Update src/model.py

src/model.py

@@ -1,92 +1,60 @@
-import torch
-import torch as th
-import torch.nn as nn
-import torch.nn.functional as F
+import torch as t, torch.nn as nn, torch.nn.functional as F
 
-def conv(n_in, n_out, **kwargs):
-    return nn.Conv2d(n_in, n_out, 3, padding=1, **kwargs)
+def cv(n_i, n_o, **kw): return nn.Conv2d(n_i, n_o, 3, padding=1, **kw)
 
-class Clamp(nn.Module):
-    def forward(self, x):
-        return torch.tanh(x / 3) * 3
+class C(nn.Module):
+    def forward(self, x): return t.tanh(x / 3) * 3
 
-class Block(nn.Module):
-    def __init__(self, n_in, n_out):
+class B(nn.Module):
+    def __init__(s, n_i, n_o):
         super().__init__()
-        self.conv = nn.Sequential(conv(n_in, n_out), nn.ReLU(), conv(n_out, n_out), nn.ReLU(), conv(n_out, n_out))
-        self.skip = nn.Conv2d(n_in, n_out, 1, bias=False) if n_in != n_out else nn.Identity()
-        self.fuse = nn.ReLU()
-    def forward(self, x):
-        return self.fuse(self.conv(x) + self.skip(x))
+        s.c = nn.Sequential(cv(n_i, n_o), nn.ReLU(), cv(n_o, n_o), nn.ReLU(), cv(n_o, n_o))
+        s.s = nn.Conv2d(n_i, n_o, 1, bias=False) if n_i != n_o else nn.Identity()
+        s.f = nn.ReLU()
+    def forward(s, x): return s.f(s.c(x) + s.s(x))
 
-def Encoder(latent_channels=4):
+def E(lc=4):
     return nn.Sequential(
-        conv(3, 64), Block(64, 64),
-        conv(64, 64, stride=2, bias=False), Block(64, 64), Block(64, 64), Block(64, 64),
-        conv(64, 64, stride=2, bias=False), Block(64, 64), Block(64, 64), Block(64, 64),
-        conv(64, 64, stride=2, bias=False), Block(64, 64), Block(64, 64), Block(64, 64),
-        conv(64, latent_channels),
+        cv(3, 64), B(64, 64), cv(64, 64, stride=2, bias=False), B(64, 64), B(64, 64), B(64, 64),
+        cv(64, 64, stride=2, bias=False), B(64, 64), B(64, 64), B(64, 64),
+        cv(64, 64, stride=2, bias=False), B(64, 64), B(64, 64), B(64, 64),
+        cv(64, lc),
     )
 
-def Decoder(latent_channels=16):  # Adjusted to match expected input channels
+def D(lc=16):
     return nn.Sequential(
-        Clamp(), 
-        conv(latent_channels, 48),  # Reduced from 64 to 48 channels
-        nn.ReLU(),
-        Block(48, 48), Block(48, 48),  # Reduced number of blocks
-        nn.Upsample(scale_factor=2), conv(48, 48, bias=False),
-        Block(48, 48), Block(48, 48),  # Reduced number of blocks
-        nn.Upsample(scale_factor=2), conv(48, 48, bias=False),
-        Block(48, 48),  # Further reduction in blocks
-        nn.Upsample(scale_factor=2), conv(48, 48, bias=False),
-        Block(48, 48), 
-        conv(48, 3),  # Final convolution to output channels
+        C(), cv(lc, 48), nn.ReLU(), B(48, 48), B(48, 48), nn.Upsample(scale_factor=2),
+        cv(48, 48, bias=False), B(48, 48), B(48, 48), nn.Upsample(scale_factor=2),
+        cv(48, 48, bias=False), B(48, 48), nn.Upsample(scale_factor=2),
+        cv(48, 48, bias=False), B(48, 48), cv(48, 3),
     )
 
-        
+class M(nn.Module):
+    lm, ls = 3, 0.5
+    def __init__(s, ep="encoder.pth", dp="decoder.pth", lc=None):
+        super().__init__()
+        if lc is None: lc = s.glc(str(ep))
+        s.e, s.d = E(lc), D(lc)
 
-class Model(nn.Module):
-    latent_magnitude = 3
-    latent_shift = 0.5
+        def f(sd, mod, pfx):
+            f_sd = {k.strip(pfx): v for k, v in sd.items() if k.strip(pfx) in mod.state_dict() and v.size() == mod.state_dict()[k.strip(pfx)].size()}
+            print(f"num keys: {len(f_sd)} of {len(mod.state_dict())}")
+            mod.load_state_dict(f_sd, strict=False)
 
-    def __init__(self, encoder_path="encoder.pth", decoder_path="decoder.pth", latent_channels=None):
-        super().__init__()
-        if latent_channels is None:
-            latent_channels = self.guess_latent_channels(str(encoder_path))
-        self.encoder = Encoder(latent_channels)
-        self.decoder = Decoder(latent_channels)
-        
-        if encoder_path is not None:
-            encoder_state_dict = torch.load(encoder_path, map_location="cpu", weights_only=True)
-            filtered_state_dict = {k.strip('encoder.'): v for k, v in encoder_state_dict.items() if k.strip('encoder.') in self.encoder.state_dict() and v.size() == self.encoder.state_dict()[k.strip('encoder.')].size()}
-            print(f" num of keys in filtered: {len(filtered_state_dict)} and in decoder: {len(self.encoder.state_dict())}")
-            self.encoder.load_state_dict(filtered_state_dict, strict=False)
-            
-        if decoder_path is not None:
-            decoder_state_dict = torch.load(decoder_path, map_location="cpu", weights_only=True)
-            filtered_state_dict = {k.strip('decoder.'): v for k, v in decoder_state_dict.items() if k.strip('decoder.') in self.decoder.state_dict() and v.size() == self.decoder.state_dict()[k.strip('decoder.')].size()}
-            print(f" num of keys in filtered: {len(filtered_state_dict)} and in decoder: {len(self.decoder.state_dict())}")
-            self.decoder.load_state_dict(filtered_state_dict, strict=False)
-            
-        self.encoder.requires_grad_(False)
-        self.decoder.requires_grad_(False)
+        if ep: f(t.load(ep, map_location="cpu", weights_only=True), s.e, "encoder.")
+        if dp: f(t.load(dp, map_location="cpu", weights_only=True), s.d, "decoder.")
+
+        s.e.requires_grad_(False)
+        s.d.requires_grad_(False)
+
+    def glc(s, ep): return 16 if "taef1" in ep or "taesd3" in ep else 4
 
-    def guess_latent_channels(self, encoder_path):
-        if "taef1" in encoder_path:return 16
-        if "taesd3" in encoder_path:return 16
-        return 4
-        
     @staticmethod
-    def scale_latents(x):
-        return x.div(2 * Model.latent_magnitude).add(Model.latent_shift).clamp(0, 1)
+    def sl(x): return x.div(2 * M.lm).add(M.ls).clamp(0, 1)
 
     @staticmethod
-    def unscale_latents(x):
-        return x.sub(Model.latent_shift).mul(2 * Model.latent_magnitude)
+    def ul(x): return x.sub(M.ls).mul(2 * M.lm)
 
-    def forward(self, x, return_latent=False):
-        latent = self.encoder(x)
-        out = self.decoder(latent)
-        if return_latent:
-            return out.clamp(0, 1), latent
-        return out.clamp(0, 1)
+    def forward(s, x, rl=False):
+        l, o = s.e(x), s.d(s.e(x))
+        return (o.clamp(0, 1), l) if rl else o.clamp(0, 1)