-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathconfig.yaml
91 lines (73 loc) · 2.86 KB
/
config.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
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
# Config parameters for LLM generation:
api_base: ''
api_key: ""
max_trials: 5
temperature: 0.1
max_tokens: 4000
random_seed: 42
reset_time: 1.2
# Config parameters for viewpoint extraction:
llm_model: "mistralai/Mistral-7B-Instruct-v0.3"
save_node_list: True
node_list_save_path: "./save"
# Config parameters for embedding-based viewpoint-graph construction:
intra_graph_deg: 5
inter_graph_deg: 10
batch_size: -1 # The number of nodes processed per batch in embedding-based viewpoint-graph construction, where -1 indicates processing all nodes simultaneously.
save_edge_list: True
edge_list_save_path: "./save"
## Config parameters for ICLR datasets:
dataset_path: "Data/ICLR_Dataset/ICLR_dataset.jsonl" # The dataset containing all research ideas.
training_set_path: "Data/ICLR_Dataset/ICLR_training_set.jsonl" # For GraphEval and Finetuned BERT, use the ideas in this subset for training or label propagation.
test_set_path: "Data/ICLR_Dataset/ICLR_test_set.jsonl" # For all methods, conduct testing on this subset.
node_list_paths:
- "Data/ICLR_Dataset/node_list/training_node_list.jsonl"
- "Data/ICLR_Dataset/node_list/test_node_list.jsonl"
edge_list_paths:
- "Data/ICLR_Dataset/edge_list/bert_based_edge_list.jsonl"
# Config parameters for AI Researcher dataset:
#dataset_path: "Data/AI_Researcher/dataset.jsonl" # The dataset containing all research ideas.
#training_set_path: "Data/AI_Researcher/training_set.jsonl" # For GraphEval and Finetuned BERT, use the ideas in this subset for training or label propagation.
#test_set_path: "Data/AI_Researcher/test_set.jsonl" # For all methods, conduct testing on this subset.
#node_list_paths:
# - "Data/AI_Researcher/node_list/training_node_list.jsonl"
# - "Data/AI_Researcher/node_list/test_node_list.jsonl"
#edge_list_paths:
# - "Data/AI_Researcher/edge_list/edge_list.jsonl"
# Config parameters for Task:
Task_name: "ICLR" # AI_Researcher ICLR
seed: 666
# Config parameters for GraphEval-GNN:
GNN_max_iters: 100
GNN_hidden_features_dim: 64
GNN_output_dim: 4 # AI_Researcher:3 ICLR:4
GNN_edge_dim: 1
mask_rate: 0.5
GNN_batch_size: 32
model_path: 'model_path/best_model.pth'
learning_rate: 1e-3
wandb_key: ''
# Config parameters for GraphEval-LP:
LP_max_iters: 5
LP_label_weights:
0: 1
1: 1.068
2: 1.1
3: 1.128
# Config parameters for all Baselines:
baseline: "Finetuned Bert" #["Prompted LLM", "CoT", "CoT-SC", "ToT", "Research Agent", "Finetuned Bert"]
save_baseline_results: True
baseline_results_save_path: "./results/baseline_results.jsonl"
# Config parameters for CoT-SC Baseline:
CoT_SC_k: 5
# Config parameters for ToT Baseline:
ToT_branch_num: 3
# Config parameters for Finetuned Bert Baseline:
finetuned_Bert_model_name: "distilbert-base-uncased"
Bert_evaluation_strategy: "epoch"
Bert_train_batch_size: 8
Bert_eval_batch_size: 8
Bert_num_train_epochs: 3
Bert_learning_rate: 3e-5
Bert_weight_decay: 0.01
Bert_logging_steps: 10