yimeiyang's picture
Training in progress, step 500
c2ace78 verified
raw
history blame contribute delete
495 Bytes
{
"_name_or_path": "RWKV/rwkv-4-169m-pile",
"architectures": [
"RwkvForCausalLM"
],
"attention_hidden_size": 768,
"bos_token_id": 0,
"context_length": 1024,
"eos_token_id": 0,
"hidden_size": 768,
"intermediate_size": 3072,
"layer_norm_epsilon": 1e-05,
"model_type": "rwkv",
"num_hidden_layers": 12,
"rescale_every": 6,
"tie_word_embeddings": false,
"torch_dtype": "float32",
"transformers_version": "4.39.0.dev0",
"use_cache": true,
"vocab_size": 50277
}