Delta-Vector
commited on
Upload sdprompterv3.yaml
Browse files- sdprompterv3.yaml +73 -0
sdprompterv3.yaml
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
base_model: Delta-Vector/Holland-4B
|
2 |
+
model_type: AutoModelForCausalLM
|
3 |
+
tokenizer_type: AutoTokenizer
|
4 |
+
|
5 |
+
trust_remote_code: true
|
6 |
+
|
7 |
+
load_in_8bit: false
|
8 |
+
load_in_4bit: false
|
9 |
+
strict: false
|
10 |
+
|
11 |
+
datasets:
|
12 |
+
- path: NewEden/CivitAI-SD-Prompts
|
13 |
+
type: chatml
|
14 |
+
conversation: mpt-30b-instruct
|
15 |
+
chat_template: chatml
|
16 |
+
|
17 |
+
dataset_prepared_path:
|
18 |
+
val_set_size: 0.02
|
19 |
+
output_dir: ./outputs/out2
|
20 |
+
sequence_len: 8192
|
21 |
+
sample_packing: true
|
22 |
+
eval_sample_packing: false
|
23 |
+
pad_to_sequence_len: true
|
24 |
+
|
25 |
+
plugins:
|
26 |
+
- axolotl.integrations.liger.LigerPlugin
|
27 |
+
liger_rope: true
|
28 |
+
liger_rms_norm: true
|
29 |
+
liger_swiglu: true
|
30 |
+
liger_fused_linear_cross_entropy: true
|
31 |
+
|
32 |
+
wandb_project: SDprompterV3
|
33 |
+
wandb_entity:
|
34 |
+
wandb_watch:
|
35 |
+
wandb_name: SDprompterV3-attempt1
|
36 |
+
wandb_log_model:
|
37 |
+
|
38 |
+
hub_model_id: NewEden/SDprompterV3
|
39 |
+
hub_strategy: "all_checkpoints"
|
40 |
+
hf_use_auth_token: true
|
41 |
+
|
42 |
+
gradient_accumulation_steps: 2
|
43 |
+
micro_batch_size: 2
|
44 |
+
num_epochs: 4
|
45 |
+
optimizer: adamw_torch
|
46 |
+
lr_scheduler: cosine
|
47 |
+
learning_rate: 0.00001
|
48 |
+
|
49 |
+
train_on_inputs: false
|
50 |
+
group_by_length: false
|
51 |
+
bf16: auto
|
52 |
+
fp16:
|
53 |
+
tf32: true
|
54 |
+
|
55 |
+
gradient_checkpointing: true
|
56 |
+
gradient_checkpointing_kwargs:
|
57 |
+
use_reentrant: false
|
58 |
+
early_stopping_patience:
|
59 |
+
resume_from_checkpoint:
|
60 |
+
local_rank:
|
61 |
+
logging_steps: 1
|
62 |
+
xformers_attention:
|
63 |
+
flash_attention: true
|
64 |
+
|
65 |
+
warmup_ratio: 0.05
|
66 |
+
evals_per_epoch: 4
|
67 |
+
saves_per_epoch: 1
|
68 |
+
debug:
|
69 |
+
weight_decay: 0.05
|
70 |
+
deepspeed: /workspace/axolotl/deepspeed_configs/zero2.json
|
71 |
+
special_tokens:
|
72 |
+
pad_token: <|finetune_right_pad_id|>
|
73 |
+
eos_token: <|eot_id|>
|