deltasplice-human / config.json
ZhiyuanChen's picture
Upload folder using huggingface_hub
c16a7a1 verified
{
"architectures": [
"DeltaSpliceModel"
],
"batch_norm_eps": 1e-05,
"batch_norm_momentum": 0.1,
"bos_token_id": null,
"context": 30000,
"dropout": 0.3,
"dtype": "float32",
"eos_token_id": null,
"head": {
"act": null,
"bias": true,
"dropout": 0.0,
"hidden_size": 64,
"layer_norm_eps": 1e-12,
"loss_weight": null,
"num_labels": 3,
"output_name": null,
"problem_type": "regression",
"transform": null,
"transform_act": "gelu",
"type": null
},
"hidden_act": "relu",
"hidden_size": 64,
"id2label": null,
"label2id": null,
"layers": [
{
"dilation": 1,
"kernel_size": 11
},
{
"dilation": 1,
"kernel_size": 11
},
{
"dilation": 1,
"kernel_size": 11
},
{
"dilation": 1,
"kernel_size": 11
},
{
"dilation": 1,
"kernel_size": 19
},
{
"dilation": 1,
"kernel_size": 19
},
{
"dilation": 1,
"kernel_size": 19
},
{
"dilation": 1,
"kernel_size": 19
},
{
"dilation": 2,
"kernel_size": 25
},
{
"dilation": 2,
"kernel_size": 25
},
{
"dilation": 2,
"kernel_size": 25
},
{
"dilation": 2,
"kernel_size": 25
},
{
"dilation": 8,
"kernel_size": 33
},
{
"dilation": 8,
"kernel_size": 33
},
{
"dilation": 8,
"kernel_size": 33
},
{
"dilation": 8,
"kernel_size": 33
},
{
"dilation": 16,
"kernel_size": 43
},
{
"dilation": 16,
"kernel_size": 43
},
{
"dilation": 16,
"kernel_size": 85
},
{
"dilation": 16,
"kernel_size": 85
},
{
"dilation": 16,
"kernel_size": 85
},
{
"dilation": 16,
"kernel_size": 85
},
{
"dilation": 32,
"kernel_size": 85
},
{
"dilation": 32,
"kernel_size": 85
}
],
"mask_token_id": null,
"model_type": "deltasplice",
"null_token_id": null,
"num_ensemble": 5,
"num_labels": 3,
"output_contexts": false,
"pad_token_id": 4,
"problem_type": "regression",
"tie_word_embeddings": true,
"transformers_version": "5.9.0",
"unk_token_id": 4,
"vocab_size": 4
}