fastplus-125m-test / config.json
aixk's picture
FastPlus-125M Relativistic Pure Geometric Lorentz Sequence Model Train Completed Safely
6610dca
Raw
History Blame Contribute Delete
439 Bytes
{
"architectures": [
"FastPlus125mForCausalLM"
],
"dtype": "float32",
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 2048,
"kd_alpha": 0.4,
"kd_temperature": 2.5,
"max_position_embeddings": 512,
"model_type": "fastplus_125m",
"num_attention_heads": 12,
"num_hidden_layers": 16,
"tie_word_embeddings": true,
"transformers_version": "5.12.0",
"use_cache": false,
"vocab_size": 1792
}