| { |
| "config": { |
| "model": { |
| "model_dim": 128, |
| "attention_heads": 4, |
| "ffn_dim": 512, |
| "base_attention_radii": [ |
| 32, |
| 128 |
| ], |
| "segment_layers": 6, |
| "base_embedding_dim": 16, |
| "dropout": 0.1, |
| "max_segment_length": 512, |
| "rope_theta": 10000.0, |
| "rope_position_scale": 5.0 |
| }, |
| "data": { |
| "genome_windows": "data/GRCh38.p14.acgt.w4096.s4096", |
| "context_length": 4096, |
| "min_context_length": 256, |
| "crops_per_read": 2, |
| "train_sources": [ |
| "NC_000003.12", |
| "NC_000004.12", |
| "NC_000005.10", |
| "NC_000006.12", |
| "NC_000007.14", |
| "NC_000008.11", |
| "NC_000009.12", |
| "NC_000010.11", |
| "NC_000011.10", |
| "NC_000012.12", |
| "NC_000013.11", |
| "NC_000014.9", |
| "NC_000015.10", |
| "NC_000016.10", |
| "NC_000017.11", |
| "NC_000018.10", |
| "NC_000019.10", |
| "NC_000020.11", |
| "NC_000021.9", |
| "NC_000022.11" |
| ], |
| "validation_sources": [ |
| "NC_000002.12" |
| ], |
| "test_sources": [ |
| "NC_000001.11" |
| ], |
| "target_margin": 32, |
| "mask_probability": 0.15, |
| "mask_replacement_probability": 0.8, |
| "random_replacement_probability": 0.1, |
| "segmenter": "lcp", |
| "lcp_level": 2, |
| "lcp_library": "build/lcptools/lib/liblcptools.so", |
| "fixed_length": 10, |
| "fixed_stride": 5, |
| "bpe_tokenizer": "data/tokenizers/bpe-4096.json", |
| "example_workers": 3 |
| }, |
| "training": { |
| "run_name": "locus-2m", |
| "seed": 42, |
| "batch_size": 16, |
| "max_steps": 359466, |
| "learning_rate": 0.0006, |
| "warmup_steps": 3600, |
| "weight_decay": 0.01, |
| "adam_beta1": 0.9, |
| "adam_beta2": 0.98, |
| "adam_epsilon": 1e-06, |
| "gradient_clip_norm": 1.0, |
| "shuffle_buffer_size": 512, |
| "mixed_precision": true, |
| "device": "cuda", |
| "log_every": 10, |
| "checkpoint_every": 1000, |
| "keep_checkpoint_every": 10000, |
| "keep_checkpoint_steps": [] |
| } |
| }, |
| "step": 359466 |
| } |
|
|
