-
Notifications
You must be signed in to change notification settings - Fork 0
/
config.py
64 lines (43 loc) · 1.79 KB
/
config.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
from os import cpu_count
# Location of the LLM file to serve as PFC
model_path = r"llama-2-13b-chat.Q8_0.gguf"
# Model temperature
model_temp = 1
# Time between last interaction and activating Default Mode Network (in seconds)
dmn_countdown = 120
# Time between last input and ending interaction session
interaction_timeout = 360
# End conversation keyword to be used to avoid waiting till interaction timeout
interaction_break = 'end_chat'
# Number of threads to be used in self-finetuning session
available_threads = cpu_count()
# Number of self-finetuning session training materials to be generated
dreams_to_generate_num = 150
# Location of folder with finetune-realted binaries
finetune_dir = r"finetune_bins"
# Binary used for finetuning
finetune_tool = r"finetune"
# Binary used for merging LoRA weights with the base model
lora_tool = r"export-lora"
# Number of epochs in a self-finetuning session
epochs = 5
# File with lora parameters to integrate into base model
lora_to_integrate = r"ggml-lora-LATEST-f32.gguf"
# Weight of lora parameters at integration into base model
lora_weight = 1
# Granularity level of logs saved to a file
file_log_level = 10
# Granularity level of logs printed to console
console_log_level = 10
# Directory containing conversation's histories
conversations_dir = r"conversations"
# Directory containing conversation conclusions
conclusions_dir = r"conclusions"
# Directory containing generated training materials
context_dir = r"dreams"
#Log directory
log_dir = r"logs"
# Location of the JSON file serving as short term memory
stm_path = r"conversations/short-term-memory.json"
# Markers dividing different parts of generated dreams - needs to be aligned with a relevant prompt
dream_markers = {'stimulus': '**QUESTION**', 'reaction': '**RESPONSE**', 'end': '**END**'}