Repository navigation
Expand file tree
/
Copy pathconfig.py
More file actions
185 lines (155 loc) · 5.86 KB
/
Copy pathconfig.py
File metadata and controls
185 lines (155 loc) · 5.86 KB
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
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
import os
from pathlib import Path
from dotenv import load_dotenv
# Load .env from project root
PROJECT_ROOT = Path(__file__).resolve().parent.parent
load_dotenv(PROJECT_ROOT / ".env")
# ==================== Path Configuration ====================
DATA_DIR = PROJECT_ROOT / "data"
RESULTS_DIR = PROJECT_ROOT / "results"
class ExperimentConfig:
# ==================== General ====================
experiment_name: str = os.environ.get("HYPERMEM_EXPERIMENT_NAME", "HyperMem-v3")
dataset_path: str = str(DATA_DIR / "locomo10.json")
num_conv: int = 10
# ==================== Stage 2: Index Building ====================
hyperedge_emb_aggregate_type: str = "sum"
node_emb_update_weight: float = 0.5
embedding_config: dict = {
"model_name": "Qwen3-Embedding-4B",
"base_url": os.environ.get("EMBEDDING_BASE_URL", "http://localhost:11810/v1"),
}
embedding_max_retries: int = 10
# ==================== Stage 3: Retrieval ====================
retrieval_type: str = "rrf"
rerank_type: str = "fix"
hypergraph_retrieval_output_type: str = "011"
retrieval_config: dict = {
"initial_candidates": int(os.environ.get("HYPERMEM_INITIAL_CANDIDATES", "100")),
"topic_top_k": int(os.environ.get("HYPERMEM_TOPIC_TOP_K", "15")),
"episode_top_k": int(os.environ.get("HYPERMEM_EPISODE_TOP_K", "20")),
"fact_top_k": int(os.environ.get("HYPERMEM_FACT_TOP_K", "30")),
}
adaptive_retrieval_config: dict = {
"factual": {
"initial_candidates": 180,
"topic_top_k": 8,
"episode_top_k": 16,
"fact_top_k": 24,
},
"temporal": {
"initial_candidates": 200,
"topic_top_k": 10,
"episode_top_k": 20,
"fact_top_k": 30,
},
"reasoning": {
"initial_candidates": 250,
"topic_top_k": 12,
"episode_top_k": 25,
"fact_top_k": 35,
},
"commonsense": {
"initial_candidates": 180,
"topic_top_k": 8,
"episode_top_k": 16,
"fact_top_k": 24,
},
"default": {
"initial_candidates": 200,
"topic_top_k": 10,
"episode_top_k": 20,
"fact_top_k": 30,
}
}
temporal_enhancement: bool = True
use_reranker: bool = os.environ.get("HYPERMEM_USE_RERANKER", "false").lower() == "true"
reranker_config: dict = {
"model_name": "Qwen3-Reranker-4B",
"base_url": os.environ.get("RERANKER_BASE_URL", "http://localhost:12810"),
}
reranker_max_retries: int = 10
# ==================== Stage 4: Response Generation ====================
answer_type: str = "cot"
llm_service: str = "openai"
llm_config: dict = {
"openai": {
"llm_provider": "openai",
"model": "gpt-4.1-mini-2025-04-14",
"base_url": "https://openrouter.ai/api/v1",
"api_key": os.environ.get("OPENROUTER_API_KEY", ""),
"temperature": 0,
"max_tokens": 16384,
},
"gemini": {
"llm_provider": "openai",
"model": "google/gemini-2.5-flash",
"base_url": "https://openrouter.ai/api/v1",
"api_key": os.environ.get("OPENROUTER_API_KEY", ""),
"temperature": 0.3,
"max_tokens": 16384,
},
"vllm": {
"llm_provider": "openai",
"model": "Qwen3-30B",
"base_url": os.environ.get("VLLM_BASE_URL", "http://localhost:8000/v1"),
"api_key": "unused",
"temperature": 0,
"max_tokens": 20000,
}
}
judge_llm_config: dict = {
"model": "gpt-4o-mini",
"base_url": "https://openrouter.ai/api/v1",
"api_key": os.environ.get("OPENROUTER_API_KEY", ""),
"temperature": 0,
}
llm_max_retries: int = 10
max_concurrent_requests: int = 10
# ==================== Derived Paths ====================
@classmethod
def experiment_dir(cls) -> Path:
return RESULTS_DIR / cls.experiment_name
@classmethod
def episodes_dir(cls) -> Path:
return cls.experiment_dir() / "episodes"
@classmethod
def hypergraph_dir(cls) -> Path:
return cls.experiment_dir() / "hypergraphs"
@classmethod
def facts_dir(cls) -> Path:
return cls.experiment_dir() / "facts"
@classmethod
def topics_dir(cls) -> Path:
return cls.experiment_dir() / "topics"
@classmethod
def token_stats_dir(cls) -> Path:
return cls.experiment_dir() / "token_stats"
@classmethod
def bm25_index_dir(cls) -> Path:
return cls.experiment_dir() / "bm25_index"
@classmethod
def vectors_dir(cls) -> Path:
return cls.experiment_dir() / "vectors"
def _build_experiment_name():
parts = [ExperimentConfig.experiment_name]
if ExperimentConfig.retrieval_type in ('vector', 'rrf'):
parts.append(ExperimentConfig.hyperedge_emb_aggregate_type)
parts.append(f"alpha{ExperimentConfig.node_emb_update_weight}")
if ExperimentConfig.retrieval_type in ('vector', 'rrf'):
parts.append(f"{ExperimentConfig.retrieval_type.upper()}")
if ExperimentConfig.rerank_type == "ada":
parts.append(ExperimentConfig.rerank_type)
else:
rc = ExperimentConfig.retrieval_config
parts.append(f"{ExperimentConfig.rerank_type}-{rc['initial_candidates']}-{rc['topic_top_k']}-{rc['episode_top_k']}-{rc['fact_top_k']}")
else:
parts.append(ExperimentConfig.retrieval_type)
parts.append(f"r{ExperimentConfig.hypergraph_retrieval_output_type}")
if ExperimentConfig.use_reranker:
parts.append("rerank")
else:
parts.append("wo-rerank")
parts.append(f"{ExperimentConfig.answer_type}")
return "_".join(parts)
ExperimentConfig.experiment_name = _build_experiment_name()