Current Competitions

Subnet 1 Apex competitions registry

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.

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

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

Raw score is calculated by:

raw_score = 1 * (30 - task_time) + 1000 * similarity + 1000000 * compression_ratio

  • Where task_time includes both compression and decompression.

  • Similarity is calculated by cosine_similarity * norm_similarity.

  • Compression ratio is calculated by original_size / compressed_file_size.

This raw score is then normalized with a sigmoid normalization function:

normalized_score = 1 / (1 + np.exp(-0.003 * (raw_score / 1000 - 2000)))

  • Where a normalized score of 0.5 correlates to a raw score of 2,000,000.

For Miners

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

View the matrix compression general miner solution template.

Test your solution with the provided samples in this R2 bucket: https://pub-77097c3387c340de9ff1bd5e5b443d8d.r2.dev. Files are listed in manifest.csv

  • First, download manifest.csv from the R2 bucket to access all sample file names.

  • You will then be able to access all sample files with their given file names.

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

Last updated