From d782a95967c9eea753df3333cd1954b6ec73eba0 Mon Sep 17 00:00:00 2001 From: brkirch Date: Tue, 27 Dec 2022 08:50:55 -0500 Subject: Add Birch-san's sub-quadratic attention implementation --- modules/sd_hijack_optimizations.py | 124 +++++++++++++++++++++++++++++-------- 1 file changed, 99 insertions(+), 25 deletions(-) (limited to 'modules/sd_hijack_optimizations.py') diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 02c87f40..f5c153e8 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -1,7 +1,7 @@ import math import sys import traceback -import importlib +import psutil import torch from torch import einsum @@ -12,6 +12,8 @@ from einops import rearrange from modules import shared from modules.hypernetworks import hypernetwork +from .sub_quadratic_attention import efficient_dot_product_attention + if shared.cmd_opts.xformers or shared.cmd_opts.force_enable_xformers: try: @@ -22,6 +24,19 @@ if shared.cmd_opts.xformers or shared.cmd_opts.force_enable_xformers: print(traceback.format_exc(), file=sys.stderr) +def get_available_vram(): + if shared.device.type == 'cuda': + stats = torch.cuda.memory_stats(shared.device) + mem_active = stats['active_bytes.all.current'] + mem_reserved = stats['reserved_bytes.all.current'] + mem_free_cuda, _ = torch.cuda.mem_get_info(torch.cuda.current_device()) + mem_free_torch = mem_reserved - mem_active + mem_free_total = mem_free_cuda + mem_free_torch + return mem_free_total + else: + return psutil.virtual_memory().available + + # see https://github.com/basujindal/stable-diffusion/pull/117 for discussion def split_cross_attention_forward_v1(self, x, context=None, mask=None): h = self.heads @@ -76,12 +91,7 @@ def split_cross_attention_forward(self, x, context=None, mask=None): r1 = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device, dtype=q.dtype) - stats = torch.cuda.memory_stats(q.device) - mem_active = stats['active_bytes.all.current'] - mem_reserved = stats['reserved_bytes.all.current'] - mem_free_cuda, _ = torch.cuda.mem_get_info(torch.cuda.current_device()) - mem_free_torch = mem_reserved - mem_active - mem_free_total = mem_free_cuda + mem_free_torch + mem_free_total = get_available_vram() gb = 1024 ** 3 tensor_size = q.shape[0] * q.shape[1] * k.shape[1] * q.element_size() @@ -118,19 +128,8 @@ def split_cross_attention_forward(self, x, context=None, mask=None): return self.to_out(r2) -def check_for_psutil(): - try: - spec = importlib.util.find_spec('psutil') - return spec is not None - except ModuleNotFoundError: - return False - -invokeAI_mps_available = check_for_psutil() - # -- Taken from https://github.com/invoke-ai/InvokeAI and modified -- -if invokeAI_mps_available: - import psutil - mem_total_gb = psutil.virtual_memory().total // (1 << 30) +mem_total_gb = psutil.virtual_memory().total // (1 << 30) def einsum_op_compvis(q, k, v): s = einsum('b i d, b j d -> b i j', q, k) @@ -215,6 +214,70 @@ def split_cross_attention_forward_invokeAI(self, x, context=None, mask=None): # -- End of code from https://github.com/invoke-ai/InvokeAI -- + +# Based on Birch-san's modified implementation of sub-quadratic attention from https://github.com/Birch-san/diffusers/pull/1 +def sub_quad_attention_forward(self, x, context=None, mask=None): + assert mask is None, "attention-mask not currently implemented for SubQuadraticCrossAttnProcessor." + + h = self.heads + + q = self.to_q(x) + context = default(context, x) + + context_k, context_v = hypernetwork.apply_hypernetwork(shared.loaded_hypernetwork, context) + k = self.to_k(context_k) + v = self.to_v(context_v) + del context, context_k, context_v, x + + q = q.unflatten(-1, (h, -1)).transpose(1,2).flatten(end_dim=1) + k = k.unflatten(-1, (h, -1)).transpose(1,2).flatten(end_dim=1) + v = v.unflatten(-1, (h, -1)).transpose(1,2).flatten(end_dim=1) + + x = sub_quad_attention(q, k, v, q_chunk_size=shared.cmd_opts.sub_quad_q_chunk_size, kv_chunk_size=shared.cmd_opts.sub_quad_kv_chunk_size, chunk_threshold_bytes=shared.cmd_opts.sub_quad_chunk_threshold, use_checkpoint=self.training) + + x = x.unflatten(0, (-1, h)).transpose(1,2).flatten(start_dim=2) + + out_proj, dropout = self.to_out + x = out_proj(x) + x = dropout(x) + + return x + +def sub_quad_attention(q, k, v, q_chunk_size=1024, kv_chunk_size=None, kv_chunk_size_min=None, chunk_threshold_bytes=None, use_checkpoint=True): + bytes_per_token = torch.finfo(q.dtype).bits//8 + batch_x_heads, q_tokens, _ = q.shape + _, k_tokens, _ = k.shape + qk_matmul_size_bytes = batch_x_heads * bytes_per_token * q_tokens * k_tokens + + available_vram = int(get_available_vram() * 0.9) if q.device.type == 'mps' else int(get_available_vram() * 0.7) + + if chunk_threshold_bytes is None: + chunk_threshold_bytes = available_vram + elif chunk_threshold_bytes == 0: + chunk_threshold_bytes = None + + if kv_chunk_size_min is None: + kv_chunk_size_min = chunk_threshold_bytes // (batch_x_heads * bytes_per_token * (k.shape[2] + v.shape[2])) + elif kv_chunk_size_min == 0: + kv_chunk_size_min = None + + if chunk_threshold_bytes is not None and qk_matmul_size_bytes <= chunk_threshold_bytes: + # the big matmul fits into our memory limit; do everything in 1 chunk, + # i.e. send it down the unchunked fast-path + query_chunk_size = q_tokens + kv_chunk_size = k_tokens + + return efficient_dot_product_attention( + q, + k, + v, + query_chunk_size=q_chunk_size, + kv_chunk_size=kv_chunk_size, + kv_chunk_size_min = kv_chunk_size_min, + use_checkpoint=use_checkpoint, + ) + + def xformers_attention_forward(self, x, context=None, mask=None): h = self.heads q_in = self.to_q(x) @@ -252,12 +315,7 @@ def cross_attention_attnblock_forward(self, x): h_ = torch.zeros_like(k, device=q.device) - stats = torch.cuda.memory_stats(q.device) - mem_active = stats['active_bytes.all.current'] - mem_reserved = stats['reserved_bytes.all.current'] - mem_free_cuda, _ = torch.cuda.mem_get_info(torch.cuda.current_device()) - mem_free_torch = mem_reserved - mem_active - mem_free_total = mem_free_cuda + mem_free_torch + mem_free_total = get_available_vram() tensor_size = q.shape[0] * q.shape[1] * k.shape[2] * q.element_size() mem_required = tensor_size * 2.5 @@ -312,3 +370,19 @@ def xformers_attnblock_forward(self, x): return x + out except NotImplementedError: return cross_attention_attnblock_forward(self, x) + +def sub_quad_attnblock_forward(self, x): + h_ = x + h_ = self.norm(h_) + q = self.q(h_) + k = self.k(h_) + v = self.v(h_) + b, c, h, w = q.shape + q, k, v = map(lambda t: rearrange(t, 'b c h w -> b (h w) c'), (q, k, v)) + q = q.contiguous() + k = k.contiguous() + v = v.contiguous() + out = sub_quad_attention(q, k, v, q_chunk_size=shared.cmd_opts.sub_quad_q_chunk_size, kv_chunk_size=shared.cmd_opts.sub_quad_kv_chunk_size, chunk_threshold_bytes=shared.cmd_opts.sub_quad_chunk_threshold, use_checkpoint=self.training) + out = rearrange(out, 'b (h w) c -> b c h w', h=h) + out = self.proj_out(out) + return x + out -- cgit v1.2.3 From b95a4c0ce5ab9c414e0494193bfff665f45e9e65 Mon Sep 17 00:00:00 2001 From: brkirch Date: Fri, 6 Jan 2023 01:01:51 -0500 Subject: Change sub-quad chunk threshold to use percentage --- modules/sd_hijack_optimizations.py | 18 +++++++++--------- modules/shared.py | 2 +- 2 files changed, 10 insertions(+), 10 deletions(-) (limited to 'modules/sd_hijack_optimizations.py') diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index f5c153e8..b416e9ac 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -233,7 +233,7 @@ def sub_quad_attention_forward(self, x, context=None, mask=None): k = k.unflatten(-1, (h, -1)).transpose(1,2).flatten(end_dim=1) v = v.unflatten(-1, (h, -1)).transpose(1,2).flatten(end_dim=1) - x = sub_quad_attention(q, k, v, q_chunk_size=shared.cmd_opts.sub_quad_q_chunk_size, kv_chunk_size=shared.cmd_opts.sub_quad_kv_chunk_size, chunk_threshold_bytes=shared.cmd_opts.sub_quad_chunk_threshold, use_checkpoint=self.training) + x = sub_quad_attention(q, k, v, q_chunk_size=shared.cmd_opts.sub_quad_q_chunk_size, kv_chunk_size=shared.cmd_opts.sub_quad_kv_chunk_size, chunk_threshold=shared.cmd_opts.sub_quad_chunk_threshold, use_checkpoint=self.training) x = x.unflatten(0, (-1, h)).transpose(1,2).flatten(start_dim=2) @@ -243,20 +243,20 @@ def sub_quad_attention_forward(self, x, context=None, mask=None): return x -def sub_quad_attention(q, k, v, q_chunk_size=1024, kv_chunk_size=None, kv_chunk_size_min=None, chunk_threshold_bytes=None, use_checkpoint=True): +def sub_quad_attention(q, k, v, q_chunk_size=1024, kv_chunk_size=None, kv_chunk_size_min=None, chunk_threshold=None, use_checkpoint=True): bytes_per_token = torch.finfo(q.dtype).bits//8 batch_x_heads, q_tokens, _ = q.shape _, k_tokens, _ = k.shape qk_matmul_size_bytes = batch_x_heads * bytes_per_token * q_tokens * k_tokens - available_vram = int(get_available_vram() * 0.9) if q.device.type == 'mps' else int(get_available_vram() * 0.7) - - if chunk_threshold_bytes is None: - chunk_threshold_bytes = available_vram - elif chunk_threshold_bytes == 0: + if chunk_threshold is None: + chunk_threshold_bytes = int(get_available_vram() * 0.9) if q.device.type == 'mps' else int(get_available_vram() * 0.7) + elif chunk_threshold == 0: chunk_threshold_bytes = None + else: + chunk_threshold_bytes = int(0.01 * chunk_threshold * get_available_vram()) - if kv_chunk_size_min is None: + if kv_chunk_size_min is None and chunk_threshold_bytes is not None: kv_chunk_size_min = chunk_threshold_bytes // (batch_x_heads * bytes_per_token * (k.shape[2] + v.shape[2])) elif kv_chunk_size_min == 0: kv_chunk_size_min = None @@ -382,7 +382,7 @@ def sub_quad_attnblock_forward(self, x): q = q.contiguous() k = k.contiguous() v = v.contiguous() - out = sub_quad_attention(q, k, v, q_chunk_size=shared.cmd_opts.sub_quad_q_chunk_size, kv_chunk_size=shared.cmd_opts.sub_quad_kv_chunk_size, chunk_threshold_bytes=shared.cmd_opts.sub_quad_chunk_threshold, use_checkpoint=self.training) + out = sub_quad_attention(q, k, v, q_chunk_size=shared.cmd_opts.sub_quad_q_chunk_size, kv_chunk_size=shared.cmd_opts.sub_quad_kv_chunk_size, chunk_threshold=shared.cmd_opts.sub_quad_chunk_threshold, use_checkpoint=self.training) out = rearrange(out, 'b (h w) c -> b c h w', h=h) out = self.proj_out(out) return x + out diff --git a/modules/shared.py b/modules/shared.py index cb1dc312..d7a81db1 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -59,7 +59,7 @@ parser.add_argument("--opt-split-attention", action='store_true', help="force-en parser.add_argument("--opt-sub-quad-attention", action='store_true', help="enable memory efficient sub-quadratic cross-attention layer optimization") parser.add_argument("--sub-quad-q-chunk-size", type=int, help="query chunk size for the sub-quadratic cross-attention layer optimization to use", default=1024) parser.add_argument("--sub-quad-kv-chunk-size", type=int, help="kv chunk size for the sub-quadratic cross-attention layer optimization to use", default=None) -parser.add_argument("--sub-quad-chunk-threshold", type=int, help="the size threshold in bytes for the sub-quadratic cross-attention layer optimization to use chunking", default=None) +parser.add_argument("--sub-quad-chunk-threshold", type=int, help="the percentage of VRAM threshold for the sub-quadratic cross-attention layer optimization to use chunking", default=None) parser.add_argument("--opt-split-attention-invokeai", action='store_true', help="force-enables InvokeAI's cross-attention layer optimization. By default, it's on when cuda is unavailable.") parser.add_argument("--opt-split-attention-v1", action='store_true', help="enable older version of split attention optimization that does not consume all the VRAM it can find") parser.add_argument("--disable-opt-split-attention", action='store_true', help="force-disables cross-attention layer optimization") -- cgit v1.2.3 From c18add68ef7d2de3617cbbaff864b0c74cfdf6c0 Mon Sep 17 00:00:00 2001 From: brkirch Date: Fri, 6 Jan 2023 16:42:47 -0500 Subject: Added license --- html/licenses.html | 29 ++++++++++++++++++++++++++++- modules/sd_hijack_optimizations.py | 1 + modules/sub_quadratic_attention.py | 2 +- 3 files changed, 30 insertions(+), 2 deletions(-) (limited to 'modules/sd_hijack_optimizations.py') diff --git a/html/licenses.html b/html/licenses.html index 9eeaa072..570630eb 100644 --- a/html/licenses.html +++ b/html/licenses.html @@ -184,7 +184,7 @@ SOFTWARE.

SwinIR

-Code added by contirubtors, most likely copied from this repository. +Code added by contributors, most likely copied from this repository. 
                                  Apache License
@@ -390,3 +390,30 @@ SOFTWARE.
    limitations under the License.
+

Memory Efficient Attention

+The sub-quadratic cross attention optimization uses modified code from the Memory Efficient Attention package that Alex Birch optimized for 3D tensors. This license is updated to reflect that. +
+MIT License
+
+Copyright (c) 2023 Alex Birch
+Copyright (c) 2023 Amin Rezaei
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+ diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index b416e9ac..cdc63ed7 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -216,6 +216,7 @@ def split_cross_attention_forward_invokeAI(self, x, context=None, mask=None): # Based on Birch-san's modified implementation of sub-quadratic attention from https://github.com/Birch-san/diffusers/pull/1 +# The sub_quad_attention_forward function is under the MIT License listed under Memory Efficient Attention in the Licenses section of the web UI interface def sub_quad_attention_forward(self, x, context=None, mask=None): assert mask is None, "attention-mask not currently implemented for SubQuadraticCrossAttnProcessor." diff --git a/modules/sub_quadratic_attention.py b/modules/sub_quadratic_attention.py index 95924d24..fea7aaac 100644 --- a/modules/sub_quadratic_attention.py +++ b/modules/sub_quadratic_attention.py @@ -1,7 +1,7 @@ # original source: # https://github.com/AminRezaei0x443/memory-efficient-attention/blob/1bc0d9e6ac5f82ea43a375135c4e1d3896ee1694/memory_efficient_attention/attention_torch.py # license: -# unspecified +# MIT License (see Memory Efficient Attention under the Licenses section in the web UI interface for the full license) # credit: # Amin Rezaei (original author) # Alex Birch (optimized algorithm for 3D tensors, at the expense of removing bias, masking and callbacks) -- cgit v1.2.3 From 40ff6db5325fc34ad4fa35e80cb1e7768d9f7e75 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sat, 21 Jan 2023 08:36:07 +0300 Subject: extra networks UI rework of hypernets: rather than via settings, hypernets are added directly to prompt as --- html/card-no-preview.png | Bin 0 -> 84440 bytes html/extra-networks-card.html | 11 ++ html/extra-networks-no-cards.html | 8 ++ javascript/extraNetworks.js | 60 ++++++++ javascript/hints.js | 2 + javascript/ui.js | 9 +- modules/api/api.py | 7 +- modules/extra_networks.py | 147 +++++++++++++++++++ modules/extra_networks_hypernet.py | 21 +++ modules/generation_parameters_copypaste.py | 12 +- modules/hypernetworks/hypernetwork.py | 107 +++++++++----- modules/hypernetworks/ui.py | 5 +- modules/processing.py | 24 ++-- modules/sd_hijack_optimizations.py | 10 +- modules/shared.py | 21 ++- modules/textual_inversion/textual_inversion.py | 2 + modules/ui.py | 50 ++++--- modules/ui_components.py | 10 ++ modules/ui_extra_networks.py | 149 +++++++++++++++++++ modules/ui_extra_networks_hypernets.py | 34 +++++ modules/ui_extra_networks_textual_inversion.py | 32 +++++ script.js | 13 +- scripts/xy_grid.py | 29 ---- style.css | 190 +++++++++++++------------ webui.py | 26 +++- 25 files changed, 765 insertions(+), 214 deletions(-) create mode 100644 html/card-no-preview.png create mode 100644 html/extra-networks-card.html create mode 100644 html/extra-networks-no-cards.html create mode 100644 javascript/extraNetworks.js create mode 100644 modules/extra_networks.py create mode 100644 modules/extra_networks_hypernet.py create mode 100644 modules/ui_extra_networks.py create mode 100644 modules/ui_extra_networks_hypernets.py create mode 100644 modules/ui_extra_networks_textual_inversion.py (limited to 'modules/sd_hijack_optimizations.py') diff --git a/html/card-no-preview.png b/html/card-no-preview.png new file mode 100644 index 00000000..e2beb269 Binary files /dev/null and b/html/card-no-preview.png differ diff --git a/html/extra-networks-card.html b/html/extra-networks-card.html new file mode 100644 index 00000000..7314b063 --- /dev/null +++ b/html/extra-networks-card.html @@ -0,0 +1,11 @@ +
+
+
+ +
+ {name} +
+
+ diff --git a/html/extra-networks-no-cards.html b/html/extra-networks-no-cards.html new file mode 100644 index 00000000..389358d6 --- /dev/null +++ b/html/extra-networks-no-cards.html @@ -0,0 +1,8 @@ +
+

Nothing here. Add some content to the following directories:

+ +
    +{dirs} +
+
+ diff --git a/javascript/extraNetworks.js b/javascript/extraNetworks.js new file mode 100644 index 00000000..71e522d1 --- /dev/null +++ b/javascript/extraNetworks.js @@ -0,0 +1,60 @@ + +function setupExtraNetworksForTab(tabname){ + gradioApp().querySelector('#'+tabname+'_extra_tabs').classList.add('extra-networks') + + gradioApp().querySelector('#'+tabname+'_extra_tabs > div').appendChild(gradioApp().getElementById(tabname+'_extra_refresh')) + gradioApp().querySelector('#'+tabname+'_extra_tabs > div').appendChild(gradioApp().getElementById(tabname+'_extra_close')) +} + +var activePromptTextarea = null; +var activePositivePromptTextarea = null; + +function setupExtraNetworks(){ + setupExtraNetworksForTab('txt2img') + setupExtraNetworksForTab('img2img') + + function registerPrompt(id, isNegative){ + var textarea = gradioApp().querySelector("#" + id + " > label > textarea"); + + if (activePromptTextarea == null){ + activePromptTextarea = textarea + } + if (activePositivePromptTextarea == null && ! isNegative){ + activePositivePromptTextarea = textarea + } + + textarea.addEventListener("focus", function(){ + activePromptTextarea = textarea; + if(! isNegative) activePositivePromptTextarea = textarea; + }); + } + + registerPrompt('txt2img_prompt') + registerPrompt('txt2img_neg_prompt', true) + registerPrompt('img2img_prompt') + registerPrompt('img2img_neg_prompt', true) +} + +onUiLoaded(setupExtraNetworks) + +function cardClicked(textToAdd, allowNegativePrompt){ + textarea = allowNegativePrompt ? activePromptTextarea : activePositivePromptTextarea + + textarea.value = textarea.value + " " + textToAdd + updateInput(textarea) + + return false +} + +function saveCardPreview(event, tabname, filename){ + textarea = gradioApp().querySelector("#" + tabname + '_preview_filename > label > textarea') + button = gradioApp().getElementById(tabname + '_save_preview') + + textarea.value = filename + updateInput(textarea) + + button.click() + + event.stopPropagation() + event.preventDefault() +} diff --git a/javascript/hints.js b/javascript/hints.js index e746e20d..f4079f96 100644 --- a/javascript/hints.js +++ b/javascript/hints.js @@ -21,6 +21,8 @@ titles = { "\U0001F5D1": "Clear prompt", "\u{1f4cb}": "Apply selected styles to current prompt", "\u{1f4d2}": "Paste available values into the field", + "\u{1f3b4}": "Show extra networks", + "Inpaint a part of image": "Draw a mask over an image, and the script will regenerate the masked area with content according to prompt", "SD upscale": "Upscale image normally, split result into tiles, improve each tile using img2img, merge whole image back", diff --git a/javascript/ui.js b/javascript/ui.js index 3ba90ca8..a7e75439 100644 --- a/javascript/ui.js +++ b/javascript/ui.js @@ -196,8 +196,6 @@ function confirm_clear_prompt(prompt, negative_prompt) { return [prompt, negative_prompt] } - - opts = {} onUiUpdate(function(){ if(Object.keys(opts).length != 0) return; @@ -239,11 +237,14 @@ onUiUpdate(function(){ return } + prompt.parentElement.insertBefore(counter, prompt) counter.classList.add("token-counter") prompt.parentElement.style.position = "relative" - textarea.addEventListener("input", () => update_token_counter(id_button)); + textarea.addEventListener("input", function(){ + update_token_counter(id_button); + }); } registerTextarea('txt2img_prompt', 'txt2img_token_counter', 'txt2img_token_button') @@ -261,10 +262,8 @@ onUiUpdate(function(){ }) } } - }) - onOptionsChanged(function(){ elem = gradioApp().getElementById('sd_checkpoint_hash') sd_checkpoint_hash = opts.sd_checkpoint_hash || "" diff --git a/modules/api/api.py b/modules/api/api.py index 9814bbc2..2c371e6e 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -480,7 +480,7 @@ class Api: def train_hypernetwork(self, args: dict): try: shared.state.begin() - initial_hypernetwork = shared.loaded_hypernetwork + shared.loaded_hypernetworks = [] apply_optimizations = shared.opts.training_xattention_optimizations error = None filename = '' @@ -491,16 +491,15 @@ class Api: except Exception as e: error = e finally: - shared.loaded_hypernetwork = initial_hypernetwork shared.sd_model.cond_stage_model.to(devices.device) shared.sd_model.first_stage_model.to(devices.device) if not apply_optimizations: sd_hijack.apply_optimizations() shared.state.end() - return TrainResponse(info = "train embedding complete: filename: {filename} error: {error}".format(filename = filename, error = error)) + return TrainResponse(info="train embedding complete: filename: {filename} error: {error}".format(filename=filename, error=error)) except AssertionError as msg: shared.state.end() - return TrainResponse(info = "train embedding error: {error}".format(error = error)) + return TrainResponse(info="train embedding error: {error}".format(error=error)) def get_memory(self): try: diff --git a/modules/extra_networks.py b/modules/extra_networks.py new file mode 100644 index 00000000..1978673d --- /dev/null +++ b/modules/extra_networks.py @@ -0,0 +1,147 @@ +import re +from collections import defaultdict + +from modules import errors + +extra_network_registry = {} + + +def initialize(): + extra_network_registry.clear() + + +def register_extra_network(extra_network): + extra_network_registry[extra_network.name] = extra_network + + +class ExtraNetworkParams: + def __init__(self, items=None): + self.items = items or [] + + +class ExtraNetwork: + def __init__(self, name): + self.name = name + + def activate(self, p, params_list): + """ + Called by processing on every run. Whatever the extra network is meant to do should be activated here. + Passes arguments related to this extra network in params_list. + User passes arguments by specifying this in his prompt: + + + + Where name matches the name of this ExtraNetwork object, and arg1:arg2:arg3 are any natural number of text arguments + separated by colon. + + Even if the user does not mention this ExtraNetwork in his prompt, the call will stil be made, with empty params_list - + in this case, all effects of this extra networks should be disabled. + + Can be called multiple times before deactivate() - each new call should override the previous call completely. + + For example, if this ExtraNetwork's name is 'hypernet' and user's prompt is: + + > "1girl, " + + params_list will be: + + [ + ExtraNetworkParams(items=["agm", "1.1"]), + ExtraNetworkParams(items=["ray"]) + ] + + """ + raise NotImplementedError + + def deactivate(self, p): + """ + Called at the end of processing for housekeeping. No need to do anything here. + """ + + raise NotImplementedError + + +def activate(p, extra_network_data): + """call activate for extra networks in extra_network_data in specified order, then call + activate for all remaining registered networks with an empty argument list""" + + for extra_network_name, extra_network_args in extra_network_data.items(): + extra_network = extra_network_registry.get(extra_network_name, None) + if extra_network is None: + print(f"Skipping unknown extra network: {extra_network_name}") + continue + + try: + extra_network.activate(p, extra_network_args) + except Exception as e: + errors.display(e, f"activating extra network {extra_network_name} with arguments {extra_network_args}") + + for extra_network_name, extra_network in extra_network_registry.items(): + args = extra_network_data.get(extra_network_name, None) + if args is not None: + continue + + try: + extra_network.activate(p, []) + except Exception as e: + errors.display(e, f"activating extra network {extra_network_name}") + + +def deactivate(p, extra_network_data): + """call deactivate for extra networks in extra_network_data in specified order, then call + deactivate for all remaining registered networks""" + + for extra_network_name, extra_network_args in extra_network_data.items(): + extra_network = extra_network_registry.get(extra_network_name, None) + if extra_network is None: + continue + + try: + extra_network.deactivate(p) + except Exception as e: + errors.display(e, f"deactivating extra network {extra_network_name}") + + for extra_network_name, extra_network in extra_network_registry.items(): + args = extra_network_data.get(extra_network_name, None) + if args is not None: + continue + + try: + extra_network.deactivate(p) + except Exception as e: + errors.display(e, f"deactivating unmentioned extra network {extra_network_name}") + + +re_extra_net = re.compile(r"<(\w+):([^>]+)>") + + +def parse_prompt(prompt): + res = defaultdict(list) + + def found(m): + name = m.group(1) + args = m.group(2) + + res[name].append(ExtraNetworkParams(items=args.split(":"))) + + return "" + + prompt = re.sub(re_extra_net, found, prompt) + + return prompt, res + + +def parse_prompts(prompts): + res = [] + extra_data = None + + for prompt in prompts: + updated_prompt, parsed_extra_data = parse_prompt(prompt) + + if extra_data is None: + extra_data = parsed_extra_data + + res.append(updated_prompt) + + return res, extra_data + diff --git a/modules/extra_networks_hypernet.py b/modules/extra_networks_hypernet.py new file mode 100644 index 00000000..6a0c4ba8 --- /dev/null +++ b/modules/extra_networks_hypernet.py @@ -0,0 +1,21 @@ +from modules import extra_networks +from modules.hypernetworks import hypernetwork + + +class ExtraNetworkHypernet(extra_networks.ExtraNetwork): + def __init__(self): + super().__init__('hypernet') + + def activate(self, p, params_list): + names = [] + multipliers = [] + for params in params_list: + assert len(params.items) > 0 + + names.append(params.items[0]) + multipliers.append(float(params.items[1]) if len(params.items) > 1 else 1.0) + + hypernetwork.load_hypernetworks(names, multipliers) + + def deactivate(p, self): + pass diff --git a/modules/generation_parameters_copypaste.py b/modules/generation_parameters_copypaste.py index a381ff59..46e12dc6 100644 --- a/modules/generation_parameters_copypaste.py +++ b/modules/generation_parameters_copypaste.py @@ -79,8 +79,6 @@ def integrate_settings_paste_fields(component_dict): from modules import ui settings_map = { - 'sd_hypernetwork': 'Hypernet', - 'sd_hypernetwork_strength': 'Hypernet strength', 'CLIP_stop_at_last_layers': 'Clip skip', 'inpainting_mask_weight': 'Conditional mask weight', 'sd_model_checkpoint': 'Model hash', @@ -275,13 +273,9 @@ Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 965400086, Size: 512x512, Model if "Clip skip" not in res: res["Clip skip"] = "1" - if "Hypernet strength" not in res: - res["Hypernet strength"] = "1" - - if "Hypernet" in res: - hypernet_name = res["Hypernet"] - hypernet_hash = res.get("Hypernet hash", None) - res["Hypernet"] = find_hypernetwork_key(hypernet_name, hypernet_hash) + hypernet = res.get("Hypernet", None) + if hypernet is not None: + res["Prompt"] += f"""""" if "Hires resize-1" not in res: res["Hires resize-1"] = 0 diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 74e78582..80a47c79 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -25,7 +25,6 @@ from statistics import stdev, mean optimizer_dict = {optim_name : cls_obj for optim_name, cls_obj in inspect.getmembers(torch.optim, inspect.isclass) if optim_name != "Optimizer"} class HypernetworkModule(torch.nn.Module): - multiplier = 1.0 activation_dict = { "linear": torch.nn.Identity, "relu": torch.nn.ReLU, @@ -41,6 +40,8 @@ class HypernetworkModule(torch.nn.Module): add_layer_norm=False, activate_output=False, dropout_structure=None): super().__init__() + self.multiplier = 1.0 + assert layer_structure is not None, "layer_structure must not be None" assert layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" assert layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" @@ -115,7 +116,7 @@ class HypernetworkModule(torch.nn.Module): state_dict[to] = x def forward(self, x): - return x + self.linear(x) * (HypernetworkModule.multiplier if not self.training else 1) + return x + self.linear(x) * (self.multiplier if not self.training else 1) def trainables(self): layer_structure = [] @@ -125,9 +126,6 @@ class HypernetworkModule(torch.nn.Module): return layer_structure -def apply_strength(value=None): - HypernetworkModule.multiplier = value if value is not None else shared.opts.sd_hypernetwork_strength - #param layer_structure : sequence used for length, use_dropout : controlling boolean, last_layer_dropout : for compatibility check. def parse_dropout_structure(layer_structure, use_dropout, last_layer_dropout): if layer_structure is None: @@ -192,6 +190,20 @@ class Hypernetwork: for param in layer.parameters(): param.requires_grad = mode + def to(self, device): + for k, layers in self.layers.items(): + for layer in layers: + layer.to(device) + + return self + + def set_multiplier(self, multiplier): + for k, layers in self.layers.items(): + for layer in layers: + layer.multiplier = multiplier + + return self + def eval(self): for k, layers in self.layers.items(): for layer in layers: @@ -269,11 +281,13 @@ class Hypernetwork: self.optimizer_state_dict = None if self.optimizer_state_dict: self.optimizer_name = optimizer_saved_dict.get('optimizer_name', 'AdamW') - print("Loaded existing optimizer from checkpoint") - print(f"Optimizer name is {self.optimizer_name}") + if shared.opts.print_hypernet_extra: + print("Loaded existing optimizer from checkpoint") + print(f"Optimizer name is {self.optimizer_name}") else: self.optimizer_name = "AdamW" - print("No saved optimizer exists in checkpoint") + if shared.opts.print_hypernet_extra: + print("No saved optimizer exists in checkpoint") for size, sd in state_dict.items(): if type(size) == int: @@ -306,23 +320,43 @@ def list_hypernetworks(path): return res -def load_hypernetwork(filename): - path = shared.hypernetworks.get(filename, None) - # Prevent any file named "None.pt" from being loaded. - if path is not None and filename != "None": - print(f"Loading hypernetwork {filename}") - try: - shared.loaded_hypernetwork = Hypernetwork() - shared.loaded_hypernetwork.load(path) +def load_hypernetwork(name): + path = shared.hypernetworks.get(name, None) - except Exception: - print(f"Error loading hypernetwork {path}", file=sys.stderr) - print(traceback.format_exc(), file=sys.stderr) - else: - if shared.loaded_hypernetwork is not None: - print("Unloading hypernetwork") + if path is None: + return None + + hypernetwork = Hypernetwork() + + try: + hypernetwork.load(path) + except Exception: + print(f"Error loading hypernetwork {path}", file=sys.stderr) + print(traceback.format_exc(), file=sys.stderr) + return None + + return hypernetwork + + +def load_hypernetworks(names, multipliers=None): + already_loaded = {} + + for hypernetwork in shared.loaded_hypernetworks: + if hypernetwork.name in names: + already_loaded[hypernetwork.name] = hypernetwork - shared.loaded_hypernetwork = None + shared.loaded_hypernetworks.clear() + + for i, name in enumerate(names): + hypernetwork = already_loaded.get(name, None) + if hypernetwork is None: + hypernetwork = load_hypernetwork(name) + + if hypernetwork is None: + continue + + hypernetwork.set_multiplier(multipliers[i] if multipliers else 1.0) + shared.loaded_hypernetworks.append(hypernetwork) def find_closest_hypernetwork_name(search: str): @@ -336,18 +370,27 @@ def find_closest_hypernetwork_name(search: str): return applicable[0] -def apply_hypernetwork(hypernetwork, context, layer=None): - hypernetwork_layers = (hypernetwork.layers if hypernetwork is not None else {}).get(context.shape[2], None) +def apply_single_hypernetwork(hypernetwork, context_k, context_v, layer=None): + hypernetwork_layers = (hypernetwork.layers if hypernetwork is not None else {}).get(context_k.shape[2], None) if hypernetwork_layers is None: - return context, context + return context_k, context_v if layer is not None: layer.hyper_k = hypernetwork_layers[0] layer.hyper_v = hypernetwork_layers[1] - context_k = hypernetwork_layers[0](context) - context_v = hypernetwork_layers[1](context) + context_k = hypernetwork_layers[0](context_k) + context_v = hypernetwork_layers[1](context_v) + return context_k, context_v + + +def apply_hypernetworks(hypernetworks, context, layer=None): + context_k = context + context_v = context + for hypernetwork in hypernetworks: + context_k, context_v = apply_single_hypernetwork(hypernetwork, context_k, context_v, layer) + return context_k, context_v @@ -357,7 +400,7 @@ def attention_CrossAttention_forward(self, x, context=None, mask=None): q = self.to_q(x) context = default(context, x) - context_k, context_v = apply_hypernetwork(shared.loaded_hypernetwork, context, self) + context_k, context_v = apply_hypernetworks(shared.loaded_hypernetworks, context, self) k = self.to_k(context_k) v = self.to_v(context_v) @@ -464,8 +507,9 @@ def train_hypernetwork(id_task, hypernetwork_name, learn_rate, batch_size, gradi template_file = template_file.path path = shared.hypernetworks.get(hypernetwork_name, None) - shared.loaded_hypernetwork = Hypernetwork() - shared.loaded_hypernetwork.load(path) + hypernetwork = Hypernetwork() + hypernetwork.load(path) + shared.loaded_hypernetworks = [hypernetwork] shared.state.job = "train-hypernetwork" shared.state.textinfo = "Initializing hypernetwork training..." @@ -489,7 +533,6 @@ def train_hypernetwork(id_task, hypernetwork_name, learn_rate, batch_size, gradi else: images_dir = None - hypernetwork = shared.loaded_hypernetwork checkpoint = sd_models.select_checkpoint() initial_step = hypernetwork.step or 0 diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 81e3f519..76599f5a 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -9,6 +9,7 @@ from modules import devices, sd_hijack, shared not_available = ["hardswish", "multiheadattention"] keys = list(x for x in modules.hypernetworks.hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) + def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False, dropout_structure=None): filename = modules.hypernetworks.hypernetwork.create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure, activation_func, weight_init, add_layer_norm, use_dropout, dropout_structure) @@ -16,8 +17,7 @@ def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, def train_hypernetwork(*args): - - initial_hypernetwork = shared.loaded_hypernetwork + shared.loaded_hypernetworks = [] assert not shared.cmd_opts.lowvram, 'Training models with lowvram is not possible' @@ -34,7 +34,6 @@ Hypernetwork saved to {html.escape(filename)} except Exception: raise finally: - shared.loaded_hypernetwork = initial_hypernetwork shared.sd_model.cond_stage_model.to(devices.device) shared.sd_model.first_stage_model.to(devices.device) sd_hijack.apply_optimizations() diff --git a/modules/processing.py b/modules/processing.py index a3e9f709..b5deeacf 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -13,7 +13,7 @@ from skimage import exposure from typing import Any, Dict, List, Optional import modules.sd_hijack -from modules import devices, prompt_parser, masking, sd_samplers, lowvram, generation_parameters_copypaste, script_callbacks +from modules import devices, prompt_parser, masking, sd_samplers, lowvram, generation_parameters_copypaste, script_callbacks, extra_networks from modules.sd_hijack import model_hijack from modules.shared import opts, cmd_opts, state import modules.shared as shared @@ -438,9 +438,6 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments=None, iter "Size": f"{p.width}x{p.height}", "Model hash": getattr(p, 'sd_model_hash', None if not opts.add_model_hash_to_info or not shared.sd_model.sd_model_hash else shared.sd_model.sd_model_hash), "Model": (None if not opts.add_model_name_to_info or not shared.sd_model.sd_checkpoint_info.model_name else shared.sd_model.sd_checkpoint_info.model_name.replace(',', '').replace(':', '')), - "Hypernet": (None if shared.loaded_hypernetwork is None else shared.loaded_hypernetwork.name), - "Hypernet hash": (None if shared.loaded_hypernetwork is None else shared.loaded_hypernetwork.shorthash()), - "Hypernet strength": (None if shared.loaded_hypernetwork is None or shared.opts.sd_hypernetwork_strength >= 1 else shared.opts.sd_hypernetwork_strength), "Batch size": (None if p.batch_size < 2 else p.batch_size), "Batch pos": (None if p.batch_size < 2 else position_in_batch), "Variation seed": (None if p.subseed_strength == 0 else all_subseeds[index]), @@ -468,14 +465,12 @@ def process_images(p: StableDiffusionProcessing) -> Processed: try: for k, v in p.override_settings.items(): setattr(opts, k, v) - if k == 'sd_hypernetwork': - shared.reload_hypernetworks() # make onchange call for changing hypernet if k == 'sd_model_checkpoint': - sd_models.reload_model_weights() # make onchange call for changing SD model + sd_models.reload_model_weights() if k == 'sd_vae': - sd_vae.reload_vae_weights() # make onchange call for changing VAE + sd_vae.reload_vae_weights() res = process_images_inner(p) @@ -484,9 +479,11 @@ def process_images(p: StableDiffusionProcessing) -> Processed: if p.override_settings_restore_afterwards: for k, v in stored_opts.items(): setattr(opts, k, v) - if k == 'sd_hypernetwork': shared.reload_hypernetworks() - if k == 'sd_model_checkpoint': sd_models.reload_model_weights() - if k == 'sd_vae': sd_vae.reload_vae_weights() + if k == 'sd_model_checkpoint': + sd_models.reload_model_weights() + + if k == 'sd_vae': + sd_vae.reload_vae_weights() return res @@ -564,10 +561,14 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed: cache[0] = (required_prompts, steps) return cache[1] + p.all_prompts, extra_network_data = extra_networks.parse_prompts(p.all_prompts) + with torch.no_grad(), p.sd_model.ema_scope(): with devices.autocast(): p.init(p.all_prompts, p.all_seeds, p.all_subseeds) + extra_networks.activate(p, extra_network_data) + with open(os.path.join(shared.script_path, "params.txt"), "w", encoding="utf8") as file: processed = Processed(p, [], p.seed, "") file.write(processed.infotext(p, 0)) @@ -681,6 +682,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed: if opts.grid_save: images.save_image(grid, p.outpath_grids, "grid", p.all_seeds[0], p.all_prompts[0], opts.grid_format, info=infotext(), short_filename=not opts.grid_extended_filename, p=p, grid=True) + extra_networks.deactivate(p, extra_network_data) devices.torch_gc() res = Processed(p, output_images, p.all_seeds[0], infotext(), comments="".join(["\n\n" + x for x in comments]), subseed=p.all_subseeds[0], index_of_first_image=index_of_first_image, infotexts=infotexts) diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index cdc63ed7..4fa54329 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -44,7 +44,7 @@ def split_cross_attention_forward_v1(self, x, context=None, mask=None): q_in = self.to_q(x) context = default(context, x) - context_k, context_v = hypernetwork.apply_hypernetwork(shared.loaded_hypernetwork, context) + context_k, context_v = hypernetwork.apply_hypernetworks(shared.loaded_hypernetworks, context) k_in = self.to_k(context_k) v_in = self.to_v(context_v) del context, context_k, context_v, x @@ -78,7 +78,7 @@ def split_cross_attention_forward(self, x, context=None, mask=None): q_in = self.to_q(x) context = default(context, x) - context_k, context_v = hypernetwork.apply_hypernetwork(shared.loaded_hypernetwork, context) + context_k, context_v = hypernetwork.apply_hypernetworks(shared.loaded_hypernetworks, context) k_in = self.to_k(context_k) v_in = self.to_v(context_v) @@ -203,7 +203,7 @@ def split_cross_attention_forward_invokeAI(self, x, context=None, mask=None): q = self.to_q(x) context = default(context, x) - context_k, context_v = hypernetwork.apply_hypernetwork(shared.loaded_hypernetwork, context) + context_k, context_v = hypernetwork.apply_hypernetworks(shared.loaded_hypernetworks, context) k = self.to_k(context_k) * self.scale v = self.to_v(context_v) del context, context_k, context_v, x @@ -225,7 +225,7 @@ def sub_quad_attention_forward(self, x, context=None, mask=None): q = self.to_q(x) context = default(context, x) - context_k, context_v = hypernetwork.apply_hypernetwork(shared.loaded_hypernetwork, context) + context_k, context_v = hypernetwork.apply_hypernetworks(shared.loaded_hypernetworks, context) k = self.to_k(context_k) v = self.to_v(context_v) del context, context_k, context_v, x @@ -284,7 +284,7 @@ def xformers_attention_forward(self, x, context=None, mask=None): q_in = self.to_q(x) context = default(context, x) - context_k, context_v = hypernetwork.apply_hypernetwork(shared.loaded_hypernetwork, context) + context_k, context_v = hypernetwork.apply_hypernetworks(shared.loaded_hypernetworks, context) k_in = self.to_k(context_k) v_in = self.to_v(context_v) diff --git a/modules/shared.py b/modules/shared.py index 2f366454..c0e11f18 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -23,6 +23,7 @@ demo = None sd_default_config = os.path.join(script_path, "configs/v1-inference.yaml") sd_model_file = os.path.join(script_path, 'model.ckpt') default_sd_model_file = sd_model_file + parser = argparse.ArgumentParser() parser.add_argument("--config", type=str, default=sd_default_config, help="path to config which constructs model",) parser.add_argument("--ckpt", type=str, default=sd_model_file, help="path to checkpoint of stable diffusion model; if specified, this checkpoint will be added to the list of checkpoints and loaded",) @@ -145,7 +146,7 @@ config_filename = cmd_opts.ui_settings_file os.makedirs(cmd_opts.hypernetwork_dir, exist_ok=True) hypernetworks = {} -loaded_hypernetwork = None +loaded_hypernetworks = [] def reload_hypernetworks(): @@ -153,8 +154,6 @@ def reload_hypernetworks(): global hypernetworks hypernetworks = hypernetwork.list_hypernetworks(cmd_opts.hypernetwork_dir) - hypernetwork.load_hypernetwork(opts.sd_hypernetwork) - class State: @@ -399,8 +398,6 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "sd_vae_checkpoint_cache": OptionInfo(0, "VAE Checkpoints to cache in RAM", gr.Slider, {"minimum": 0, "maximum": 10, "step": 1}), "sd_vae": OptionInfo("Automatic", "SD VAE", gr.Dropdown, lambda: {"choices": ["Automatic", "None"] + list(sd_vae.vae_dict)}, refresh=sd_vae.refresh_vae_list), "sd_vae_as_default": OptionInfo(True, "Ignore selected VAE for stable diffusion checkpoints that have their own .vae.pt next to them"), - "sd_hypernetwork": OptionInfo("None", "Hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}, refresh=reload_hypernetworks), - "sd_hypernetwork_strength": OptionInfo(1.0, "Hypernetwork strength", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.001}), "inpainting_mask_weight": OptionInfo(1.0, "Inpainting conditioning mask strength", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}), "initial_noise_multiplier": OptionInfo(1.0, "Noise multiplier for img2img", gr.Slider, {"minimum": 0.5, "maximum": 1.5, "step": 0.01 }), "img2img_color_correction": OptionInfo(False, "Apply color correction to img2img results to match original colors."), @@ -661,3 +658,17 @@ mem_mon.start() def listfiles(dirname): filenames = [os.path.join(dirname, x) for x in sorted(os.listdir(dirname)) if not x.startswith(".")] return [file for file in filenames if os.path.isfile(file)] + + +def html_path(filename): + return os.path.join(script_path, "html", filename) + + +def html(filename): + path = html_path(filename) + + if os.path.exists(path): + with open(path, encoding="utf8") as file: + return file.read() + + return "" diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index 5a7be422..4e90f690 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -50,6 +50,7 @@ class Embedding: self.sd_checkpoint = None self.sd_checkpoint_name = None self.optimizer_state_dict = None + self.filename = None def save(self, filename): embedding_data = { @@ -182,6 +183,7 @@ class EmbeddingDatabase: embedding.sd_checkpoint_name = data.get('sd_checkpoint_name', None) embedding.vectors = vec.shape[0] embedding.shape = vec.shape[-1] + embedding.filename = path if self.expected_shape == -1 or self.expected_shape == embedding.shape: self.register_embedding(embedding, shared.sd_model) diff --git a/modules/ui.py b/modules/ui.py index 06c11848..d23b2b8e 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -20,7 +20,7 @@ import numpy as np from PIL import Image, PngImagePlugin from modules.call_queue import wrap_gradio_gpu_call, wrap_queued_call, wrap_gradio_call -from modules import sd_hijack, sd_models, localization, script_callbacks, ui_extensions, deepbooru, sd_vae +from modules import sd_hijack, sd_models, localization, script_callbacks, ui_extensions, deepbooru, sd_vae, extra_networks from modules.ui_components import FormRow, FormGroup, ToolButton, FormHTML from modules.paths import script_path @@ -90,6 +90,7 @@ refresh_symbol = '\U0001f504' # 🔄 save_style_symbol = '\U0001f4be' # 💾 apply_style_symbol = '\U0001f4cb' # 📋 clear_prompt_symbol = '\U0001F5D1' # 🗑️ +extra_networks_symbol = '\U0001F3B4' # 🎴 def plaintext_to_html(text): @@ -324,6 +325,8 @@ def connect_reuse_seed(seed: gr.Number, reuse_seed: gr.Button, generation_info: def update_token_counter(text, steps): try: + text, _ = extra_networks.parse_prompt(text) + _, prompt_flat_list, _ = prompt_parser.get_multicond_prompt_list([text]) prompt_schedules = prompt_parser.get_learned_conditioning_prompt_schedules(prompt_flat_list, steps) @@ -354,10 +357,10 @@ def create_toprow(is_img2img): negative_prompt = gr.Textbox(label="Negative prompt", elem_id=f"{id_part}_neg_prompt", show_label=False, lines=2, placeholder="Negative prompt (press Ctrl+Enter or Alt+Enter to generate)") with gr.Column(scale=1, elem_id="roll_col"): - paste = gr.Button(value=paste_symbol, elem_id="paste") - save_style = gr.Button(value=save_style_symbol, elem_id="style_create") - prompt_style_apply = gr.Button(value=apply_style_symbol, elem_id="style_apply") - clear_prompt_button = gr.Button(value=clear_prompt_symbol, elem_id=f"{id_part}_clear_prompt") + paste = ToolButton(value=paste_symbol, elem_id="paste") + clear_prompt_button = ToolButton(value=clear_prompt_symbol, elem_id=f"{id_part}_clear_prompt") + extra_networks_button = ToolButton(value=extra_networks_symbol, elem_id=f"{id_part}_extra_networks") + token_counter = gr.HTML(value="", elem_id=f"{id_part}_token_counter") token_button = gr.Button(visible=False, elem_id=f"{id_part}_token_button") negative_token_counter = gr.HTML(value="", elem_id=f"{id_part}_negative_token_counter") @@ -395,11 +398,14 @@ def create_toprow(is_img2img): outputs=[], ) - with gr.Row(): + with gr.Row(elem_id=f"{id_part}_styles_row"): prompt_styles = gr.Dropdown(label="Styles", elem_id=f"{id_part}_styles", choices=[k for k, v in shared.prompt_styles.styles.items()], value=[], multiselect=True) create_refresh_button(prompt_styles, shared.prompt_styles.reload, lambda: {"choices": [k for k, v in shared.prompt_styles.styles.items()]}, f"refresh_{id_part}_styles") - return prompt, prompt_styles, negative_prompt, submit, button_interrogate, button_deepbooru, prompt_style_apply, save_style, paste, token_counter, token_button, negative_token_counter, negative_token_button + prompt_style_apply = ToolButton(value=apply_style_symbol, elem_id="style_apply") + save_style = ToolButton(value=save_style_symbol, elem_id="style_create") + + return prompt, prompt_styles, negative_prompt, submit, button_interrogate, button_deepbooru, prompt_style_apply, save_style, paste, extra_networks_button, token_counter, token_button, negative_token_counter, negative_token_button def setup_progressbar(*args, **kwargs): @@ -616,11 +622,15 @@ def create_ui(): modules.scripts.scripts_txt2img.initialize_scripts(is_img2img=False) with gr.Blocks(analytics_enabled=False) as txt2img_interface: - txt2img_prompt, txt2img_prompt_styles, txt2img_negative_prompt, submit, _, _, txt2img_prompt_style_apply, txt2img_save_style, txt2img_paste, token_counter, token_button, negative_token_counter, negative_token_button = create_toprow(is_img2img=False) + txt2img_prompt, txt2img_prompt_styles, txt2img_negative_prompt, submit, _, _, txt2img_prompt_style_apply, txt2img_save_style, txt2img_paste, extra_networks_button, token_counter, token_button, negative_token_counter, negative_token_button = create_toprow(is_img2img=False) dummy_component = gr.Label(visible=False) txt_prompt_img = gr.File(label="", elem_id="txt2img_prompt_image", file_count="single", type="binary", visible=False) + with FormRow(variant='compact', elem_id="txt2img_extra_networks", visible=False) as extra_networks: + from modules import ui_extra_networks + extra_networks_ui = ui_extra_networks.create_ui(extra_networks, extra_networks_button, 'txt2img') + with gr.Row().style(equal_height=False): with gr.Column(variant='compact', elem_id="txt2img_settings"): for category in ordered_ui_categories(): @@ -794,14 +804,20 @@ def create_ui(): token_button.click(fn=wrap_queued_call(update_token_counter), inputs=[txt2img_prompt, steps], outputs=[token_counter]) negative_token_button.click(fn=wrap_queued_call(update_token_counter), inputs=[txt2img_negative_prompt, steps], outputs=[negative_token_counter]) + ui_extra_networks.setup_ui(extra_networks_ui, txt2img_gallery) + modules.scripts.scripts_current = modules.scripts.scripts_img2img modules.scripts.scripts_img2img.initialize_scripts(is_img2img=True) with gr.Blocks(analytics_enabled=False) as img2img_interface: - img2img_prompt, img2img_prompt_styles, img2img_negative_prompt, submit, img2img_interrogate, img2img_deepbooru, img2img_prompt_style_apply, img2img_save_style, img2img_paste, token_counter, token_button, negative_token_counter, negative_token_button = create_toprow(is_img2img=True) + img2img_prompt, img2img_prompt_styles, img2img_negative_prompt, submit, img2img_interrogate, img2img_deepbooru, img2img_prompt_style_apply, img2img_save_style, img2img_paste, extra_networks_button, token_counter, token_button, negative_token_counter, negative_token_button = create_toprow(is_img2img=True) img2img_prompt_img = gr.File(label="", elem_id="img2img_prompt_image", file_count="single", type="binary", visible=False) + with FormRow(variant='compact', elem_id="img2img_extra_networks", visible=False) as extra_networks: + from modules import ui_extra_networks + extra_networks_ui_img2img = ui_extra_networks.create_ui(extra_networks, extra_networks_button, 'img2img') + with FormRow().style(equal_height=False): with gr.Column(variant='compact', elem_id="img2img_settings"): copy_image_buttons = [] @@ -1064,6 +1080,8 @@ def create_ui(): token_button.click(fn=update_token_counter, inputs=[img2img_prompt, steps], outputs=[token_counter]) negative_token_button.click(fn=wrap_queued_call(update_token_counter), inputs=[txt2img_negative_prompt, steps], outputs=[negative_token_counter]) + ui_extra_networks.setup_ui(extra_networks_ui_img2img, img2img_gallery) + img2img_paste_fields = [ (img2img_prompt, "Prompt"), (img2img_negative_prompt, "Negative prompt"), @@ -1666,10 +1684,8 @@ def create_ui(): download_localization = gr.Button(value='Download localization template', elem_id="download_localization") reload_script_bodies = gr.Button(value='Reload custom script bodies (No ui updates, No restart)', variant='secondary', elem_id="settings_reload_script_bodies") - if os.path.exists("html/licenses.html"): - with open("html/licenses.html", encoding="utf8") as file: - with gr.TabItem("Licenses"): - gr.HTML(file.read(), elem_id="licenses") + with gr.TabItem("Licenses"): + gr.HTML(shared.html("licenses.html"), elem_id="licenses") gr.Button(value="Show all pages", elem_id="settings_show_all_pages") @@ -1756,11 +1772,9 @@ def create_ui(): if os.path.exists(os.path.join(script_path, "notification.mp3")): audio_notification = gr.Audio(interactive=False, value=os.path.join(script_path, "notification.mp3"), elem_id="audio_notification", visible=False) - if os.path.exists("html/footer.html"): - with open("html/footer.html", encoding="utf8") as file: - footer = file.read() - footer = footer.format(versions=versions_html()) - gr.HTML(footer, elem_id="footer") + footer = shared.html("footer.html") + footer = footer.format(versions=versions_html()) + gr.HTML(footer, elem_id="footer") text_settings = gr.Textbox(elem_id="settings_json", value=lambda: opts.dumpjson(), visible=False) settings_submit.click( diff --git a/modules/ui_components.py b/modules/ui_components.py index 97acff06..46324425 100644 --- a/modules/ui_components.py +++ b/modules/ui_components.py @@ -11,6 +11,16 @@ class ToolButton(gr.Button, gr.components.FormComponent): return "button" +class ToolButtonTop(gr.Button, gr.components.FormComponent): + """Small button with single emoji as text, with extra margin at top, fits inside gradio forms""" + + def __init__(self, **kwargs): + super().__init__(variant="tool-top", **kwargs) + + def get_block_name(self): + return "button" + + class FormRow(gr.Row, gr.components.FormComponent): """Same as gr.Row but fits inside gradio forms""" diff --git a/modules/ui_extra_networks.py b/modules/ui_extra_networks.py new file mode 100644 index 00000000..253e90f7 --- /dev/null +++ b/modules/ui_extra_networks.py @@ -0,0 +1,149 @@ +import os.path + +from modules import shared +import gradio as gr +import json + +from modules.generation_parameters_copypaste import image_from_url_text + +extra_pages = [] + + +def register_page(page): + """registers extra networks page for the UI; recommend doing it in on_app_started() callback for extensions""" + + extra_pages.append(page) + + +class ExtraNetworksPage: + def __init__(self, title): + self.title = title + self.card_page = shared.html("extra-networks-card.html") + self.allow_negative_prompt = False + + def refresh(self): + pass + + def create_html(self, tabname): + items_html = '' + + for item in self.list_items(): + items_html += self.create_html_for_item(item, tabname) + + if items_html == '': + dirs = "".join([f"
  • {x}
    • " for x in self.allowed_directories_for_previews()]) + items_html = shared.html("extra-networks-no-cards.html").format(dirs=dirs) + + res = "
    " + items_html + "
    " + + return res + + def list_items(self): + raise NotImplementedError() + + def allowed_directories_for_previews(self): + return [] + + def create_html_for_item(self, item, tabname): + preview = item.get("preview", None) + + args = { + "preview_html": "style='background-image: url(" + json.dumps(preview) + ")'" if preview else '', + "prompt": json.dumps(item["prompt"]), + "tabname": json.dumps(tabname), + "local_preview": json.dumps(item["local_preview"]), + "name": item["name"], + "allow_negative_prompt": "true" if self.allow_negative_prompt else "false", + } + + return self.card_page.format(**args) + + +def intialize(): + extra_pages.clear() + + +class ExtraNetworksUi: + def __init__(self): + self.pages = None + self.stored_extra_pages = None + + self.button_save_preview = None + self.preview_target_filename = None + + self.tabname = None + + +def create_ui(container, button, tabname): + ui = ExtraNetworksUi() + ui.pages = [] + ui.stored_extra_pages = extra_pages.copy() + ui.tabname = tabname + + with gr.Tabs(elem_id=tabname+"_extra_tabs") as tabs: + button_refresh = gr.Button('Refresh', elem_id=tabname+"_extra_refresh") + button_close = gr.Button('Close', elem_id=tabname+"_extra_close") + + for page in ui.stored_extra_pages: + with gr.Tab(page.title): + page_elem = gr.HTML(page.create_html(ui.tabname)) + ui.pages.append(page_elem) + + ui.button_save_preview = gr.Button('Save preview', elem_id=tabname+"_save_preview", visible=False) + ui.preview_target_filename = gr.Textbox('Preview save filename', elem_id=tabname+"_preview_filename", visible=False) + + button.click(fn=lambda: gr.update(visible=True), inputs=[], outputs=[container]) + button_close.click(fn=lambda: gr.update(visible=False), inputs=[], outputs=[container]) + + def refresh(): + res = [] + + for pg in ui.stored_extra_pages: + pg.refresh() + res.append(pg.create_html(ui.tabname)) + + return res + + button_refresh.click(fn=refresh, inputs=[], outputs=ui.pages) + + return ui + + +def path_is_parent(parent_path, child_path): + parent_path = os.path.abspath(parent_path) + child_path = os.path.abspath(child_path) + + return os.path.commonpath([parent_path]) == os.path.commonpath([parent_path, child_path]) + + +def setup_ui(ui, gallery): + def save_preview(index, images, filename): + if len(images) == 0: + print("There is no image in gallery to save as a preview.") + return [page.create_html(ui.tabname) for page in ui.stored_extra_pages] + + index = int(index) + index = 0 if index < 0 else index + index = len(images) - 1 if index >= len(images) else index + + img_info = images[index if index >= 0 else 0] + image = image_from_url_text(img_info) + + is_allowed = False + for extra_page in ui.stored_extra_pages: + if any([path_is_parent(x, filename) for x in extra_page.allowed_directories_for_previews()]): + is_allowed = True + break + + assert is_allowed, f'writing to {filename} is not allowed' + + image.save(filename) + + return [page.create_html(ui.tabname) for page in ui.stored_extra_pages] + + ui.button_save_preview.click( + fn=save_preview, + _js="function(x, y, z){console.log(x, y, z); return [selected_gallery_index(), y, z]}", + inputs=[ui.preview_target_filename, gallery, ui.preview_target_filename], + outputs=[*ui.pages] + ) diff --git a/modules/ui_extra_networks_hypernets.py b/modules/ui_extra_networks_hypernets.py new file mode 100644 index 00000000..312dbaf0 --- /dev/null +++ b/modules/ui_extra_networks_hypernets.py @@ -0,0 +1,34 @@ +import os + +from modules import shared, ui_extra_networks + + +class ExtraNetworksPageHypernetworks(ui_extra_networks.ExtraNetworksPage): + def __init__(self): + super().__init__('Hypernetworks') + + def refresh(self): + shared.reload_hypernetworks() + + def list_items(self): + for name, path in shared.hypernetworks.items(): + path, ext = os.path.splitext(path) + previews = [path + ".png", path + ".preview.png"] + + preview = None + for file in previews: + if os.path.isfile(file): + preview = "./file=" + file.replace('\\', '/') + "?mtime=" + str(os.path.getmtime(file)) + break + + yield { + "name": name, + "filename": path, + "preview": preview, + "prompt": f"", + "local_preview": path + ".png", + } + + def allowed_directories_for_previews(self): + return [shared.cmd_opts.hypernetwork_dir] + diff --git a/modules/ui_extra_networks_textual_inversion.py b/modules/ui_extra_networks_textual_inversion.py new file mode 100644 index 00000000..e4a6e3bf --- /dev/null +++ b/modules/ui_extra_networks_textual_inversion.py @@ -0,0 +1,32 @@ +import os + +from modules import ui_extra_networks, sd_hijack + + +class ExtraNetworksPageTextualInversion(ui_extra_networks.ExtraNetworksPage): + def __init__(self): + super().__init__('Textual Inversion') + self.allow_negative_prompt = True + + def refresh(self): + sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings(force_reload=True) + + def list_items(self): + for embedding in sd_hijack.model_hijack.embedding_db.word_embeddings.values(): + path, ext = os.path.splitext(embedding.filename) + preview_file = path + ".preview.png" + + preview = None + if os.path.isfile(preview_file): + preview = "./file=" + preview_file.replace('\\', '/') + "?mtime=" + str(os.path.getmtime(preview_file)) + + yield { + "name": embedding.name, + "filename": embedding.filename, + "preview": preview, + "prompt": embedding.name, + "local_preview": path + ".preview.png", + } + + def allowed_directories_for_previews(self): + return list(sd_hijack.model_hijack.embedding_db.embedding_dirs) diff --git a/script.js b/script.js index 3345e32b..97e0bfcf 100644 --- a/script.js +++ b/script.js @@ -13,6 +13,7 @@ function get_uiCurrentTabContent() { } uiUpdateCallbacks = [] +uiLoadedCallbacks = [] uiTabChangeCallbacks = [] optionsChangedCallbacks = [] let uiCurrentTab = null @@ -20,6 +21,9 @@ let uiCurrentTab = null function onUiUpdate(callback){ uiUpdateCallbacks.push(callback) } +function onUiLoaded(callback){ + uiLoadedCallbacks.push(callback) +} function onUiTabChange(callback){ uiTabChangeCallbacks.push(callback) } @@ -38,8 +42,15 @@ function executeCallbacks(queue, m) { queue.forEach(function(x){runCallback(x, m)}) } +var executedOnLoaded = false; + document.addEventListener("DOMContentLoaded", function() { var mutationObserver = new MutationObserver(function(m){ + if(!executedOnLoaded && gradioApp().querySelector('#txt2img_prompt')){ + executedOnLoaded = true; + executeCallbacks(uiLoadedCallbacks); + } + executeCallbacks(uiUpdateCallbacks, m); const newTab = get_uiCurrentTab(); if ( newTab && ( newTab !== uiCurrentTab ) ) { @@ -53,7 +64,7 @@ document.addEventListener("DOMContentLoaded", function() { /** * Add a ctrl+enter as a shortcut to start a generation */ - document.addEventListener('keydown', function(e) { +document.addEventListener('keydown', function(e) { var handled = false; if (e.key !== undefined) { if((e.key == "Enter" && (e.metaKey || e.ctrlKey || e.altKey))) handled = true; diff --git a/scripts/xy_grid.py b/scripts/xy_grid.py index 6629f5d5..b1badec9 100644 --- a/scripts/xy_grid.py +++ b/scripts/xy_grid.py @@ -11,7 +11,6 @@ import modules.scripts as scripts import gradio as gr from modules import images, paths, sd_samplers, processing, sd_models, sd_vae -from modules.hypernetworks import hypernetwork from modules.processing import process_images, Processed, StableDiffusionProcessingTxt2Img from modules.shared import opts, cmd_opts, state import modules.shared as shared @@ -94,28 +93,6 @@ def confirm_checkpoints(p, xs): raise RuntimeError(f"Unknown checkpoint: {x}") -def apply_hypernetwork(p, x, xs): - if x.lower() in ["", "none"]: - name = None - else: - name = hypernetwork.find_closest_hypernetwork_name(x) - if not name: - raise RuntimeError(f"Unknown hypernetwork: {x}") - hypernetwork.load_hypernetwork(name) - - -def apply_hypernetwork_strength(p, x, xs): - hypernetwork.apply_strength(x) - - -def confirm_hypernetworks(p, xs): - for x in xs: - if x.lower() in ["", "none"]: - continue - if not hypernetwork.find_closest_hypernetwork_name(x): - raise RuntimeError(f"Unknown hypernetwork: {x}") - - def apply_clip_skip(p, x, xs): opts.data["CLIP_stop_at_last_layers"] = x @@ -208,8 +185,6 @@ axis_options = [ AxisOption("Prompt order", str_permutations, apply_order, format_value=format_value_join_list), AxisOption("Sampler", str, apply_sampler, format_value=format_value, confirm=confirm_samplers, choices=lambda: [x.name for x in sd_samplers.samplers]), AxisOption("Checkpoint name", str, apply_checkpoint, format_value=format_value, confirm=confirm_checkpoints, cost=1.0, choices=lambda: list(sd_models.checkpoints_list)), - AxisOption("Hypernetwork", str, apply_hypernetwork, format_value=format_value, confirm=confirm_hypernetworks, cost=0.2, choices=lambda: list(shared.hypernetworks)), - AxisOption("Hypernet str.", float, apply_hypernetwork_strength), AxisOption("Sigma Churn", float, apply_field("s_churn")), AxisOption("Sigma min", float, apply_field("s_tmin")), AxisOption("Sigma max", float, apply_field("s_tmax")), @@ -291,7 +266,6 @@ def draw_xy_grid(p, xs, ys, x_labels, y_labels, cell, draw_legend, include_lone_ class SharedSettingsStackHelper(object): def __enter__(self): self.CLIP_stop_at_last_layers = opts.CLIP_stop_at_last_layers - self.hypernetwork = opts.sd_hypernetwork self.vae = opts.sd_vae def __exit__(self, exc_type, exc_value, tb): @@ -299,9 +273,6 @@ class SharedSettingsStackHelper(object): modules.sd_models.reload_model_weights() modules.sd_vae.reload_vae_weights() - hypernetwork.load_hypernetwork(self.hypernetwork) - hypernetwork.apply_strength() - opts.data["CLIP_stop_at_last_layers"] = self.CLIP_stop_at_last_layers diff --git a/style.css b/style.css index 3a515ebd..5e8bc2ca 100644 --- a/style.css +++ b/style.css @@ -132,13 +132,6 @@ } #roll_col > button { - min-width: 2em; - min-height: 2em; - max-width: 2em; - max-height: 2em; - flex-grow: 0; - padding-left: 0.25em; - padding-right: 0.25em; margin: 0.1em 0; } @@ -146,9 +139,10 @@ min-width: 0 !important; max-width: 8em !important; margin-right: 1em; + gap: 0; } #interrogate, #deepbooru{ - margin: 0em 0.25em 0.9em 0.25em; + margin: 0em 0.25em 0.5em 0.25em; min-width: 8em; max-width: 8em; } @@ -157,8 +151,17 @@ min-width: 8em !important; } +#txt2img_styles_row, #img2img_styles_row{ + gap: 0.25em; + margin-top: 0.5em; +} + +#txt2img_styles_row > button, #img2img_styles_row > button{ + margin: 0; +} + #txt2img_styles, #img2img_styles{ - margin-top: 1em; + padding: 0; } #txt2img_styles ul, #img2img_styles ul{ @@ -635,17 +638,21 @@ canvas[key="mask"] { background-color: rgb(31 41 55 / var(--tw-bg-opacity)); } -.gr-button-tool{ +.gr-button-tool, .gr-button-tool-top{ max-width: 2.5em; min-width: 2.5em !important; height: 2.4em; - margin: 1.6em 0.7em 0.55em 0; } -#tab_modelmerger .gr-button-tool{ +.gr-button-tool{ margin: 0.6em 0em 0.55em 0; } +.gr-button-tool-top, #settings .gr-button-tool{ + margin: 1.6em 0.7em 0.55em 0; +} + + #modelmerger_results_container{ margin-top: 1em; overflow: visible; @@ -763,81 +770,88 @@ footer { line-height: 2.4em; } -/* The following handles localization for right-to-left (RTL) languages like Arabic. -The rtl media type will only be activated by the logic in javascript/localization.js. -If you change anything above, you need to make sure it is RTL compliant by just running -your changes through converters like https://cssjanus.github.io/ or https://rtlcss.com/. -Then, you will need to add the RTL counterpart only if needed in the rtl section below.*/ -@media rtl { - /* this part was added manually */ - :host { - direction: rtl; - } - select, .file-preview, .gr-text-input, .output-html:has(.performance), #ti_progress { - direction: ltr; - } - #script_list > label > select, - #x_type > label > select, - #y_type > label > select { - direction: rtl; - } - .gr-radio, .gr-checkbox{ - margin-left: 0.25em; - } +#txt2img_extra_networks, #img2img_extra_networks{ + margin-top: -1em; +} - /* automatically generated with few manual modifications */ - .performance .time { - margin-right: unset; - margin-left: 0; - } - .justify-center.overflow-x-scroll { - justify-content: right; - } - .justify-center.overflow-x-scroll button:first-of-type { - margin-left: unset; - margin-right: auto; - } - .justify-center.overflow-x-scroll button:last-of-type { - margin-right: unset; - margin-left: auto; - } - #settings fieldset span.text-gray-500, #settings .gr-block.gr-box span.text-gray-500, #settings label.block span{ - margin-right: unset; - margin-left: 8em; - } - #txt2img_progressbar, #img2img_progressbar, #ti_progressbar{ - right: unset; - left: 0; - } - .progressDiv .progress{ - padding: 0 0 0 8px; - text-align: left; - } - #lightboxModal{ - left: unset; - right: 0; - } - .modalPrev, .modalNext{ - border-radius: 3px 0 0 3px; - } - .modalNext { - right: unset; - left: 0; - border-radius: 0 3px 3px 0; - } - #imageARPreview{ - left:unset; - right:0px; - } - #txt2img_skip, #img2img_skip{ - right: unset; - left: 0px; - } - #context-menu{ - box-shadow:-1px 1px 2px #CE6400; - } - .gr-box > div > div > input.gr-text-input{ - right: unset; - left: 0.5em; - } +.extra-networks > div > [id *= '_extra_']{ + margin: 0.3em; } + +.extra-network-cards .nocards{ + margin: 1.25em 0.5em 0.5em 0.5em; +} + +.extra-network-cards .nocards h1{ + font-size: 1.5em; + margin-bottom: 1em; +} + +.extra-network-cards .nocards li{ + margin-left: 0.5em; +} + +.extra-network-cards .card{ + display: inline-block; + margin: 0.5em; + width: 16em; + height: 24em; + box-shadow: 0 0 5px rgba(128, 128, 128, 0.5); + border-radius: 0.2em; + position: relative; + + background-size: auto 100%; + background-position: center; + overflow: hidden; + cursor: pointer; + + background-image: url('./file=html/card-no-preview.png') +} + +.extra-network-cards .card:hover{ + box-shadow: 0 0 2px 0.3em rgba(0, 128, 255, 0.35); +} + +.extra-network-cards .card .actions .additional{ + display: none; +} + +.extra-network-cards .card .actions{ + position: absolute; + bottom: 0; + left: 0; + right: 0; + padding: 0.5em; + color: white; + background: rgba(0,0,0,0.5); + box-shadow: 0 0 0.25em 0.25em rgba(0,0,0,0.5); + text-shadow: 0 0 0.2em black; +} + +.extra-network-cards .card .actions:hover{ + box-shadow: 0 0 0.75em 0.75em rgba(0,0,0,0.5) !important; +} + +.extra-network-cards .card .actions .name{ + font-size: 1.7em; + font-weight: bold; + line-break: anywhere; +} + +.extra-network-cards .card .actions:hover .additional{ + display: block; +} + +.extra-network-cards .card ul{ + margin: 0.25em 0 0.75em 0.25em; + cursor: unset; +} + +.extra-network-cards .card ul a{ + cursor: pointer; +} + +.extra-network-cards .card ul a:hover{ + color: red; +} + diff --git a/webui.py b/webui.py index 865a7300..e8dd822a 100644 --- a/webui.py +++ b/webui.py @@ -9,16 +9,18 @@ from fastapi import FastAPI from fastapi.middleware.cors import CORSMiddleware from fastapi.middleware.gzip import GZipMiddleware -from modules import import_hook, errors +from modules import import_hook, errors, extra_networks +from modules import extra_networks_hypernet, ui_extra_networks_hypernets, ui_extra_networks_textual_inversion from modules.call_queue import wrap_queued_call, queue_lock, wrap_gradio_gpu_call from modules.paths import script_path import torch + # Truncate version number of nightly/local build of PyTorch to not cause exceptions with CodeFormer or Safetensors if ".dev" in torch.__version__ or "+git" in torch.__version__: torch.__version__ = re.search(r'[\d.]+[\d]', torch.__version__).group(0) -from modules import shared, devices, sd_samplers, upscaler, extensions, localization, ui_tempdir +from modules import shared, devices, sd_samplers, upscaler, extensions, localization, ui_tempdir, ui_extra_networks import modules.codeformer_model as codeformer import modules.extras import modules.face_restoration @@ -84,10 +86,17 @@ def initialize(): shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights())) shared.opts.onchange("sd_vae", wrap_queued_call(lambda: modules.sd_vae.reload_vae_weights()), call=False) shared.opts.onchange("sd_vae_as_default", wrap_queued_call(lambda: modules.sd_vae.reload_vae_weights()), call=False) - shared.opts.onchange("sd_hypernetwork", wrap_queued_call(lambda: shared.reload_hypernetworks())) - shared.opts.onchange("sd_hypernetwork_strength", modules.hypernetworks.hypernetwork.apply_strength) shared.opts.onchange("temp_dir", ui_tempdir.on_tmpdir_changed) + shared.reload_hypernetworks() + + ui_extra_networks.intialize() + ui_extra_networks.register_page(ui_extra_networks_textual_inversion.ExtraNetworksPageTextualInversion()) + ui_extra_networks.register_page(ui_extra_networks_hypernets.ExtraNetworksPageHypernetworks()) + + extra_networks.initialize() + extra_networks.register_extra_network(extra_networks_hypernet.ExtraNetworkHypernet()) + if cmd_opts.tls_keyfile is not None and cmd_opts.tls_keyfile is not None: try: @@ -209,6 +218,15 @@ def webui(): modules.sd_models.list_models() + shared.reload_hypernetworks() + + ui_extra_networks.intialize() + ui_extra_networks.register_page(ui_extra_networks_textual_inversion.ExtraNetworksPageTextualInversion()) + ui_extra_networks.register_page(ui_extra_networks_hypernets.ExtraNetworksPageHypernetworks()) + + extra_networks.initialize() + extra_networks.register_extra_network(extra_networks_hypernet.ExtraNetworkHypernet()) + if __name__ == "__main__": if cmd_opts.nowebui: -- cgit v1.2.3 From 3262e825cc542ff634e6ba2e3a162eafdc6c1bba Mon Sep 17 00:00:00 2001 From: Takuma Mori Date: Sat, 21 Jan 2023 17:42:04 +0900 Subject: add --xformers-flash-attention option & impl --- modules/sd_hijack_optimizations.py | 26 ++++++++++++++++++++++++-- modules/shared.py | 1 + 2 files changed, 25 insertions(+), 2 deletions(-) (limited to 'modules/sd_hijack_optimizations.py') diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 4fa54329..9967359b 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -290,7 +290,19 @@ def xformers_attention_forward(self, x, context=None, mask=None): q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b n h d', h=h), (q_in, k_in, v_in)) del q_in, k_in, v_in - out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None) + + if shared.cmd_opts.xformers_flash_attention: + op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp + fw, bw = op + if not fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v, attn_bias=None)): + # print('xformers_attention_forward', q.shape, k.shape, v.shape) + # Flash Attention is not availabe for the input arguments. + # Fallback to default xFormers' backend. + op = None + else: + op = None + + out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=op) out = rearrange(out, 'b n h d -> b n (h d)', h=h) return self.to_out(out) @@ -365,7 +377,17 @@ def xformers_attnblock_forward(self, x): q = q.contiguous() k = k.contiguous() v = v.contiguous() - out = xformers.ops.memory_efficient_attention(q, k, v) + if shared.cmd_opts.xformers_flash_attention: + op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp + fw, bw = op + if not fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v)): + # print('xformers_attnblock_forward', q.shape, k.shape, v.shape) + # Flash Attention is not availabe for the input arguments. + # Fallback to default xFormers' backend. + op = None + else: + op = None + out = xformers.ops.memory_efficient_attention(q, k, v, op=op) out = rearrange(out, 'b (h w) c -> b c h w', h=h) out = self.proj_out(out) return x + out diff --git a/modules/shared.py b/modules/shared.py index 72fb1934..23328adf 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -57,6 +57,7 @@ parser.add_argument("--realesrgan-models-path", type=str, help="Path to director parser.add_argument("--clip-models-path", type=str, help="Path to directory with CLIP model file(s).", default=None) parser.add_argument("--xformers", action='store_true', help="enable xformers for cross attention layers") parser.add_argument("--force-enable-xformers", action='store_true', help="enable xformers for cross attention layers regardless of whether the checking code thinks you can run it; do not make bug reports if this fails to work") +parser.add_argument("--xformers-flash-attention", action='store_true', help="enable xformers with Flash Attention to improve reproducibility (supported for SD2.x or variant only)") parser.add_argument("--deepdanbooru", action='store_true', help="does not do anything") parser.add_argument("--opt-split-attention", action='store_true', help="force-enables Doggettx's cross-attention layer optimization. By default, it's on for torch cuda.") parser.add_argument("--opt-sub-quad-attention", action='store_true', help="enable memory efficient sub-quadratic cross-attention layer optimization") -- cgit v1.2.3 From 59146621e256269b85feb536edeb745da20daf68 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Mon, 23 Jan 2023 16:40:20 +0300 Subject: better support for xformers flash attention on older versions of torch --- modules/errors.py | 12 +++++++++++ modules/sd_hijack_optimizations.py | 42 ++++++++++++++++---------------------- 2 files changed, 30 insertions(+), 24 deletions(-) (limited to 'modules/sd_hijack_optimizations.py') diff --git a/modules/errors.py b/modules/errors.py index a10e8708..f6b80dbb 100644 --- a/modules/errors.py +++ b/modules/errors.py @@ -24,6 +24,18 @@ See https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#stable """) +already_displayed = {} + + +def display_once(e: Exception, task): + if task in already_displayed: + return + + display(e, task) + + already_displayed[task] = 1 + + def run(code, task): try: code() diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 9967359b..74452709 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -9,7 +9,7 @@ from torch import einsum from ldm.util import default from einops import rearrange -from modules import shared +from modules import shared, errors from modules.hypernetworks import hypernetwork from .sub_quadratic_attention import efficient_dot_product_attention @@ -279,6 +279,21 @@ def sub_quad_attention(q, k, v, q_chunk_size=1024, kv_chunk_size=None, kv_chunk_ ) +def get_xformers_flash_attention_op(q, k, v): + if not shared.cmd_opts.xformers_flash_attention: + return None + + try: + flash_attention_op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp + fw, bw = flash_attention_op + if fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v, attn_bias=None)): + return flash_attention_op + except Exception as e: + errors.display_once(e, "enabling flash attention") + + return None + + def xformers_attention_forward(self, x, context=None, mask=None): h = self.heads q_in = self.to_q(x) @@ -291,18 +306,7 @@ def xformers_attention_forward(self, x, context=None, mask=None): q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b n h d', h=h), (q_in, k_in, v_in)) del q_in, k_in, v_in - if shared.cmd_opts.xformers_flash_attention: - op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp - fw, bw = op - if not fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v, attn_bias=None)): - # print('xformers_attention_forward', q.shape, k.shape, v.shape) - # Flash Attention is not availabe for the input arguments. - # Fallback to default xFormers' backend. - op = None - else: - op = None - - out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=op) + out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=get_xformers_flash_attention_op(q, k, v)) out = rearrange(out, 'b n h d -> b n (h d)', h=h) return self.to_out(out) @@ -377,17 +381,7 @@ def xformers_attnblock_forward(self, x): q = q.contiguous() k = k.contiguous() v = v.contiguous() - if shared.cmd_opts.xformers_flash_attention: - op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp - fw, bw = op - if not fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v)): - # print('xformers_attnblock_forward', q.shape, k.shape, v.shape) - # Flash Attention is not availabe for the input arguments. - # Fallback to default xFormers' backend. - op = None - else: - op = None - out = xformers.ops.memory_efficient_attention(q, k, v, op=op) + out = xformers.ops.memory_efficient_attention(q, k, v, op=get_xformers_flash_attention_op(q, k, v)) out = rearrange(out, 'b (h w) c -> b c h w', h=h) out = self.proj_out(out) return x + out -- cgit v1.2.3 From e3b53fd295aca784253dfc8668ec87b537a72f43 Mon Sep 17 00:00:00 2001 From: brkirch Date: Wed, 25 Jan 2023 00:23:10 -0500 Subject: Add UI setting for upcasting attention to float32 Adds "Upcast cross attention layer to float32" option in Stable Diffusion settings. This allows for generating images using SD 2.1 models without --no-half or xFormers. In order to make upcasting cross attention layer optimizations possible it is necessary to indent several sections of code in sd_hijack_optimizations.py so that a context manager can be used to disable autocast. Also, even though Stable Diffusion (and Diffusers) only upcast q and k, unfortunately my findings were that most of the cross attention layer optimizations could not function unless v is upcast also. --- modules/devices.py | 6 +- modules/processing.py | 2 +- modules/sd_hijack_optimizations.py | 159 +++++++++++++++++++++++-------------- modules/shared.py | 1 + modules/sub_quadratic_attention.py | 4 +- 5 files changed, 108 insertions(+), 64 deletions(-) (limited to 'modules/sd_hijack_optimizations.py') diff --git a/modules/devices.py b/modules/devices.py index 0981ef80..6b36622c 100644 --- a/modules/devices.py +++ b/modules/devices.py @@ -108,6 +108,10 @@ def autocast(disable=False): return torch.autocast("cuda") +def without_autocast(disable=False): + return torch.autocast("cuda", enabled=False) if torch.is_autocast_enabled() and not disable else contextlib.nullcontext() + + class NansException(Exception): pass @@ -125,7 +129,7 @@ def test_for_nans(x, where): message = "A tensor with all NaNs was produced in Unet." if not shared.cmd_opts.no_half: - message += " This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try using --no-half commandline argument to fix this." + message += " This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the \"Upcast cross attention layer to float32\" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this." elif where == "vae": message = "A tensor with all NaNs was produced in VAE." diff --git a/modules/processing.py b/modules/processing.py index 2d186ba0..a850082d 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -611,7 +611,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed: if p.n_iter > 1: shared.state.job = f"Batch {n+1} out of {p.n_iter}" - with devices.autocast(disable=devices.unet_needs_upcast): + with devices.without_autocast() if devices.unet_needs_upcast else devices.autocast(): samples_ddim = p.sample(conditioning=c, unconditional_conditioning=uc, seeds=seeds, subseeds=subseeds, subseed_strength=p.subseed_strength, prompts=prompts) x_samples_ddim = [decode_first_stage(p.sd_model, samples_ddim[i:i+1].to(dtype=devices.dtype_vae))[0].cpu() for i in range(samples_ddim.size(0))] diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 74452709..c02d954c 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -9,7 +9,7 @@ from torch import einsum from ldm.util import default from einops import rearrange -from modules import shared, errors +from modules import shared, errors, devices from modules.hypernetworks import hypernetwork from .sub_quadratic_attention import efficient_dot_product_attention @@ -52,18 +52,25 @@ def split_cross_attention_forward_v1(self, x, context=None, mask=None): 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): - end = i + 2 - s1 = einsum('b i d, b j d -> b i j', q[i:end], k[i:end]) - s1 *= self.scale + dtype = q.dtype + if shared.opts.upcast_attn: + q, k, v = q.float(), k.float(), v.float() - s2 = s1.softmax(dim=-1) - del s1 + with devices.without_autocast(disable=not shared.opts.upcast_attn): + r1 = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device, dtype=q.dtype) + for i in range(0, q.shape[0], 2): + end = i + 2 + s1 = einsum('b i d, b j d -> b i j', q[i:end], k[i:end]) + s1 *= self.scale + + s2 = s1.softmax(dim=-1) + del s1 + + r1[i:end] = einsum('b i j, b j d -> b i d', s2, v[i:end]) + del s2 + del q, k, v - r1[i:end] = einsum('b i j, b j d -> b i d', s2, v[i:end]) - del s2 - del q, k, v + r1 = r1.to(dtype) r2 = rearrange(r1, '(b h) n d -> b n (h d)', h=h) del r1 @@ -82,45 +89,52 @@ def split_cross_attention_forward(self, x, context=None, mask=None): k_in = self.to_k(context_k) v_in = self.to_v(context_v) - k_in *= self.scale - - del context, x - - 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, dtype=q.dtype) - - mem_free_total = get_available_vram() - - gb = 1024 ** 3 - tensor_size = q.shape[0] * q.shape[1] * k.shape[1] * q.element_size() - modifier = 3 if q.element_size() == 2 else 2.5 - mem_required = tensor_size * modifier - steps = 1 - - if mem_required > mem_free_total: - steps = 2 ** (math.ceil(math.log(mem_required / mem_free_total, 2))) - # print(f"Expected tensor size:{tensor_size/gb:0.1f}GB, cuda free:{mem_free_cuda/gb:0.1f}GB " - # f"torch free:{mem_free_torch/gb:0.1f} total:{mem_free_total/gb:0.1f} steps:{steps}") + dtype = q_in.dtype + if shared.opts.upcast_attn: + q_in, k_in, v_in = q_in.float(), k_in.float(), v_in if v_in.device.type == 'mps' else v_in.float() - if steps > 64: - max_res = math.floor(math.sqrt(math.sqrt(mem_free_total / 2.5)) / 8) * 64 - raise RuntimeError(f'Not enough memory, use lower resolution (max approx. {max_res}x{max_res}). ' - f'Need: {mem_required / 64 / gb:0.1f}GB free, Have:{mem_free_total / gb:0.1f}GB free') - - slice_size = q.shape[1] // steps if (q.shape[1] % steps) == 0 else q.shape[1] - for i in range(0, q.shape[1], slice_size): - end = i + slice_size - s1 = einsum('b i d, b j d -> b i j', q[:, i:end], k) - - s2 = s1.softmax(dim=-1, dtype=q.dtype) - del s1 - - r1[:, i:end] = einsum('b i j, b j d -> b i d', s2, v) - del s2 + with devices.without_autocast(disable=not shared.opts.upcast_attn): + k_in = k_in * self.scale + + del context, x + + 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, dtype=q.dtype) + + mem_free_total = get_available_vram() + + gb = 1024 ** 3 + tensor_size = q.shape[0] * q.shape[1] * k.shape[1] * q.element_size() + modifier = 3 if q.element_size() == 2 else 2.5 + mem_required = tensor_size * modifier + steps = 1 + + if mem_required > mem_free_total: + steps = 2 ** (math.ceil(math.log(mem_required / mem_free_total, 2))) + # print(f"Expected tensor size:{tensor_size/gb:0.1f}GB, cuda free:{mem_free_cuda/gb:0.1f}GB " + # f"torch free:{mem_free_torch/gb:0.1f} total:{mem_free_total/gb:0.1f} steps:{steps}") + + if steps > 64: + max_res = math.floor(math.sqrt(math.sqrt(mem_free_total / 2.5)) / 8) * 64 + raise RuntimeError(f'Not enough memory, use lower resolution (max approx. {max_res}x{max_res}). ' + f'Need: {mem_required / 64 / gb:0.1f}GB free, Have:{mem_free_total / gb:0.1f}GB free') + + slice_size = q.shape[1] // steps if (q.shape[1] % steps) == 0 else q.shape[1] + for i in range(0, q.shape[1], slice_size): + end = i + slice_size + s1 = einsum('b i d, b j d -> b i j', q[:, i:end], k) + + s2 = s1.softmax(dim=-1, dtype=q.dtype) + del s1 + + r1[:, i:end] = einsum('b i j, b j d -> b i d', s2, v) + del s2 + + del q, k, v - del q, k, v + r1 = r1.to(dtype) r2 = rearrange(r1, '(b h) n d -> b n (h d)', h=h) del r1 @@ -204,12 +218,20 @@ def split_cross_attention_forward_invokeAI(self, x, context=None, mask=None): context = default(context, x) context_k, context_v = hypernetwork.apply_hypernetworks(shared.loaded_hypernetworks, context) - k = self.to_k(context_k) * self.scale + k = self.to_k(context_k) v = self.to_v(context_v) del context, context_k, context_v, x - q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v)) - r = einsum_op(q, k, v) + dtype = q.dtype + if shared.opts.upcast_attn: + q, k, v = q.float(), k.float(), v if v.device.type == 'mps' else v.float() + + with devices.without_autocast(disable=not shared.opts.upcast_attn): + k = k * self.scale + + q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v)) + r = einsum_op(q, k, v) + r = r.to(dtype) return self.to_out(rearrange(r, '(b h) n d -> b n (h d)', h=h)) # -- End of code from https://github.com/invoke-ai/InvokeAI -- @@ -234,8 +256,14 @@ def sub_quad_attention_forward(self, x, context=None, mask=None): k = k.unflatten(-1, (h, -1)).transpose(1,2).flatten(end_dim=1) v = v.unflatten(-1, (h, -1)).transpose(1,2).flatten(end_dim=1) + dtype = q.dtype + if shared.opts.upcast_attn: + q, k = q.float(), k.float() + x = sub_quad_attention(q, k, v, q_chunk_size=shared.cmd_opts.sub_quad_q_chunk_size, kv_chunk_size=shared.cmd_opts.sub_quad_kv_chunk_size, chunk_threshold=shared.cmd_opts.sub_quad_chunk_threshold, use_checkpoint=self.training) + x = x.to(dtype) + x = x.unflatten(0, (-1, h)).transpose(1,2).flatten(start_dim=2) out_proj, dropout = self.to_out @@ -268,15 +296,16 @@ def sub_quad_attention(q, k, v, q_chunk_size=1024, kv_chunk_size=None, kv_chunk_ query_chunk_size = q_tokens kv_chunk_size = k_tokens - return efficient_dot_product_attention( - q, - k, - v, - query_chunk_size=q_chunk_size, - kv_chunk_size=kv_chunk_size, - kv_chunk_size_min = kv_chunk_size_min, - use_checkpoint=use_checkpoint, - ) + with devices.without_autocast(disable=q.dtype == v.dtype): + return efficient_dot_product_attention( + q, + k, + v, + query_chunk_size=q_chunk_size, + kv_chunk_size=kv_chunk_size, + kv_chunk_size_min = kv_chunk_size_min, + use_checkpoint=use_checkpoint, + ) def get_xformers_flash_attention_op(q, k, v): @@ -306,8 +335,14 @@ def xformers_attention_forward(self, x, context=None, mask=None): q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b n h d', h=h), (q_in, k_in, v_in)) del q_in, k_in, v_in + dtype = q.dtype + if shared.opts.upcast_attn: + q, k = q.float(), k.float() + out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=get_xformers_flash_attention_op(q, k, v)) + out = out.to(dtype) + out = rearrange(out, 'b n h d -> b n (h d)', h=h) return self.to_out(out) @@ -378,10 +413,14 @@ def xformers_attnblock_forward(self, x): v = self.v(h_) b, c, h, w = q.shape q, k, v = map(lambda t: rearrange(t, 'b c h w -> b (h w) c'), (q, k, v)) + dtype = q.dtype + if shared.opts.upcast_attn: + q, k = q.float(), k.float() q = q.contiguous() k = k.contiguous() v = v.contiguous() out = xformers.ops.memory_efficient_attention(q, k, v, op=get_xformers_flash_attention_op(q, k, v)) + out = out.to(dtype) out = rearrange(out, 'b (h w) c -> b c h w', h=h) out = self.proj_out(out) return x + out diff --git a/modules/shared.py b/modules/shared.py index 4ce1209b..6a0b96cb 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -410,6 +410,7 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "comma_padding_backtrack": OptionInfo(20, "Increase coherency by padding from the last comma within n tokens when using more than 75 tokens", gr.Slider, {"minimum": 0, "maximum": 74, "step": 1 }), "CLIP_stop_at_last_layers": OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}), "extra_networks_default_multiplier": OptionInfo(1.0, "Multiplier for extra networks", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}), + "upcast_attn": OptionInfo(False, "Upcast cross attention layer to float32"), })) options_templates.update(options_section(('compatibility', "Compatibility"), { diff --git a/modules/sub_quadratic_attention.py b/modules/sub_quadratic_attention.py index 55052815..05595323 100644 --- a/modules/sub_quadratic_attention.py +++ b/modules/sub_quadratic_attention.py @@ -67,7 +67,7 @@ def _summarize_chunk( max_score, _ = torch.max(attn_weights, -1, keepdim=True) max_score = max_score.detach() exp_weights = torch.exp(attn_weights - max_score) - exp_values = torch.bmm(exp_weights, value) + exp_values = torch.bmm(exp_weights, value) if query.device.type == 'mps' else torch.bmm(exp_weights, value.to(exp_weights.dtype)).to(value.dtype) max_score = max_score.squeeze(-1) return AttnChunk(exp_values, exp_weights.sum(dim=-1), max_score) @@ -129,7 +129,7 @@ def _get_attention_scores_no_kv_chunking( ) attn_probs = attn_scores.softmax(dim=-1) del attn_scores - hidden_states_slice = torch.bmm(attn_probs, value) + hidden_states_slice = torch.bmm(attn_probs, value) if query.device.type == 'mps' else torch.bmm(attn_probs, value.to(attn_probs.dtype)).to(value.dtype) return hidden_states_slice -- cgit v1.2.3