From 0cc39f01a3c67f32873b7888e1e032d33626e269 Mon Sep 17 00:00:00 2001 From: damian0815 Date: Wed, 2 Nov 2022 01:18:50 +0100 Subject: [PATCH] report full size for fast latents and update conversion matrix for v1.5 --- backend/invoke_ai_web_server.py | 4 +++- ldm/invoke/generator/base.py | 18 +++++++++--------- 2 files changed, 12 insertions(+), 10 deletions(-) diff --git a/backend/invoke_ai_web_server.py b/backend/invoke_ai_web_server.py index 6f6837301d..b9fc147911 100644 --- a/backend/invoke_ai_web_server.py +++ b/backend/invoke_ai_web_server.py @@ -640,9 +640,11 @@ class InvokeAIWebServer: if generation_parameters['progress_latents']: image = self.generate.sample_to_lowres_estimated_image(sample) (width, height) = image.size + width *= 8 + height *= 8 buffered = io.BytesIO() image.save(buffered, format="PNG") - img_base64 = "data:image/jpeg;base64," + base64.b64encode(buffered.getvalue()).decode('UTF-8') + img_base64 = "data:image/png;base64," + base64.b64encode(buffered.getvalue()).decode('UTF-8') self.socketio.emit( "intermediateResult", { diff --git a/ldm/invoke/generator/base.py b/ldm/invoke/generator/base.py index 534fe2547c..3c6eca08a2 100644 --- a/ldm/invoke/generator/base.py +++ b/ldm/invoke/generator/base.py @@ -119,19 +119,19 @@ class Generator(): # write an approximate RGB image from latent samples for a single step to PNG def sample_to_lowres_estimated_image(self,samples): - # adapted from code by @erucipe and @keturn here: + # origingally adapted from code by @erucipe and @keturn here: # https://discuss.huggingface.co/t/decoding-latents-to-rgb-without-upscaling/23204/7 - # these numbers were determined empirically by @keturn - v1_4_latent_rgb_factors = torch.tensor([ - # R G B - [ 0.298, 0.207, 0.208], # L1 - [ 0.187, 0.286, 0.173], # L2 - [-0.158, 0.189, 0.264], # L3 - [-0.184, -0.271, -0.473], # L4 + # these updated numbers for v1.5 are from @torridgristle + v1_5_latent_rgb_factors = torch.tensor([ + # R G B + [ 0.3444, 0.1385, 0.0670], # L1 + [ 0.1247, 0.4027, 0.1494], # L2 + [-0.3192, 0.2513, 0.2103], # L3 + [-0.1307, -0.1874, -0.7445] # L4 ], dtype=samples.dtype, device=samples.device) - latent_image = samples[0].permute(1, 2, 0) @ v1_4_latent_rgb_factors + latent_image = samples[0].permute(1, 2, 0) @ v1_5_latent_rgb_factors latents_ubyte = (((latent_image + 1) / 2) .clamp(0, 1) # change scale from -1..1 to 0..1 .mul(0xFF) # to 0..255