eGPT-1B-untrained

Randomly initialized eGPT decoder-only model (1.13B parameters). Not trained.

Architecture

Field	Value
Parameters	1.13B
Layers	24
Dim	2048
Heads (Q)	16
Heads (KV)	8
Head dim	128
FFN hidden	5632
Max seq len	2048
Vocab size	256
Tokenizer	google/byt5-small

Loading

from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer

tok   = AutoTokenizer.from_pretrained("LLMsHub/eGPT-1B-untrained", trust_remote_code=True)
cfg   = AutoConfig.from_pretrained("LLMsHub/eGPT-1B-untrained", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("LLMsHub/eGPT-1B-untrained", trust_remote_code=True)