fix lowvram (#562)

fooocus_extras/ip_adapter.py

@@ -213,6 +213,7 @@ def patch_model(model, ip_tasks):
             org_dtype = n.dtype
             current_step = float(model.model.diffusion_model.current_step.detach().cpu().numpy()[0])
             cond_or_uncond = extra_options['cond_or_uncond']
+            batch_size = int(context_attn2.shape[0])
 
             with torch.autocast("cuda", dtype=ip_adapter.dtype):
                 q = n
@@ -223,9 +224,16 @@ def patch_model(model, ip_tasks):
 
                 for cn_img, cn_stop, cn_weight, cache in tasks:
                     if current_step < cn_stop:
+                        ip_k, ip_v = None, None
+
                         if ip_index in cache:
                             ip_k, ip_v = cache[ip_index]
-                        else:
+                            if int(ip_k.shape[0]) != batch_size:
+                                ip_k = None
+                            if int(ip_v.shape[0]) != batch_size:
+                                ip_v = None
+
+                        if ip_k is None or ip_v is None:
                             ip_model_k.to(device=ip_adapter.load_device, dtype=ip_adapter.dtype)
                             ip_model_v.to(device=ip_adapter.load_device, dtype=ip_adapter.dtype)
                             cond = cn_img.to(device=ip_adapter.load_device, dtype=ip_adapter.dtype).repeat(batch_prompt, 1, 1)

fooocus_version.py

@@ -1 +1 @@
-version = '2.1.1'
+version = '2.1.2'