forked from MineDojo/Voyager
-
Notifications
You must be signed in to change notification settings - Fork 0
/
run.py
100 lines (82 loc) · 4.08 KB
/
run.py
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
from voyager import Voyager
import dotenv
import os
import dotenv
dotenv.load_dotenv()
openai_api_key = os.getenv("OPENAI_API_KEY")
voyager = Voyager(
#azure_login=azure_login,
mc_port=55130,
openai_api_key=openai_api_key,
ckpt_dir='skill_library/gpt-4o-mini-2171441100564293352',
#resume=True,
#ckpt_dir='skill_library/claude--2171441100564293352',
# action_agent_model_name='gpt-4-0125-preview',
# critic_agent_model_name='gpt-4-0125-preview',
# curriculum_agent_model_name='gpt-4-0125-preview',
# curriculum_agent_qa_model_name='gpt-4-0125-preview',
# skill_manager_model_name='gpt-4-0125-preview',
# action_agent_model_name='gpt-4-turbo',
# critic_agent_model_name='gpt-4-turbo',
# curriculum_agent_model_name='gpt-4-turbo',
# curriculum_agent_qa_model_name='gpt-4-turbo',
# skill_manager_model_name='gpt-4-turbo',
# action_agent_model_name='gpt-4o',
# critic_agent_model_name='gpt-4o',
# curriculum_agent_model_name='gpt-4o',
# curriculum_agent_qa_model_name='gpt-4o',
# skill_manager_model_name='gpt-4o',
action_agent_model_name='gpt-4o-mini',
critic_agent_model_name='gpt-4o-mini',
curriculum_agent_model_name='gpt-4o-mini',
curriculum_agent_qa_model_name='gpt-4o-mini',
skill_manager_model_name='gpt-4o-mini',
# action_agent_model_name='models/gemini-1.0-pro-latest',
# critic_agent_model_name='models/gemini-1.0-pro-latest',
# curriculum_agent_model_name='models/gemini-1.0-pro-latest',
# curriculum_agent_qa_model_name='models/gemini-1.0-pro-latest',
# skill_manager_model_name='models/gemini-1.0-pro-latest',
# action_agent_model_name='models/gemini-1.5-flash',
# critic_agent_model_name='models/gemini-1.5-flash',
# curriculum_agent_model_name='models/gemini-1.5-flash',
# curriculum_agent_qa_model_name='models/gemini-1.5-flash',
# skill_manager_model_name='models/gemini-1.5-flash',
# action_agent_model_name='mistral-large-latest',
# critic_agent_model_name='mistral-large-latest',
# curriculum_agent_model_name='mistral-large-latest',
# curriculum_agent_qa_model_name='mistral-large-latest',
# skill_manager_model_name='mistral-large-latest',
# action_agent_model_name='claude-3-sonnet@20240229',
# critic_agent_model_name='claude-3-sonnet@20240229',
# curriculum_agent_model_name='claude-3-sonnet@20240229',
# curriculum_agent_qa_model_name='claude-3-sonnet@20240229',
# skill_manager_model_name='claude-3-sonnet@20240229',/
# action_agent_model_name='claude-3-5-sonnet@20240620',
# critic_agent_model_name='claude-3-5-sonnet@20240620',
# curriculum_agent_model_name='claude-3-5-sonnet@20240620',
# curriculum_agent_qa_model_name='claude-3-5-sonnet@20240620',
# skill_manager_model_name='claude-3-5-sonnet@20240620',
# action_agent_model_name='claude-3-opus@20240229',
# critic_agent_model_name='claude-3-opus@20240229',
# curriculum_agent_model_name='claude-3-opus@20240229',
# curriculum_agent_qa_model_name='claude-3-opus@20240229',
# skill_manager_model_name='claude-3-opus@20240229',
# action_agent_model_name='command-r-plus',
# critic_agent_model_name='command-r-plus',
# curriculum_agent_model_name='command-r-plus',
# curriculum_agent_qa_model_name='command-r-plus',
# skill_manager_model_name='command-r-plus',
# action_agent_model_name='reka-core',
# critic_agent_model_name='reka-core',
# curriculum_agent_model_name='reka-core',
# curriculum_agent_qa_model_name='reka-core',
# skill_manager_model_name='reka-core',
)
# start lifelong learning
voyager.learn()
# seed for training: -2171441100564293352 (Gpt-4 turbo didn't do well on this seed)
# for gemini pro, use 'models/gemini-1.0-pro-latest' for the latest model
# looks like I can use gemini ultra w/ 'models/gemini-1.0-ultra-latest' and gemini pro 1.5 w/ 'gemini-1.5-pro-latest', best to use 'models/embedding' for embedding models,
# gpt-4 should be 'gpt-4-0125-preview', gpt-4-turbo (worse) should be 'gpt-4-turbo-preview'
# for mistral, it is 'mistral-large-latest'
# for claude, it's 'claude-3-sonnet@20240229'