Upload 2 files
Browse files- config.yaml +32 -0
- image_adapter.pt +3 -0
config.yaml
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
wandb_project: joy-caption-1
|
2 |
+
device_batch_size: 2
|
3 |
+
batch_size: 256
|
4 |
+
learning_rate: 0.001
|
5 |
+
warmup_samples: 18000
|
6 |
+
max_samples: 600000
|
7 |
+
save_every: 50000
|
8 |
+
test_every: 50000
|
9 |
+
use_amp: true
|
10 |
+
grad_scaler: true
|
11 |
+
lr_scheduler_type: cosine
|
12 |
+
min_lr_ratio: 0.0
|
13 |
+
allow_tf32: true
|
14 |
+
seed: 42
|
15 |
+
num_workers: 8
|
16 |
+
optimizer_type: adamw
|
17 |
+
adam_beta1: 0.9
|
18 |
+
adam_beta2: 0.999
|
19 |
+
adam_eps: 1.0e-08
|
20 |
+
adam_weight_decay: 0.0
|
21 |
+
clip_grad_norm: 1.0
|
22 |
+
dataset: fancyfeast/joy-captioning-20240729a
|
23 |
+
clip_model: google/siglip-so400m-patch14-384
|
24 |
+
text_model: meta-llama/Meta-Llama-3.1-8B
|
25 |
+
resume: null
|
26 |
+
gradient_checkpointing: false
|
27 |
+
test_size: 2048
|
28 |
+
grad_scaler_init: 65536.0
|
29 |
+
max_caption_length: 257
|
30 |
+
num_image_tokens: 32
|
31 |
+
adapter_type: mlp
|
32 |
+
text_model_dtype: float16
|
image_adapter.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2ebb1d1437bbb3264a6f25a896b25a7c7dd06c570c5de909dc2f19d3a5c5c110
|
3 |
+
size 86018240
|