AlexSimon06/Hybridmodel5
03
1{2 "architectures": [3 "EuroBertForSequenceClassification"4 ],5 "attention_bias": false,6 "attention_dropout": 0.0,7 "auto_map": {8 "AutoConfig": "EuroBERT/EuroBERT-210m--configuration_eurobert.EuroBertConfig",9 "AutoModel": "EuroBERT/EuroBERT-210m--modeling_eurobert.EuroBertModel",10 "AutoModelForMaskedLM": "EuroBERT/EuroBERT-210m--modeling_eurobert.EuroBertForMaskedLM",11 "AutoModelForPreTraining": "EuroBERT/EuroBERT-210m--modeling_eurobert.EuroBertPreTrainedModel",12 "AutoModelForSequenceClassification": "EuroBERT/EuroBERT-210m--modeling_eurobert.EuroBertForSequenceClassification",13 "AutoModelForTokenClassification": "EuroBERT/EuroBERT-210m--modeling_eurobert.EuroBertForTokenClassification"14 },15 "bos_token": "<|begin_of_text|>",16 "bos_token_id": 128000,17 "clf_pooling": "late",18 "eos_token": "<|end_of_text|>",19 "eos_token_id": 128001,20 "head_dim": 64,21 "hidden_act": "silu",22 "hidden_dropout": 0.0,23 "hidden_size": 768,24 "initializer_range": 0.02,25 "intermediate_size": 3072,26 "mask_token": "<|mask|>",27 "mask_token_id": 128002,28 "max_position_embeddings": 8192,29 "mlp_bias": false,30 "model_type": "eurobert",31 "num_attention_heads": 12,32 "num_hidden_layers": 12,33 "num_key_value_heads": 12,34 "pad_token": "<|end_of_text|>",35 "pad_token_id": 128001,36 "pretraining_tp": 1,37 "rms_norm_eps": 1e-05,38 "rope_scaling": null,39 "rope_theta": 250000,40 "tie_word_embeddings": false,41 "torch_dtype": "float32",42 "transformers_version": "4.52.4",43 "use_cache": false,44 "vocab_size": 12825645}46 