File size: 1,208 Bytes
6ef46f8
 
 
a7a9a14
6ef46f8
 
 
 
782b6a4
 
6ef46f8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5a5d474
6ef46f8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
base_model: 01-ai/Yi-34B-Chat
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false
sequence_len: 1024
bf16: auto
fp16:
tf32: false
flash_attention: true
special_tokens:
  bos_token: "<|startoftext|>"
  eos_token: "<|endoftext|>"
  unk_token: "<unk>"

# Data
datasets:
  - path: mhenrichsen/alpaca_2k_test
    type: alpaca
warmup_steps: 10

# Iterations
num_epochs: 1

# Evaluation
val_set_size: 0.1
evals_per_epoch: 5
eval_table_size:
eval_max_new_tokens: 128
eval_sample_packing: false
eval_batch_size: 1

# LoRA
output_dir: ./qlora-out
adapter: qlora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:

# Sampling
sample_packing: false
pad_to_sequence_len: false

# Batching
gradient_accumulation_steps: 4
micro_batch_size: 1
gradient_checkpointing: true

# wandb
wandb_project:

# Optimizer
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 0.0002

# Misc
train_on_inputs: false
group_by_length: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
debug:
deepspeed:
weight_decay: 0
fsdp:
fsdp_config: