raavig's picture
Training in progress, epoch 1
82ddd1f verified
{
"alibi": true,
"architectures": [
"BertForSpanCategorization"
],
"attention_bias": false,
"attention_dropout": 0.0,
"attention_probs_dropout_prob": 0.0,
"bos_token_id": 50281,
"classifier_activation": "gelu",
"classifier_bias": false,
"classifier_dropout": 0.0,
"classifier_pooling": "mean",
"cls_token_id": 50281,
"context_length": 1024,
"decoder_bias": true,
"deterministic_flash_attn": false,
"dtype": "float32",
"embedding_dropout": 0.0,
"eos_token_id": 50282,
"flash_attention": true,
"gated_linear_units": true,
"global_attn_every_n_layers": 3,
"global_rope_theta": 160000.0,
"gradient_checkpointing": false,
"hidden_act": "gelu",
"hidden_activation": "gelu",
"hidden_dropout_prob": 0.0,
"hidden_size": 768,
"id2label": {
"0": "O",
"1": "B-BENZO",
"2": "I-BENZO",
"3": "B-COCAINE",
"4": "I-COCAINE",
"5": "B-IV",
"6": "I-IV",
"7": "B-KETAMINE",
"8": "I-KETAMINE",
"9": "B-OPIOIDS",
"10": "I-OPIOIDS",
"11": "B-PCP",
"12": "I-PCP",
"13": "B-OTHER_RX",
"14": "I-OTHER_RX",
"15": "B-AMPHETAMINE",
"16": "I-AMPHETAMINE",
"17": "B-UNSPECIFIED_SUBSTANCE",
"18": "I-UNSPECIFIED_SUBSTANCE",
"19": "B-TREATMENT_HISTORY",
"20": "I-TREATMENT_HISTORY",
"21": "B-INJ_EQP",
"22": "I-INJ_EQP",
"23": "B-OTHER_PARAPHERNALIA",
"24": "I-OTHER_PARAPHERNALIA",
"25": "B-PACKAGING",
"26": "I-PACKAGING",
"27": "B-PILLS",
"28": "I-PILLS",
"29": "B-POWDER",
"30": "I-POWDER",
"31": "B-GLASS_PIPE",
"32": "I-GLASS_PIPE",
"33": "B-STRAW",
"34": "I-STRAW",
"35": "B-PRIOR_OD",
"36": "I-PRIOR_OD"
},
"initializer_cutoff_factor": 2.0,
"initializer_range": 0.02,
"intermediate_size": 1152,
"label2id": {
"B-AMPHETAMINE": 15,
"B-BENZO": 1,
"B-COCAINE": 3,
"B-GLASS_PIPE": 31,
"B-INJ_EQP": 21,
"B-IV": 5,
"B-KETAMINE": 7,
"B-OPIOIDS": 9,
"B-OTHER_PARAPHERNALIA": 23,
"B-OTHER_RX": 13,
"B-PACKAGING": 25,
"B-PCP": 11,
"B-PILLS": 27,
"B-POWDER": 29,
"B-PRIOR_OD": 35,
"B-STRAW": 33,
"B-TREATMENT_HISTORY": 19,
"B-UNSPECIFIED_SUBSTANCE": 17,
"I-AMPHETAMINE": 16,
"I-BENZO": 2,
"I-COCAINE": 4,
"I-GLASS_PIPE": 32,
"I-INJ_EQP": 22,
"I-IV": 6,
"I-KETAMINE": 8,
"I-OPIOIDS": 10,
"I-OTHER_PARAPHERNALIA": 24,
"I-OTHER_RX": 14,
"I-PACKAGING": 26,
"I-PCP": 12,
"I-PILLS": 28,
"I-POWDER": 30,
"I-PRIOR_OD": 36,
"I-STRAW": 34,
"I-TREATMENT_HISTORY": 20,
"I-UNSPECIFIED_SUBSTANCE": 18,
"O": 0
},
"layer_norm_eps": 1e-05,
"local_attention": 128,
"local_rope_theta": 10000.0,
"max_position_embeddings": 8192,
"mlp_bias": false,
"mlp_dropout": 0.0,
"model_type": "bert",
"norm_bias": false,
"norm_eps": 1e-05,
"num_attention_heads": 12,
"num_hidden_layers": 22,
"pad_token_id": 50283,
"position_embedding_type": "absolute",
"reference_compile": null,
"repad_logits_with_grad": false,
"rope_theta": 10000,
"sep_token_id": 50282,
"sparse_pred_ignore_index": -100,
"sparse_prediction": false,
"transformers_version": "4.57.1",
"type_vocab_size": 2,
"use_cache": true,
"use_low_precision_layernorm": true,
"vocab_size": 50368
}