Release trained ModernBERT-JEV model with option-query cross-attention head, calibration, and documentation

790beda verified 1 day ago

377 Bytes

	{
	"backbone_name_or_path": "answerdotai/ModernBERT-base",
	"cls_token_id": 50281,
	"cross_attention_heads": 8,
	"dtype": "bfloat16",
	"hidden_size": 768,
	"max_context_length": 512,
	"max_option_length": 32,
	"model_type": "modernjev",
	"num_kinds": 3,
	"pad_token_id": 50283,
	"score_rps_weight": 0.2,
	"sep_token_id": 50282,
	"transformers_version": "4.57.6"
	}