ea9f01294d
Model: Firemedic15/nemotron-mini-4b-travel-risk-analysis-merged Source: Original Platform
35 lines
830 B
JSON
35 lines
830 B
JSON
{
|
|
"architectures": [
|
|
"NemotronForCausalLM"
|
|
],
|
|
"attention_bias": false,
|
|
"attention_dropout": 0.0,
|
|
"bos_token_id": 2,
|
|
"dtype": "bfloat16",
|
|
"eos_token_id": 3,
|
|
"head_dim": 128,
|
|
"hidden_act": "relu2",
|
|
"hidden_size": 3072,
|
|
"initializer_range": 0.0134,
|
|
"intermediate_size": 9216,
|
|
"kv_channels": 128,
|
|
"max_position_embeddings": 4096,
|
|
"mlp_bias": false,
|
|
"model_type": "nemotron",
|
|
"norm_eps": 1e-05,
|
|
"num_attention_heads": 24,
|
|
"num_hidden_layers": 32,
|
|
"num_key_value_heads": 8,
|
|
"pad_token_id": null,
|
|
"partial_rotary_factor": 0.5,
|
|
"rope_parameters": {
|
|
"partial_rotary_factor": 0.5,
|
|
"rope_theta": 10000,
|
|
"rope_type": "default"
|
|
},
|
|
"tie_word_embeddings": false,
|
|
"transformers_version": "5.11.0",
|
|
"use_cache": true,
|
|
"vocab_size": 256000
|
|
}
|