Updates

.gitignore

@@ -2,7 +2,7 @@
 __pycache__/
 *.py[cod]
 *$py.class
-
+.idea/*
 # C extensions
 *.so
 

scripts/patch_fixer.py

@@ -4,9 +4,12 @@ import os.path
 import traceback
 from collections import namedtuple
 
+import ldm.modules.attention
 import torch
+from einops import rearrange, repeat
 from ldm.util import instantiate_from_config
 from omegaconf import OmegaConf
+from torch import einsum
 
 from modules import shared, devices, script_callbacks, sd_models
 from modules.paths import models_path
@@ -21,11 +24,56 @@ CheckpointInfo = namedtuple("CheckpointInfo", ['filename', 'title', 'hash', 'mod
 checkpoints_list = {}
 checkpoints_loaded = collections.OrderedDict()
 
+_ATTN_PRECISION = os.environ.get("ATTN_PRECISION", "fp32")
+
+
+def exists(val):
+    return val is not None
+
+
+def default(val, d):
+    return val if exists(val) else d
+
+
+def fixed_forward(self, x, context=None, mask=None):
+    h = self.heads
+
+    q = self.to_q(x)
+    context = default(context, x)
+    k = self.to_k(context)
+    v = self.to_v(context)
+
+    q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))
+
+    # force cast to fp32 to avoid overflowing
+    if _ATTN_PRECISION == "fp32":
+        with torch.autocast(enabled=False, device_type='cuda'):
+            q, k = q.float(), k.float()
+            sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+    else:
+        sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+
+    del q, k
+
+    if exists(mask):
+        mask = rearrange(mask, 'b ... -> b (...)')
+        max_neg_value = -torch.finfo(sim.dtype).max
+        mask = repeat(mask, 'b j -> (b h) () j', h=h)
+        sim.masked_fill_(~mask, max_neg_value)
+
+    # attention, what we cannot get enough of
+    sim = sim.softmax(dim=-1)
+
+    out = einsum('b i j, b j d -> b i d', sim, v)
+    out = rearrange(out, '(b h) n d -> b n (h d)', h=h)
+    return self.to_out(out)
+
 
 def get_config(checkpoint_info):
     path = checkpoint_info[0]
     model_config = checkpoint_info.config
     checkpoint_dir = os.path.join(shared.script_path, "extensions", "sd_auto_fix", "configs")
+    is_v21 = False
     if model_config == shared.cmd_opts.config:
         try:
             checkpoint = torch.load(path)
@@ -38,18 +86,24 @@ def get_config(checkpoint_info):
                     else:
                         model_config = os.path.join(checkpoint_dir, "v2-inference-v.yaml")
                 print(f"V2 Model detected, selecting model config: {model_config}")
+            v21_keys = ["callbacks", "lr_schedulers", "native_amp_scaling_state"]
+            for v22key in v21_keys:
+                if v22key in checkpoint:
+                    is_v21 = True
+                    break
+
             del checkpoint
         except Exception as e:
             print(f"Exception: {e}")
             traceback.print_exc()
             pass
-    return model_config
+    return model_config, is_v21
 
 
 def load_model(checkpoint_info=None):
     from modules import lowvram, sd_hijack
     checkpoint_info = checkpoint_info or select_checkpoint()
-    model_config = get_config(checkpoint_info)
+    model_config, is_v21 = get_config(checkpoint_info)
 
     if model_config != shared.cmd_opts.config:
         print(f"Loading config from: {model_config}")
@@ -74,9 +128,6 @@ def load_model(checkpoint_info=None):
 
     do_inpainting_hijack()
 
-    if shared.cmd_opts.no_half:
-        sd_config.model.params.unet_config.params.use_fp16 = False
-
     sd_model = instantiate_from_config(sd_config.model)
     load_model_weights(sd_model, checkpoint_info)
 
@@ -87,6 +138,10 @@ def load_model(checkpoint_info=None):
 
     sd_hijack.model_hijack.hijack(sd_model)
 
+    if is_v21 and not shared.cmd_opts.xformers and not shared.cmd_opts.force_enable_xformers and not shared.cmd_opts.no_half:
+        print("Fixing attention for v21 model.")
+        ldm.modules.attention.CrossAttention.forward = fixed_forward
+
     sd_model.eval()
     shared.sd_model = sd_model
 
@@ -106,7 +161,7 @@ def reload_model_weights(sd_model=None, info=None):
     if sd_model.sd_model_checkpoint == checkpoint_info.filename:
         return
 
-    model_config = get_config(checkpoint_info)
+    model_config, is_v21 = get_config(checkpoint_info)
     checkpoint_info = checkpoint_info._replace(config=model_config)
     if sd_model.sd_checkpoint_info.config != model_config or should_hijack_inpainting(
             checkpoint_info) != should_hijack_inpainting(sd_model.sd_checkpoint_info):
@@ -125,6 +180,11 @@ def reload_model_weights(sd_model=None, info=None):
     load_model_weights(sd_model, checkpoint_info)
 
     sd_hijack.model_hijack.hijack(sd_model)
+
+    if is_v21 and not shared.cmd_opts.xformers and not shared.cmd_opts.force_enable_xformers and not shared.cmd_opts.no_half:
+        print("Fixing attention for v21 model.")
+        ldm.modules.attention.CrossAttention.forward = fixed_forward
+
     script_callbacks.model_loaded_callback(sd_model)
 
     if not shared.cmd_opts.lowvram and not shared.cmd_opts.medvram: