sedrickkeh commited on
Commit
f1fd49c
1 Parent(s): a761cf5

Upload configs.yaml with huggingface_hub

Browse files
Files changed (1) hide show
  1. configs.yaml +41 -0
configs.yaml ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ adam_beta1: '0.9'
2
+ adam_beta2: '0.999'
3
+ bf16: 'True'
4
+ cutoff_len: '2048'
5
+ dataset: mlfoundations-dev/stackexchange_chess
6
+ dataset_dir: ONLINE
7
+ ddp_timeout: '180000000'
8
+ deepspeed: /opt/ml/code/zero3.json
9
+ do_train: 'True'
10
+ enable_liger_kernel: 'False'
11
+ eval_strategy: epoch
12
+ finetuning_type: full
13
+ formatting: sharegpt
14
+ global_batch_size: '512'
15
+ gradient_accumulation_steps: '8'
16
+ gradient_checkpointing: 'True'
17
+ hub_model_id: mlfoundations-dev/stackexchange_chess
18
+ learning_rate: 5e-06
19
+ logging_steps: '10'
20
+ lr_scheduler_type: constant
21
+ max_grad_norm: '1'
22
+ messages: conversations
23
+ model_name_or_path: meta-llama/Meta-Llama-3.1-8B
24
+ neat_packing: 'True'
25
+ num_train_epochs: '3.0'
26
+ output_dir: /opt/ml/model
27
+ overwrite_cache: 'True'
28
+ overwrite_output_dir: 'True'
29
+ packing: 'True'
30
+ per_device_train_batch_size: '8'
31
+ plot_loss: 'True'
32
+ preprocessing_num_workers: '16'
33
+ push_to_db: 'True'
34
+ push_to_hub: 'True'
35
+ report_to: wandb
36
+ run_name: stackexchange_chess
37
+ save_strategy: epoch
38
+ stage: sft
39
+ template: llama3
40
+ val_size: '0.05'
41
+ weight_decay: '0.1'