Hugging Face's logo

tyzhu
/
litgpt_pretrain_jul23

Model card Files Community
litgpt_pretrain_jul23
File size: 991 Bytes 
caf988f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
 
model_name: Llama-3.2-3B
out_dir: /home/aiops/zhuty/litgpt/out/pretrain/llama3-3B-bio1k-2
precision: bf16-mixed
initial_checkpoint_dir: checkpoints/meta-llama/Llama-3.2-3B
resume: auto
data:
  class_path: litgpt.data.LitData
  init_args:
    data_path: /home/aiops/zhuty/cont_pretrain_processed/bio_1k
    split_names:
    - train
    - valid
    seed: 42
    num_workers: 8
    seq_length: 2048
train:
  save_interval: 2500
  log_interval: 1
  global_batch_size: 256
  micro_batch_size: 1
  lr_warmup_fraction: 0.01
  max_tokens: 30000000
  max_seq_length: 1024
  max_norm: 1.0
  min_lr: 5.0e-06
eval:
  interval: 10
  max_iters: 100
  initial_validation: true
  final_validation: true
  evaluate_example: first
log:
  project: mathcont
optimizer:
  class_path: torch.optim.AdamW
  init_args:
    lr: 5.0e-05
    weight_decay: 0.1
    betas:
    - 0.9
    - 0.95
devices: auto
num_nodes: 1
logger_name: wandb
seed: 42
compiler: torch
executors:
- sdpa
- torchcompile
- torch
strategy: fsdp