Upload folder using huggingface_hub
Browse files- checkpoints/128.pt +3 -0
- config.json +28 -0
- config.yaml +45 -0
checkpoints/128.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c14aebf27004cfa6bf24fe5ce3df61d69650c3085a8b1c866ad955b7b99089e4
|
3 |
+
size 2609241188
|
config.json
ADDED
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"dim": 1024,
|
3 |
+
"n_layer": 24,
|
4 |
+
"n_head": 16,
|
5 |
+
"n_kv_head": null,
|
6 |
+
"multiple_of": 256,
|
7 |
+
"ffn_dim_multiplier": null,
|
8 |
+
"rope_base": 10000,
|
9 |
+
"norm_eps": 1e-05,
|
10 |
+
"initializer_range": 0.02,
|
11 |
+
"token_dropout_p": 0.1,
|
12 |
+
"attn_dropout_p": 0.0,
|
13 |
+
"resid_dropout_p": 0.1,
|
14 |
+
"ffn_dropout_p": 0.1,
|
15 |
+
"drop_path_rate": 0.0,
|
16 |
+
"num_classes": 1000,
|
17 |
+
"caption_dim": 2048,
|
18 |
+
"class_dropout_prob": 0.1,
|
19 |
+
"model_type": "c2i",
|
20 |
+
"vocab_size": 8192,
|
21 |
+
"cls_token_num": 1,
|
22 |
+
"block_size": 128,
|
23 |
+
"max_batch_size": 32,
|
24 |
+
"max_seq_len": 2048,
|
25 |
+
"learnable_tok_embeddings": true,
|
26 |
+
"group_size": 8,
|
27 |
+
"causal_num": null
|
28 |
+
}
|
config.yaml
ADDED
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: configs/ar_l.yaml
|
2 |
+
code_path:
|
3 |
+
gpt_model: GPT-L
|
4 |
+
gpt_ckpt:
|
5 |
+
finetune: false
|
6 |
+
model_weight_strict: true
|
7 |
+
gpt_type: c2i
|
8 |
+
ema: true
|
9 |
+
cls_token_num: 1
|
10 |
+
dropout_p: 0.1
|
11 |
+
token_dropout_p: 0.1
|
12 |
+
drop_path_rate: 0.0
|
13 |
+
compile: true
|
14 |
+
results_dir:
|
15 |
+
data_path: ImageNet2012/train
|
16 |
+
dataset: imagenet_code
|
17 |
+
vq_path:
|
18 |
+
vq_ema: false
|
19 |
+
image_size:
|
20 |
+
num_classes: 1000
|
21 |
+
epochs: 300
|
22 |
+
lr: 1e-4
|
23 |
+
warmup_steps:
|
24 |
+
warmup_ratio: 0.05
|
25 |
+
weight_decay: 5e-2
|
26 |
+
beta1: 0.9
|
27 |
+
beta2: 0.95
|
28 |
+
max_grad_norm: 1.0
|
29 |
+
global_batch_size: 512
|
30 |
+
global_seed: 0
|
31 |
+
num_workers: 24
|
32 |
+
log_every: 50
|
33 |
+
ckpt_every: 5000
|
34 |
+
save_epochs: 2
|
35 |
+
mixed_precision: bf16
|
36 |
+
learnable_tok_embeddings: true
|
37 |
+
resume_from_newest_ckpt: true
|
38 |
+
debug_mode: false
|
39 |
+
correction_prob: 0.3
|
40 |
+
rank: 0
|
41 |
+
world_size: 16
|
42 |
+
gpu: 0
|
43 |
+
dist_url: env://
|
44 |
+
distributed: true
|
45 |
+
dist_backend: nccl
|