File size: 926 Bytes
1710729
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Model arguments
model_name_or_path: sanchit-gandhi/Mistral-7B-v0.1-6-layer
teacher_model_name_or_path: mistralai/Mistral-7B-v0.1
dtype: bfloat16
load_teacher_in_4bit: true

# Data arguments
train_dataset_name: HuggingFaceTB/cosmopedia
train_dataset_config_name:
  - auto_math_text
  - khanacademy
  - openstax
  - stanford
  - stories
  - web_samples_v1
  - web_samples_v2
  - wikihow
train_split_name: train[1000:]
eval_split_name: train[:1000]
max_steps: 50000

# Training arguments
do_train: true
do_eval: true
per_device_eval_batch_size: 8
per_device_train_batch_size: 8
learning_rate: 0.0003
warmup_steps: 500
gradient_checkpointing: true
dataloader_num_workers: 4
preprocessing_num_workers: 32
ddp_timeout: 7200
save_strategy: steps
save_steps: 5000
evaluation_strategy: steps
eval_steps: 5000
logging_steps: 25
output_router_logits: true
report_to: wandb
output_dir: ./
overwrite_output_dir: true
save_total_limit: 1