nanochat-keith-sft / meta.json
keithtyser's picture
Upload sft model (d20 @ step 715)
a278d4e verified
Raw History Blame Contribute Delete
262 Bytes
{
"step": 715,
"val_loss": 1.1147265434265137,
"mmlu_acc": 0.2919921875,
"arc_easy_acc": 0.27734375,
"model_config": {
"sequence_len": 2048,
"vocab_size": 65536,
"n_layer": 20,
"n_head": 10,
"n_kv_head": 10,
"n_embd": 1280
}
}