Merge branch 'development' of github.com:pbaylies/stable-diffusion into pbaylies-development

backend/invoke_ai_web_server.py

@@ -799,9 +799,6 @@ class InvokeAIWebServer:
                 rfc_dict['init_image_path'] = parameters[
                     'init_img'
                 ]  # TODO: Noncompliant
-                rfc_dict[
-                    'sampler'
-                ] = 'ddim'  # TODO: FIX ME WHEN IMG2IMG SUPPORTS ALL SAMPLERS
                 if 'init_mask' in parameters:
                     rfc_dict['mask_hash'] = calculate_init_img_hash(
                         self.get_image_path_from_url(parameters['init_mask'])

frontend/README.md

@@ -2,24 +2,22 @@
 
 ## Run
 
-- `python backend/server.py` serves both frontend and backend at http://localhost:9090
+- `python scripts/dream.py --web` serves both frontend and backend at
+  http://localhost:9090
 
 ## Evironment
 
 Install [node](https://nodejs.org/en/download/) (includes npm) and optionally
 [yarn](https://yarnpkg.com/getting-started/install).
 
-From `frontend/` run `npm install` / `yarn install` to install the frontend packages.
+From `frontend/` run `npm install` / `yarn install` to install the frontend
+packages.
 
 ## Dev
 
 1. From `frontend/`, run `npm dev` / `yarn dev` to start the dev server.
-2. Note the address it starts up on (probably `http://localhost:5173/`).
-3. Edit `backend/server.py`'s `additional_allowed_origins` to include this address, e.g.
-   `additional_allowed_origins = ['http://localhost:5173']`.
-4. Leaving the dev server running, open a new terminal and go to the project root.
-5. Run `python backend/server.py`.
-6. Navigate to the dev server address e.g. `http://localhost:5173/`.
+2. Run `python scripts/dream.py --web`.
+3. Navigate to the dev server address e.g. `http://localhost:5173/`.
 
 To build for dev: `npm build-dev` / `yarn build-dev`
 
@@ -28,10 +26,3 @@ To build for production: `npm build` / `yarn build`
 ## TODO
 
 - Search repo for "TODO"
-- My one gripe with Chakra: no way to disable all animations right now and drop the dependence on
-  `framer-motion`. I would prefer to save the ~30kb on bundle and have zero animations. This is on
-  the Chakra roadmap. See https://github.com/chakra-ui/chakra-ui/pull/6368 for last discussion on
-  this. Need to check in on this issue periodically.
-- Mobile friendly layout
-- Proper image gallery/viewer/manager
-- Help tooltips and such

frontend/dist/index.html

@@ -6,7 +6,7 @@
   <meta name="viewport" content="width=device-width, initial-scale=1.0" />
   <title>InvokeAI - A Stable Diffusion Toolkit</title>
   <link rel="shortcut icon" type="icon" href="/assets/favicon.0d253ced.ico" />
-  <script type="module" crossorigin src="/assets/index.045a5291.js"></script>
+  <script type="module" crossorigin src="/assets/index.d9916e7a.js"></script>
   <link rel="stylesheet" href="/assets/index.853a336f.css">
 </head>
 

frontend/src/app/constants.ts

@@ -7,8 +7,6 @@ export const SAMPLERS: Array<string> = [
   'k_lms',
   'k_dpm_2',
   'k_dpm_2_a',
-  'k_dpm_fast',
-  'k_dpm_adaptive',
   'k_euler',
   'k_euler_a',
   'k_heun',

ldm/dream/args.py

@@ -189,7 +189,6 @@ class Args(object):
             switches.append(f'--perlin {a["perlin"]}')
         if a['threshold'] > 0:
             switches.append(f'--threshold {a["threshold"]}')
-            
         if a['grid']:
             switches.append('--grid')
         if a['seamless']:

ldm/dream/generator/txt2img.py

@@ -32,7 +32,7 @@ class Txt2Img(Generator):
             if self.free_gpu_mem and self.model.model.device != self.model.device:
                 self.model.model.to(self.model.device)
                                 
-            sampler.make_schedule(ddim_num_steps=steps, ddim_eta=ddim_eta, verbose=True)
+            sampler.make_schedule(ddim_num_steps=steps, ddim_eta=ddim_eta, verbose=False)
 
             samples, _ = sampler.sample(
                 batch_size                   = 1,

ldm/models/diffusion/ksampler.py

@@ -79,17 +79,9 @@ class KSampler(Sampler):
             ddim_eta=0.0,
             verbose=False,
         )
-        self.model = outer_model
+        self.model          = outer_model
         self.ddim_num_steps = ddim_num_steps
-        sigmas = K.sampling.get_sigmas_karras(
-            n=ddim_num_steps,
-            sigma_min=self.model.sigmas[0].item(),
-            sigma_max=self.model.sigmas[-1].item(),
-            rho=7.,
-            device=self.device,
-            # Birch-san recommends this, but it doesn't match the call signature in his branch of k-diffusion
-            # concat_zero=False
-        )
+        sigmas = self.model.get_sigmas(ddim_num_steps)
         self.sigmas = sigmas
         
     # ALERT: We are completely overriding the sample() method in the base class, which
@@ -133,7 +125,8 @@ class KSampler(Sampler):
 
         # sigmas = self.model.get_sigmas(S)
         # sigmas are now set up in make_schedule - we take the last steps items
-        sigmas = self.sigmas[-S:]
+        sigmas = self.sigmas[-S-1:]
+
         if x_T is not None:
             x = x_T * sigmas[0]
         else:
@@ -147,7 +140,7 @@ class KSampler(Sampler):
             'uncond': unconditional_conditioning,
             'cond_scale': unconditional_guidance_scale,
         }
-        print(f'>> Sampling with k__{self.schedule}')
+        print(f'>> Sampling with k_{self.schedule}')
         return (
             K.sampling.__dict__[f'sample_{self.schedule}'](
                 model_wrap_cfg, x, sigmas, extra_args=extra_args,

ldm/util.py

@@ -218,7 +218,7 @@ def rand_perlin_2d(shape, res, fade = lambda t: 6*t**5 - 15*t**4 + 10*t**3):
     delta = (res[0] / shape[0], res[1] / shape[1])
     d = (shape[0] // res[0], shape[1] // res[1])
 
-    grid = torch.stack(torch.meshgrid(torch.arange(0, res[0], delta[0]), torch.arange(0, res[1], delta[1])), dim = -1) % 1
+    grid = torch.stack(torch.meshgrid(torch.arange(0, res[0], delta[0]), torch.arange(0, res[1], delta[1]), indexing='ij'), dim = -1) % 1
     angles = 2*math.pi*torch.rand(res[0]+1, res[1]+1)
     gradients = torch.stack((torch.cos(angles), torch.sin(angles)), dim = -1)
 
@@ -230,4 +230,4 @@ def rand_perlin_2d(shape, res, fade = lambda t: 6*t**5 - 15*t**4 + 10*t**3):
     n01 = dot(tile_grads([0, -1],[1, None]), [0, -1])
     n11 = dot(tile_grads([1, None], [1, None]), [-1,-1])
     t = fade(grid[:shape[0], :shape[1]])
-    return math.sqrt(2) * torch.lerp(torch.lerp(n00, n10, t[..., 0]), torch.lerp(n01, n11, t[..., 0]), t[..., 1])
+    return math.sqrt(2) * torch.lerp(torch.lerp(n00, n10, t[..., 0]), torch.lerp(n01, n11, t[..., 0]), t[..., 1])