File size: 753 Bytes
ce24f5e
 
 
77762a5
ce24f5e
e50a64e
ce24f5e
 
949a27b
77762a5
 
ce24f5e
 
 
8d959a7
77762a5
8d959a7
77762a5
7019509
ce24f5e
a1da39c
abddcf4
77762a5
c2a0792
8d959a7
8b79ff0
949a27b
ce24f5e
8d959a7
782b6a4
ace70b3
d1aed4c
ce24f5e
 
77762a5
5f79b82
77762a5
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
base_model: EleutherAI/pythia-1.4b-deduped
load_in_8bit: true
datasets:
  - path: teknium/GPT4-LLM-Cleaned
    type: alpaca
dataset_prepared_path:
val_set_size: 0.05
adapter: lora
lora_model_dir:
sequence_len: 512
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_modules:
  - query_key_value
lora_target_linear:
lora_fan_in_fan_out: true  # pythia/GPTNeoX lora specific
wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
output_dir: ./lora-alpaca-pythia
gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 4
learning_rate: 0.00001
train_on_inputs: false
group_by_length: false
bf16: auto
tf32: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
weight_decay: 0.1
evals_per_epoch: 4
logging_steps: 1