Add hypernetwork support to split cross attention v1

* Add hypernetwork support to split_cross_attention_forward_v1
* Fix device check in esrgan_model.py to use devices.device_esrgan instead of shared.device

2 files changed, 15 insertions, 5 deletions

diff --git a/modules/esrgan_model.py b/modules/esrgan_model.py
index d17e730f..28548124 100644
--- a/modules/esrgan_model.py
+++ b/modules/esrgan_model.py
@@ -111,7 +111,7 @@ class UpscalerESRGAN(Upscaler):
             print("Unable to load %s from %s" % (self.model_path, filename))

             return None

 

-        pretrained_net = torch.load(filename, map_location='cpu' if shared.device.type == 'mps' else None)

+        pretrained_net = torch.load(filename, map_location='cpu' if devices.device_esrgan.type == 'mps' else None)

         crt_model = arch.RRDBNet(3, 3, 64, 23, gc=32)

 

         pretrained_net = fix_model_layers(crt_model, pretrained_net)

diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py
index d9cca485..3351c740 100644
--- a/modules/sd_hijack_optimizations.py
+++ b/modules/sd_hijack_optimizations.py
@@ -12,13 +12,22 @@ from modules import shared
 def split_cross_attention_forward_v1(self, x, context=None, mask=None):

     h = self.heads

 

-    q = self.to_q(x)

+    q_in = self.to_q(x)

     context = default(context, x)

-    k = self.to_k(context)

-    v = self.to_v(context)

+

+    hypernetwork = shared.selected_hypernetwork()

+    hypernetwork_layers = (hypernetwork.layers if hypernetwork is not None else {}).get(context.shape[2], None)

+

+    if hypernetwork_layers is not None:

+        k_in = self.to_k(hypernetwork_layers[0](context))

+        v_in = self.to_v(hypernetwork_layers[1](context))

+    else:

+        k_in = self.to_k(context)

+        v_in = self.to_v(context)

     del context, x

 

-    q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))

+    q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q_in, k_in, v_in))

+    del q_in, k_in, v_in

 

     r1 = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device)

     for i in range(0, q.shape[0], 2):

@@ -31,6 +40,7 @@ def split_cross_attention_forward_v1(self, x, context=None, mask=None):
 

         r1[i:end] = einsum('b i j, b j d -> b i d', s2, v[i:end])

         del s2

+    del q, k, v

 

     r2 = rearrange(r1, '(b h) n d -> b n (h d)', h=h)

     del r1