babyllama-wake-sleep / config.json
bkhmsi's picture
Push model using huggingface_hub.
033e238 verified
raw
history blame contribute delete
No virus
320 Bytes
{
"freeze-vision": true,
"language-model": {
"hidden_size": 512,
"intermediate_size": 1024,
"max_position_embeddings": 512,
"num_attention_heads": 8,
"num_hidden_layers": 16,
"num_key_value_heads": 8,
"vocab_size": 16000
},
"phase": 4,
"tokenizer": "tokenizers/bpe_babylm_100m_v2"
}