| { | |
| "arch": { | |
| "architectures": [ | |
| "ScriptableMaskedLM" | |
| ], | |
| "attention": { | |
| "causal_attention": false, | |
| "dropout_prob": 0.1, | |
| "high_level_fusion": false, | |
| "low_level_fusion": true, | |
| "num_attention_heads": 12, | |
| "qkv_bias": false, | |
| "rotary_embedding": false, | |
| "seq_op_in_fp32": false, | |
| "sequence_op": "torch-softmax", | |
| "skip_output_projection": false, | |
| "type": "flash-attention-impl" | |
| }, | |
| "classification_head": { | |
| "classifier_dropout": 0.1, | |
| "head_dim": 1024, | |
| "include_ff_layer": true, | |
| "nonlin": "Tanh", | |
| "pooler": "zero_index" | |
| }, | |
| "decoder_bias": false, | |
| "deepnorm_scaling": false, | |
| "embedding": { | |
| "dropout_prob": 0.1, | |
| "embedding_dim": 768, | |
| "max_seq_length": 128, | |
| "normalization": true, | |
| "pad_token_id": 0, | |
| "pos_embedding": "learned", | |
| "vocab_size": 32768 | |
| }, | |
| "ffn_layer_frequency": 1, | |
| "final_norm": true, | |
| "gradient_checkpointing": false, | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 768, | |
| "init": { | |
| "std": 0.02, | |
| "type": "normal" | |
| }, | |
| "intermed_size": 3072, | |
| "layer_drop_theta": null, | |
| "layer_fusion": true, | |
| "layer_macro_type": "transformer", | |
| "loss": "cross-entropy", | |
| "nonlin": "GELUglu", | |
| "norm": "LayerNorm", | |
| "norm_eps": 1e-12, | |
| "norm_scheme": "pre", | |
| "num_labels": null, | |
| "num_transformer_layers": 16, | |
| "recurrent_layers": null, | |
| "skip_head_transform": true, | |
| "sparse_prediction": true, | |
| "tie_weights": true, | |
| "use_bias": false, | |
| "z_loss_factor": 0 | |
| }, | |
| "architectures": [ | |
| "ScriptableLMForPreTraining" | |
| ], | |
| "model_type": "crammedBERT", | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.26.1" | |
| } | |