FastLLM / config.json
devoppro's picture
Training in progress, step 500
58c67da verified
Raw History Blame
475 Bytes
{
"architectures": [
"ModernLLMForCausalLM"
],
"bos_token_id": 1,
"dtype": "float32",
"eos_token_id": 2,
"hidden_size": 768,
"intermediate_size": 2048,
"max_position_embeddings": 2048,
"model_type": "modern_llm",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"num_key_value_heads": 4,
"pad_token_id": 0,
"rms_norm_eps": 1e-06,
"rope_theta": 1000000.0,
"transformers_version": "5.15.1",
"use_cache": false,
"vocab_size": 151936
}