fdeantoni commited on
Commit
cfc5cc3
·
verified ·
1 Parent(s): af54def

Upload chatsft_checkpoints/d18/meta_000499.json with huggingface_hub

Browse files
chatsft_checkpoints/d18/meta_000499.json ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 499,
3
+ "val_bpb": 0.23425339651983787,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 32768,
7
+ "n_layer": 18,
8
+ "n_head": 9,
9
+ "n_kv_head": 9,
10
+ "n_embd": 1152,
11
+ "window_pattern": "SSSL"
12
+ },
13
+ "user_config": {
14
+ "run": "dummy",
15
+ "device_type": "",
16
+ "model_tag": "d18",
17
+ "model_step": null,
18
+ "load_optimizer": 1,
19
+ "num_iterations": 500,
20
+ "max_seq_len": null,
21
+ "device_batch_size": 16,
22
+ "total_batch_size": 65536,
23
+ "embedding_lr": null,
24
+ "unembedding_lr": null,
25
+ "matrix_lr": null,
26
+ "init_lr_frac": 0.8,
27
+ "warmup_ratio": 0.0,
28
+ "warmdown_ratio": 0.5,
29
+ "final_lr_frac": 0.0,
30
+ "eval_every": 200,
31
+ "eval_tokens": 20971520,
32
+ "chatcore_every": -1,
33
+ "chatcore_max_cat": -1,
34
+ "chatcore_max_sample": 24,
35
+ "mmlu_epochs": 3,
36
+ "gsm8k_epochs": 4,
37
+ "save_every": 100,
38
+ "sft_file": "/workspace/.cache/nanochat/sft_data_babbelaar/sft_train.jsonl",
39
+ "sft_val_file": "/workspace/.cache/nanochat/sft_data_babbelaar/sft_val.jsonl"
40
+ }
41
+ }