-
Notifications
You must be signed in to change notification settings - Fork 0
/
settings.yaml
56 lines (56 loc) · 1.33 KB
/
settings.yaml
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
53
54
55
56
lapis_yanolja+EEVE-Korean-Instruct-10.8B-v1.0_True:
openai:
api_key:
wandb:
api_key:
project_name: CIKM_Lapis_2024
group_name: 240423_main_experiment
session_name: lapis_yanolja+EEVE-Korean-Instruct-10.8B-v1.0_True
path:
dataset: /lapis/data/
checkpoint: /lapis/checkpoint/
template: /lapis/template/
result: /lapis/result/
dataprep:
raw_dataset: CI_hypothesis_v1
finetuning_dataset: CI_hypothesis_v1
instruction_method: correct
subsample: 1.0
years_train: null
years_dev:
- 2020
years_test:
- 2021
- 2022
- 2023
finetune:
device: cuda
report_to: wandb
llm_backbone: yanolja/EEVE-Korean-Instruct-10.8B-v1.0
lora:
enabled: true
qlora: true
r: 8
alpha: 32
dropout: 0.05
bias: none
task_type: CAUSAL_LM
per_device_train_batch_size: 4
gradient_accumulation_steps: 1
num_train_epochs: 5
learning_rate: 0.0001
fp16: true
logging_steps: 1000
output_dir: null
optim: paged_adamw_8bit
inference:
checkpoint_name: null
template_method: v3_zs_reasoning+para_kor
retrieval:
embedding_library: openai
embedding_model: text-embedding-ada-002
vector_library: faiss
vector_store: faiss_all_summary_v2_index
generation:
max_new_tokens: 8
start_batch: 0