gla-340m-15B / config.json
bailin28's picture
Upload GLAForCausalLM
c0c456d verified
raw
history blame contribute delete
573 Bytes
{
"_name_or_path": "/u/yikangshen/bw_mnt/bailinw/checkpoints/gla_340m/2024-03-26/14-51-15-910951/huggingface/",
"architectures": [
"GLAForCausalLM"
],
"bos_token_id": 0,
"context_length": 2048,
"d_model": 1024,
"eos_token_id": 0,
"model_type": "gla",
"n_head": 4,
"n_layer": 24,
"pad_token_id": 0,
"tie_word_embeddings": false,
"torch_dtype": "float32",
"transformers_version": "4.38.1",
"use_cache": true,
"use_conv1d": false,
"use_gk": true,
"use_gv": false,
"use_lowrank": true,
"use_xformers": false,
"vocab_size": 32000
}