NanoChat / midtraining /meta_000813.json
code1phoenix's picture
Upload folder using huggingface_hub
908224f verified
raw
history blame contribute delete
603 Bytes
{
"step": 813,
"val_bpb": 0.43868296985286714,
"model_config": {
"sequence_len": 2048,
"vocab_size": 65536,
"n_layer": 20,
"n_head": 10,
"n_kv_head": 10,
"n_embd": 1280
},
"user_config": {
"run": "dummy",
"device_type": "",
"dtype": "bfloat16",
"num_iterations": -1,
"max_seq_len": 2048,
"device_batch_size": 8,
"unembedding_lr": 0.004,
"embedding_lr": 0.2,
"matrix_lr": 0.02,
"init_lr_frac": 1.0,
"weight_decay": 0.0,
"eval_every": 150,
"eval_tokens": 10485760,
"total_batch_size": 524288,
"dry_run": 0
}
}