fixes #3449 - VRAM leak when switching to/from inpainting model

modules/sd_samplers.py

@@ -1,4 +1,4 @@
-from collections import namedtuple
+from collections import namedtuple, deque
 import numpy as np
 from math import floor
 import torch
@@ -335,18 +335,28 @@ class CFGDenoiser(torch.nn.Module):
 
 
 class TorchHijack:
-    def __init__(self, kdiff_sampler):
-        self.kdiff_sampler = kdiff_sampler
+    def __init__(self, sampler_noises):
+        # Using a deque to efficiently receive the sampler_noises in the same order as the previous index-based
+        # implementation.
+        self.sampler_noises = deque(sampler_noises)
 
     def __getattr__(self, item):
         if item == 'randn_like':
-            return self.kdiff_sampler.randn_like
+            return self.randn_like
 
         if hasattr(torch, item):
             return getattr(torch, item)
 
         raise AttributeError("'{}' object has no attribute '{}'".format(type(self).__name__, item))
 
+    def randn_like(self, x):
+        if self.sampler_noises:
+            noise = self.sampler_noises.popleft()
+            if noise.shape == x.shape:
+                return noise
+
+        return torch.randn_like(x)
+
 
 class KDiffusionSampler:
     def __init__(self, funcname, sd_model):
@@ -356,7 +366,6 @@ class KDiffusionSampler:
         self.extra_params = sampler_extra_params.get(funcname, [])
         self.model_wrap_cfg = CFGDenoiser(self.model_wrap)
         self.sampler_noises = None
-        self.sampler_noise_index = 0
         self.stop_at = None
         self.eta = None
         self.default_eta = 1.0
@@ -389,26 +398,14 @@ class KDiffusionSampler:
     def number_of_needed_noises(self, p):
         return p.steps
 
-    def randn_like(self, x):
-        noise = self.sampler_noises[self.sampler_noise_index] if self.sampler_noises is not None and self.sampler_noise_index < len(self.sampler_noises) else None
-
-        if noise is not None and x.shape == noise.shape:
-            res = noise
-        else:
-            res = torch.randn_like(x)
-
-        self.sampler_noise_index += 1
-        return res
-
     def initialize(self, p):
         self.model_wrap_cfg.mask = p.mask if hasattr(p, 'mask') else None
         self.model_wrap_cfg.nmask = p.nmask if hasattr(p, 'nmask') else None
         self.model_wrap.step = 0
-        self.sampler_noise_index = 0
         self.eta = p.eta or opts.eta_ancestral
 
         if self.sampler_noises is not None:
-            k_diffusion.sampling.torch = TorchHijack(self)
+            k_diffusion.sampling.torch = TorchHijack(self.sampler_noises)
 
         extra_params_kwargs = {}
         for param_name in self.extra_params: