aixk's picture
Training in progress, step 200
bedd674 verified
Raw History Blame Contribute Delete
354 Bytes
{
"architectures": [
"HyperBitSSMForCausalLM"
],
"dtype": "float32",
"hidden_size": 768,
"hop": 48,
"intermediate_size": 1792,
"lowrank": 32,
"model_type": "hyper_bit_ssm_98m_v55",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"transformers_version": "5.15.0",
"use_cache": false,
"vocab_size": 8192,
"window": 96
}