay resets from 10%.

• Evaluation: each round uses 400 randomly sampled matrices from the total pool; the sample refreshes every round.

• Submission Constraints: Your submission must be less than 20KB.

• Multiple submissions:

  • The rate limit is 4 submissions in 24 hours per miner across all the competitions - Matrix Compression and Battleship. The period is individual for each user and starts to count from the first submission.

For Miners

View the matrix compression baseline miner solution provided as an example.

The manifest.json file contains a sample of 1000 matrices of a large matrix pool, including both matrices that have already been used for evaluation and matrices that have not been evaluated.

Note: The R2 bucket does not list files directly. Use the manifest to find and download samples.

1. Download the manifest file. This file contains the full list of available sample filenames.

2. Open the manifest and copy the name of any file you want to download. For example: example_001.txt.

3. In your browser (or with curl/wget), go to: https://pub-77097c3387c340de9ff1bd5e5b443d8d.r2.dev/<file_name> replacing <file_name> with the name you copied from the manifest. For example: https://pub-77097c3387c340de9ff1bd5e5b443d8d.r2.dev/example_001.txt.

Then, continue to the Apex CLI guide to submit a solution.

The information about enabled packages is in requirements.txt.

2. Battleship

This is SN1 Festive Competition Launch - a special experiment - the Battleship Competition!

Battleship Settings

Competition Dashboard

Match Structure

• A single match consists of several Battleship games played by 1 miner.

• In each game, the miner receives a hidden ship board - a unique configuration of ships that is known only to the orchestrator.

• Their task is to determine an optimal strategy to locate and hit the opponent’s ships as efficiently as possible, while working under an unknown turn constraint

• The miner that solves the most boards the fastest wins and receives all competition emissions, annealing with the burn.

• Shots may not be repeated!

Evaluation

Game Score

Every game produces a score based on two components:

1. Win Score

• Winning a game grants 1000 points.

2. Speed Bonus

• A speed bonus is added based on how quickly the miner wins: (100 − number_of_turns_to_win) × 0.1

• This rewards faster solutions.

• Important rule: If a game ends in fewer than 10 turns, we assume it failed to run correctly → no speed bonus is awarded. * One turn - is one shot made to the ship board.

Score Limits

• Maximum speed bonus: 9 - Received when the miner wins in 10 turns

• Maximum possible score per game: 1000 + 9 = 1009

Final Competition Score

• The miner’s final score is the average of all their game scores across the round.

Additional details:

• Incentive mechanism is standard for the Subnet 1

  • Miners code reveals 1 day after the evaluation.

  • Logs are opened after the round completion.

  • Emission burning for the top miner is active.

• Multiple submissions:

  • The rate limit is 4 submissions in 24 hours per miner across all the competitions - Battleship and Matrix Compression. The period is individual for each user and starts to count from the first submission.

  • If a miner makes multiple submissions to Battleship within a single round, only the latest version participates in the battle evaluation.

• Base miner example can be found at baseline.py.

• The information about enabled packages is in requirements.txt.

• All matches produce a replay file, with View only access.

• Local testing instructions can be found here.