fdeantoni commited on
Commit
aeeed46
·
verified ·
1 Parent(s): ada4c8f

Upload chatsft_checkpoints/d18/meta_001750.json with huggingface_hub

Browse files
chatsft_checkpoints/d18/meta_001750.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 1750,
3
+ "val_bpb": 0.2651569889048805,
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
+ "resume_sft_step": null,
19
+ "load_optimizer": 1,
20
+ "num_iterations": 5000,
21
+ "max_seq_len": null,
22
+ "device_batch_size": 16,
23
+ "total_batch_size": 32768,
24
+ "embedding_lr": null,
25
+ "unembedding_lr": null,
26
+ "matrix_lr": null,
27
+ "init_lr_frac": 0.8,
28
+ "warmup_ratio": 0.0,
29
+ "warmdown_ratio": 0.5,
30
+ "final_lr_frac": 0.0,
31
+ "eval_every": 250,
32
+ "eval_tokens": 20971520,
33
+ "chatcore_every": -1,
34
+ "chatcore_max_cat": -1,
35
+ "chatcore_max_sample": 24,
36
+ "mmlu_epochs": 3,
37
+ "gsm8k_epochs": 4,
38
+ "save_every": 250,
39
+ "sft_file": "/workspace/.cache/nanochat/sft_data_babbelaar/sft_train.jsonl",
40
+ "sft_val_file": "/workspace/.cache/nanochat/sft_data_babbelaar/sft_val.jsonl"
41
+ }
42
+ }