arch_budget_hours_40.json

{
    "architectures": [
        "ScriptableMaskedLM"
    ],
    "num_transformer_layers": 16,
    "hidden_size": 768,
    "intermed_size": 3072,
    "hidden_dropout_prob": 0.1,
    "norm": "LayerNorm",
    "norm_eps": 1e-12,
    "norm_scheme": "pre",
    "nonlin": "GELUglu",
    "tie_weights": true,
    "sparse_prediction": true,
    "decoder_bias": false,
    "loss": "cross-entropy",
    "z_loss_factor": 0,
    "gradient_checkpointing": false,
    "layer_fusion": true,
    "embedding": {
        "vocab_size": null,
        "pos_embedding": "scaled-sinusoidal",
        "dropout_prob": 0.1,
        "pad_token_id": 0,
        "max_seq_length": 128,
        "embedding_dim": 768,
        "normalization": true
    },
    "attention": {
        "type": "flash-attention-impl",
        "causal_attention": false,
        "num_attention_heads": 12,
        "dropout_prob": 0.1,
        "skip_output_projection": false,
        "qkv_bias": false,
        "rotary_embedding": false,
        "seq_op_in_fp32": false,
        "sequence_op": "torch-softmax",
        "high_level_fusion": false,
        "low_level_fusion": true
    },
    "init": {
        "type": "normal",
        "std": 0.02
    },
    "ffn_layer_frequency": 1,
    "deepnorm_scaling": false,
    "skip_head_transform": true,
    "layer_drop_theta": null,
    "use_bias": false,
    "final_norm": true,
    "recurrent_layers": null,
    "layer_macro_type": "transformer",
    "num_labels": null,
    "classification_head": {
        "pooler": "avg",
        "include_ff_layer": true,
        "head_dim": 1024,
        "nonlin": "Tanh",
        "classifier_dropout": 0.1
    }
}