{ | |
"vocab_size": 49152, | |
"hidden_size": 576, | |
"intermediate_size": 1536, | |
"num_hidden_layers": 30, | |
"num_attention_heads": 9, | |
"num_key_value_heads": 3, | |
"hidden_act": "silu", | |
"max_position_embeddings": 2048, | |
"initializer_range": 0.041666666666666664, | |
"rms_norm_eps": 1e-05, | |
"use_cache": true, | |
"pad_token_id": null, | |
"bos_token_id": 0, | |
"eos_token_id": 0, | |
"tie_word_embeddings": true, | |
"rope_theta": 10000.0, | |
"architectures": [ | |
"SmolLM2" | |
], | |
"model_type": "smollm2" | |
} |