RMSN-310M-Base / config.json
DuoNeural's picture
Upload RMSN-310M-Base — DuoNeural RMSN Phase 4 (Nov 2026)
17b0ed0 verified
Raw History Blame Contribute Delete
324 Bytes
{
"architectures": [
"RMSN"
],
"model_type": "rmsn",
"vocab_size": 49152,
"dim": 1152,
"depth": 16,
"slots": 64,
"seq_len": 512,
"num_parameters": 293031984,
"tokenizer_class": "PreTrainedTokenizerFast",
"tokenizer_name": "HuggingFaceTB/cosmo2-tokenizer",
"val_ppl": 68.1,
"val_lyap": -0.9714
}