From 56e557c6ff8a6480887c9c585bf908045ee8e791 Mon Sep 17 00:00:00 2001
From: AUTOMATIC <16777216c@gmail.com>
Date: Sat, 24 Dec 2022 22:39:00 +0300
Subject: added cheap NN approximation for VAE

---
 modules/sd_vae_approx.py | 58 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 58 insertions(+)
 create mode 100644 modules/sd_vae_approx.py

(limited to 'modules/sd_vae_approx.py')

diff --git a/modules/sd_vae_approx.py b/modules/sd_vae_approx.py
new file mode 100644
index 00000000..0a58542d
--- /dev/null
+++ b/modules/sd_vae_approx.py
@@ -0,0 +1,58 @@
+import os
+
+import torch
+from torch import nn
+from modules import devices, paths
+
+sd_vae_approx_model = None
+
+
+class VAEApprox(nn.Module):
+    def __init__(self):
+        super(VAEApprox, self).__init__()
+        self.conv1 = nn.Conv2d(4, 8, (7, 7))
+        self.conv2 = nn.Conv2d(8, 16, (5, 5))
+        self.conv3 = nn.Conv2d(16, 32, (3, 3))
+        self.conv4 = nn.Conv2d(32, 64, (3, 3))
+        self.conv5 = nn.Conv2d(64, 32, (3, 3))
+        self.conv6 = nn.Conv2d(32, 16, (3, 3))
+        self.conv7 = nn.Conv2d(16, 8, (3, 3))
+        self.conv8 = nn.Conv2d(8, 3, (3, 3))
+
+    def forward(self, x):
+        extra = 11
+        x = nn.functional.interpolate(x, (x.shape[2] * 2, x.shape[3] * 2))
+        x = nn.functional.pad(x, (extra, extra, extra, extra))
+
+        for layer in [self.conv1, self.conv2, self.conv3, self.conv4, self.conv5, self.conv6, self.conv7, self.conv8, ]:
+            x = layer(x)
+            x = nn.functional.leaky_relu(x, 0.1)
+
+        return x
+
+
+def model():
+    global sd_vae_approx_model
+
+    if sd_vae_approx_model is None:
+        sd_vae_approx_model = VAEApprox()
+        sd_vae_approx_model.load_state_dict(torch.load(os.path.join(paths.models_path, "VAE-approx", "model.pt")))
+        sd_vae_approx_model.eval()
+        sd_vae_approx_model.to(devices.device, devices.dtype)
+
+    return sd_vae_approx_model
+
+
+def cheap_approximation(sample):
+    # https://discuss.huggingface.co/t/decoding-latents-to-rgb-without-upscaling/23204/2
+
+    coefs = torch.tensor([
+        [0.298, 0.207, 0.208],
+        [0.187, 0.286, 0.173],
+        [-0.158, 0.189, 0.264],
+        [-0.184, -0.271, -0.473],
+    ]).to(sample.device)
+
+    x_sample = torch.einsum("lxy,lr -> rxy", sample, coefs)
+
+    return x_sample
-- 
cgit v1.2.3