Spaces:
Running
Running
Upload config.yaml with huggingface_hub
Browse files- config.yaml +38 -0
config.yaml
ADDED
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
dataset:
|
2 |
+
tokenizer_type: CharBPE
|
3 |
+
context_length: 64
|
4 |
+
image_resolution: 256
|
5 |
+
|
6 |
+
stage1:
|
7 |
+
type: vqgan
|
8 |
+
embed_dim: 256
|
9 |
+
n_embed: 16384
|
10 |
+
hparams:
|
11 |
+
double_z: False
|
12 |
+
z_channels: 256
|
13 |
+
resolution: 256
|
14 |
+
in_channels: 3
|
15 |
+
out_ch: 3
|
16 |
+
ch: 128
|
17 |
+
ch_mult: [1, 1, 2, 2, 4]
|
18 |
+
num_res_blocks: 2
|
19 |
+
attn_resolutions: [16]
|
20 |
+
pdrop: 0.0
|
21 |
+
|
22 |
+
stage2:
|
23 |
+
type: transformer1d
|
24 |
+
vocab_size_txt: 16384
|
25 |
+
vocab_size_img: 16384
|
26 |
+
hparams:
|
27 |
+
embed_dim: 1536
|
28 |
+
n_layers: 42
|
29 |
+
n_heads: 24
|
30 |
+
n_dense_layers: 42
|
31 |
+
ctx_len_img: 256
|
32 |
+
ctx_len_txt: 64
|
33 |
+
embd_pdrop: 0.0
|
34 |
+
resid_pdrop: 0.0
|
35 |
+
attn_pdrop: 0.0
|
36 |
+
mlp_bias: True
|
37 |
+
attn_bias: True
|
38 |
+
gelu_use_approx: False
|