| { | |
| "architecture": "mine-embed-57M (encoder-only, pre-norm, GELU, mean-pool, L2-norm)", | |
| "encoder_config": { | |
| "vocab": 50258, | |
| "hidden": 512, | |
| "heads": 8, | |
| "ffn": 2048, | |
| "layers": 10, | |
| "max_pos": 1024, | |
| "dropout": 0.0 | |
| }, | |
| "pooling": "mean", | |
| "normalize": true, | |
| "max_seq_len_trained": 128, | |
| "tokenizer": "openai-community/gpt2 (BPE, pad=EOT 50256, mask id 50257)", | |
| "params_M": 57.781248 | |
| } |