|
{ |
|
"_name_or_path": "omarmomen/sf_babylm_1", |
|
"architectures": [ |
|
"StructFormerModelForSequenceClassification" |
|
], |
|
"auto_map": { |
|
"AutoConfig": "structformer.StructFormerConfig", |
|
"AutoModelForMaskedLM": "structformer.StructFormerModel", |
|
"AutoModelForSequenceClassification": "structformer.StructFormerModelForSequenceClassification" |
|
}, |
|
"conv_size": 9, |
|
"dropatt": 0.1, |
|
"dropout": 0.1, |
|
"hidden_dropout_prob": 0.1, |
|
"hidden_size": 512, |
|
"id2label": { |
|
"0": 0, |
|
"1": 1 |
|
}, |
|
"initializer_range": 0.02, |
|
"label2id": { |
|
"0": 0, |
|
"1": 1 |
|
}, |
|
"model_type": "structformer", |
|
"n_parser_layers": 3, |
|
"nhead": 8, |
|
"nlayers": 8, |
|
"ntokens": 16000, |
|
"pad": 1, |
|
"pos_emb": true, |
|
"problem_type": "single_label_classification", |
|
"relations": [ |
|
"head", |
|
"child" |
|
], |
|
"relative_bias": false, |
|
"torch_dtype": "float32", |
|
"transformers_version": "4.26.1", |
|
"weight_act": "softmax" |
|
} |
|
|