support for generating images on video cards with 4GB

webui.py

@@ -2,6 +2,8 @@ import argparse
 import os
 import sys
 from collections import namedtuple
+from contextlib import nullcontext
+
 import torch
 import torch.nn as nn
 import numpy as np
@@ -51,6 +53,7 @@ parser.add_argument("--no-progressbar-hiding", action='store_true', help="do not
 parser.add_argument("--max-batch-count", type=int, default=16, help="maximum batch count value for the UI")
 parser.add_argument("--embeddings-dir", type=str, default='embeddings', help="embeddings dirtectory for textual inversion (default: embeddings)")
 parser.add_argument("--allow-code", action='store_true', help="allow custom script execution from webui")
+parser.add_argument("--lowvram", action='store_true', help="enamble optimizations for low vram")
 
 cmd_opts = parser.parse_args()
 
@@ -185,11 +188,80 @@ def load_model_from_config(config, ckpt, verbose=False):
         print("unexpected keys:")
         print(u)
 
-    model.cuda()
     model.eval()
     return model
 
 
+module_in_gpu = None
+
+
+def setup_for_low_vram(sd_model):
+    parents = {}
+
+    def send_me_to_gpu(module, _):
+        """send this module to GPU; send whatever tracked module was previous in GPU to CPU;
+        we add this as forward_pre_hook to a lot of modules and this way all but one of them will
+        be in CPU
+        """
+        global module_in_gpu
+
+        module = parents.get(module, module)
+
+        if module_in_gpu == module:
+            return
+
+        if module_in_gpu is not None:
+            print('removing from gpu:', type(module_in_gpu))
+            module_in_gpu.to(cpu)
+
+        print('adding to gpu:', type(module))
+        module.to(gpu)
+
+        print('added to gpu:', type(module))
+        module_in_gpu = module
+
+    # see below for register_forward_pre_hook;
+    # first_stage_model does not use forward(), it uses encode/decode, so register_forward_pre_hook is
+    # useless here, and we just replace those methods
+    def first_stage_model_encode_wrap(self, encoder, x):
+        send_me_to_gpu(self, None)
+        return encoder(x)
+
+    def first_stage_model_decode_wrap(self, decoder, z):
+        send_me_to_gpu(self, None)
+        return decoder(z)
+
+    # remove three big modules, cond, first_stage, and unet from the model and then
+    # send the model to GPU. Then put modules back. the modules will be in CPU.
+    stored = sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.model
+    sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.model = None, None, None
+    sd_model.to(device)
+    sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.model = stored
+
+    # register hooks for those the first two models
+    sd_model.cond_stage_model.transformer.register_forward_pre_hook(send_me_to_gpu)
+    sd_model.first_stage_model.register_forward_pre_hook(send_me_to_gpu)
+    sd_model.first_stage_model.encode = lambda x, en=sd_model.first_stage_model.encode: first_stage_model_encode_wrap(sd_model.first_stage_model, en, x)
+    sd_model.first_stage_model.decode = lambda z, de=sd_model.first_stage_model.decode: first_stage_model_decode_wrap(sd_model.first_stage_model, de, z)
+    parents[sd_model.cond_stage_model.transformer] = sd_model.cond_stage_model
+
+    # the third remaining model is still too big for 4GB, so we also do the same for its submodules
+    # so that only one of them is in GPU at a time
+    diff_model = sd_model.model.diffusion_model
+    stored = diff_model.input_blocks, diff_model.middle_block, diff_model.output_blocks, diff_model.time_embed
+    diff_model.input_blocks, diff_model.middle_block, diff_model.output_blocks, diff_model.time_embed = None, None, None, None
+    sd_model.model.to(device)
+    diff_model.input_blocks, diff_model.middle_block, diff_model.output_blocks, diff_model.time_embed = stored
+
+    # install hooks for bits of third model
+    diff_model.time_embed.register_forward_pre_hook(send_me_to_gpu)
+    for block in diff_model.input_blocks:
+        block.register_forward_pre_hook(send_me_to_gpu)
+    diff_model.middle_block.register_forward_pre_hook(send_me_to_gpu)
+    for block in diff_model.output_blocks:
+        block.register_forward_pre_hook(send_me_to_gpu)
+
+
 def create_random_tensors(shape, seeds):
     xs = []
     for seed in seeds:
@@ -838,7 +910,8 @@ def process_images(p: StableDiffusionProcessing) -> Processed:
         model_hijack.load_textual_inversion_embeddings(cmd_opts.embeddings_dir, model)
 
     output_images = []
-    with torch.no_grad(), autocast("cuda"), model.ema_scope():
+    ema_scope = (nullcontext if cmd_opts.lowvram else model.ema_scope)
+    with torch.no_grad(), autocast("cuda"), ema_scope():
         p.init()
 
         for n in range(p.n_iter):
@@ -1327,8 +1400,17 @@ interfaces = [
 sd_config = OmegaConf.load(cmd_opts.config)
 sd_model = load_model_from_config(sd_config, cmd_opts.ckpt)
 
-device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
-sd_model = (sd_model if cmd_opts.no_half else sd_model.half()).to(device)
+cpu = torch.device("cpu")
+gpu = torch.device("cuda")
+device = gpu if torch.cuda.is_available() else cpu
+
+sd_model = (sd_model if cmd_opts.no_half else sd_model.half())
+
+if not cmd_opts.lowvram:
+    sd_model = sd_model.to(device)
+
+else:
+    setup_for_low_vram(sd_model)
 
 model_hijack = StableDiffusionModelHijack()
 model_hijack.hijack(sd_model)