MiniCPM3-4B-mic / config.json
tarekziade's picture
tarekziade HF Staff
Publish MIC usage, slim fixture and reviewed Transformers expectations
e35e5fa verified
Raw History Blame Contribute Delete
1.65 kB
{
"architectures": [
"MiniCPM3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"auto_map": {
"AutoConfig": "configuration_minicpm.MiniCPM3Config",
"AutoModel": "modeling_minicpm.MiniCPM3Model",
"AutoModelForCausalLM": "modeling_minicpm.MiniCPM3ForCausalLM",
"AutoModelForSeq2SeqLM": "modeling_minicpm.MiniCPM3ForCausalLM",
"AutoModelForSequenceClassification": "modeling_minicpm.MiniCPM3ForSequenceClassification"
},
"bos_token_id": 1,
"dim_model_base": 256,
"dtype": "bfloat16",
"eos_token_id": [
2,
73440
],
"head_dim": 8,
"hidden_act": "silu",
"hidden_size": 64,
"initializer_range": 0.1,
"intermediate_size": 128,
"kv_lora_rank": 16,
"max_position_embeddings": 32768,
"mlp_bias": false,
"model_type": "minicpm3",
"num_attention_heads": 4,
"num_hidden_layers": 2,
"num_key_value_heads": 4,
"pad_token_id": null,
"pretraining_tp": 1,
"q_lora_rank": 16,
"qk_head_dim": 96,
"qk_nope_head_dim": 16,
"qk_rope_head_dim": 8,
"rms_norm_eps": 1e-05,
"rope_parameters": {
"long_factor": [
1.0591234137867171,
2.093982484148734,
10.475458000005451,
16.461961959767827
],
"original_max_position_embeddings": 32768,
"rope_theta": 10000.0,
"rope_type": "longrope",
"short_factor": [
1.0591234137867171,
2.093982484148734,
10.475458000005451,
16.461961959767827
],
"type": "longrope"
},
"scale_depth": 1.4,
"scale_emb": 12,
"tie_word_embeddings": true,
"transformers_version": "5.21.0.dev0",
"use_cache": true,
"v_head_dim": 16,
"vocab_size": 73448
}