Nachtrag 3: Es ging nicht zurück in die #VAE sondern daran vorbei Richtung Nordwest, aktuell über dem #Iran mit der groben Richtung westliches #Russland. Was bedeutet, dass die Flugzeuge offenbar nun leer sind, demzufolge etwas *nach* Madagaskar geliefert wurde, was zudem – mit Blick auf die …1/
#vae
Nachtrag 1: Am 22. Februar kamen beide Flugzeuge mit derselben Flugnummer wie beim Hinflug aus #Moskau. Der »Auftrag« begann als dort. Nachtrag 2: Für eine leere #Il-76 sollte die Strecke von #Madagaskar in die #VAE auch ohne Tankstop möglich sein. (Es gibt für einen Stop aber auch andere Gründe.)
Ich habe immer noch ein Auge auf die beiden #Frachtflugzeuge der russischen Energieministeriums, bei denen man vermutete, dass sie #russische Diplomat:innen aus der #Golfregion evakuieren.
Nachdem #RA-76841 und #RA-76362 von den #VAE nach #Dschibuti und von dort weiter nach #Madagaskar flogen, …1/6
RE: https://bsky.app/profile/did:plc:ijispkmryk5fl73ucqrkde43/post/3mfhcgzswes2b
#usa #israel #palästina #saudiarabien #ägypten #jordanien #vae #indonesien #pakistan #türkei #katar #kuwait #oman #bahrain #libanon #syrien : #krieg / #imperialismus / #zionismus / #reaktion / #usdiplomatie / #oic / #las / #gcc / #protest
„Die Äußerungen des US-Botschafters in Israel, Mike Huckabee, über eine mögliche Expansion Israels haben in der Region für heftigen Widerstand gesorgt.“
Lesenswerter Artikel: Wie #Äthiopien und die #VAE den "Bürgerkrieg" im #Sudan befeuern
https://www.dw.com/de/wie-%C3%A4thiopien-und-die-vae-den-b%C3%BCrgerkrieg-im-sudan-befeuern/a-76025552
#JeRecrute La Coopérative #LaSonnette recherche un·e Opérateur·ice de Service de Location
Longue Durée de #Vélo
Poste basé à #Redon #Bretagne - CDD à temps plein
📅 courant avril à mi-août 2026
💰️ Salaire brut : 1 923 € / mois, commun à tou·te·s les salarié·e·s de l’équipe
➡️ Détails : https://cloud.ker-redon.fr/index.php/s/KAnCMNPT8FaynTy
Mission : Gérer le quotidien
d’un service de location longue durée de #VAE pour le compte d’une collectivité.
Permis B et formation #Cycle requis.
E.Leclerc lance ses offres de recharge pour véhicules électriques
https://mac4ever.com/194785
#Mac4Ever #leclerc #Recharge #VAE
Project Arrow : le Canada veut prouver qu'il peut construire sa propre voiture électrique
https://mac4ever.com/194738
#Mac4Ever #borealis #Canada #projectarrow #VAE #vector
Marché automobile en France : l'électrique et le made in France se portent bien
https://mac4ever.com/194730
#Mac4Ever #France #Renault5 #VAE
Voiture électrique aux États-Unis : Trump coupe les aides et le marché s'effondre
https://mac4ever.com/194702
#Mac4Ever #Trump #VAE #VoitureAutonome
La Chine change de ton et pousse ses constructeurs de voitures électriques à négocier avec l'UE
https://mac4ever.com/194661
#Mac4Ever #Chine #Europe #VAE
Les voitures chinoises gagnent (un peu) de terrain en Europe
https://mac4ever.com/194610
#Mac4Ever #Chine #VAE
Je mets en vente mon #velo à assistance électrique #VAE de marque Moustache, modèle Samedi 27 Xroad 5 Open.
Acheté en 2022, 4600km, n'a pas roulé depuis 2an. Moteur Bosch 85Nm, batterie 625Wh. Taille M (utilisateur 1.59m - 1.76m). Chaîne, cassette, pneu arrière et plaquette de frein en très bon état (car remplacés).
Je suis près de #toulouse si ça intéresse quelqu'un.
Trek lance le Charter+, un vélo électrique à tout faire de 3 299 à 5 099 euros
https://mac4ever.com/194531
#Mac4Ever #trek #VAE #Velo
Je trouve ça hyper stylé ! Il était temps que quelqu'un propose les réparations de moteurs de #VAE sur #Strasbourg !
Réparons au lieu de jeter et racheter ! ✊
Some #ML für #dinosaurs: https://www.pnas.org/doi/10.1073/pnas.2527222122
🦕 🦖 🐾
This is a #VAE-based disentanglement approach to identify the main patterns of variation in dinosaur footprints. Nothing super fancy, but an interesting application nonetheless.
YES SUCCEEDED!!!
Just rendered an image at 944×1152 (slightly above 1024×1024) using Flux1-Schnell-FP8 on my 6700 XT, and it works! (Image 1 is the Real-ESRGAN 2× upscaled version)
Workflow 1: Sampling (Image 2)
Prompt executed → UNet generates the latent
Step 1 (model load + latent generation) took 419 seconds
Output: Latent tensor saved to disk
Workflow 2 : VAE Decode (Image 3)
Latent loaded → VAE decodes the image
Duration: 7.5 seconds
Advantage: UNet doesn’t need to stay in VRAM → VRAM freed, even on 12 GB GPUs
The problem with the stock LoadLatent Node
Dropdown only shows files if they were produced / annotated by a previous SaveLatent Node
Node is designed to pass latents inside a graph, not load arbitrary files from disk
Purpose: prevents accidentally loading wrong files
Workaround (Image 4)
Edited /ComfyUI/nodes.py, class LoadLatent
Hardcoded latent path → Node now loads directly from disk
Result: Workflow 2 runs instantly, UNet can be unloaded
Timing
Step 1 (model load + latent generation): 419 s
Step 2 (VAE decode): 7.5 s
Result: High-res images on a 12 GB RDNA2 GPU are now possible on Flux1-Schnell-FP8 without ComfyUI crashing! (Image 5 is the original output)
This might actually become my new Flux workflow: render quick 512×512 previews first (which works perfectly on RDNA2 GPUs), sort out the good ones, extract the seed from the PNG metadata, and then re-render only the selected images with the same seed using the split workflow at higher resolutions. This way, high-resolution Flux1-Schnell-FP8 renders become possible on 12 GB RDNA2 GPUs D:
Question at the end: Has anyone ever done this before? Because I have no clue xD
#ComfyUI #flux #Flux1SchnellFP8 #FP8 #AMD #RDNA2 #VAE #AIArt #Pixelfed #HighResolution #GPUOptimization #LatentWorkflow #AIWorkflow #AIHacks #RealESRGAN #Upscale #AIExperiment #CreativeAI #DigitalArt #AICommunity #python #linux #opensource #foss
Just rendered an image at 944×1152 (slightly above 1024×1024) using Flux1-Schnell-FP8 on my 6700 XT, and it works! (Image 1 is the Real-ESRGAN 2× upscaled version)
Workflow 1: Sampling (Image 2)
Prompt executed → UNet generates the latent
Step 1 (model load + latent generation) took 419 seconds
Output: Latent tensor saved to disk
Workflow 2 : VAE Decode (Image 3)
Latent loaded → VAE decodes the image
Duration: 7.5 seconds
Advantage: UNet doesn’t need to stay in VRAM → VRAM freed, even on 12 GB GPUs
The problem with the stock LoadLatent Node
Dropdown only shows files if they were produced / annotated by a previous SaveLatent Node
Node is designed to pass latents inside a graph, not load arbitrary files from disk
Purpose: prevents accidentally loading wrong files
Workaround (Image 4)
Edited /ComfyUI/nodes.py, class LoadLatent
Hardcoded latent path → Node now loads directly from disk
Result: Workflow 2 runs instantly, UNet can be unloaded
Timing
Step 1 (model load + latent generation): 419 s
Step 2 (VAE decode): 7.5 s
Result: High-res images on a 12 GB RDNA2 GPU are now possible on Flux1-Schnell-FP8 without ComfyUI crashing! (Image 5 is the original output)
This might actually become my new Flux workflow: render quick 512×512 previews first (which works perfectly on RDNA2 GPUs), sort out the good ones, extract the seed from the PNG metadata, and then re-render only the selected images with the same seed using the split workflow at higher resolutions. This way, high-resolution Flux1-Schnell-FP8 renders become possible on 12 GB RDNA2 GPUs D:
Question at the end: Has anyone ever done this before? Because I have no clue xD
#ComfyUI #flux #Flux1SchnellFP8 #FP8 #AMD #RDNA2 #VAE #AIArt #Pixelfed #HighResolution #GPUOptimization #LatentWorkflow #AIWorkflow #AIHacks #RealESRGAN #Upscale #AIExperiment #CreativeAI #DigitalArt #AICommunity #python #linux #opensource #foss
Die letzten #Winter - Wochen waren in #Deutschland zwar kalt, aber viel zu trocken. Das erhöht die Wahrscheinlichkeit auf einen weiteren #Hitzesommer mit #Dürre.
Die #Klimakrise führt über die eskalierende #Wasserkrise zum #Wasserbankrott - auch in #Europa.
Und währenddessen schlägt der #Fossilismus weiter um sich: Ob #Russland , #Iran , #USA , #SaudiArabien , #VAE …
Erneuerbare Energien sind #Friedensenergien , #Wohlstandsenergien , #Zukunftsenergien. #Solarpunk https://www.t-online.de/nachrichten/panorama/wetter/id_101096848/wetter-erste-hitze-prognose-fuer-den-sommer-mitten-im-winter.html
Finally got SDXL running!
I looked into the error
"modules.devices.NansException: A tensor with NaNs was produced in VAE"
and here’s what it means:
Briefly, how an image is generated with a diffusion model: The text encoder interprets the prompt, the UNet "dreams" iteratively in the latent space from noise into an image structure, and the VAE translates this latent vision into visible pixels.
A tensor is simply a multi-dimensional array of numbers, basically the data structure where the model stores all its calculations, like colors, intensities, and intermediate results of the image.
In this case, the VAE experienced a numerical instability: the latent tensor contained invalid values (NaNs), so the dreamed image could not be decoded correctly. In short: the model was still dreaming in the latent space, but the numbers “exploded” along the way (e.g., division by zero, overflow, or undefined operations).
#StableDiffusion #SDXL #AIArt #DiffusionModel #VAE #LatentSpace #Tensor #DigitalArt #CinematicArt #Kunst #KI #AI #DigitalIllustration #StilizedRealism #UrbanFantasy #Motion #DramaticLighting #FilmStill #AICommunity
I looked into the error
"modules.devices.NansException: A tensor with NaNs was produced in VAE"
and here’s what it means:
Briefly, how an image is generated with a diffusion model: The text encoder interprets the prompt, the UNet "dreams" iteratively in the latent space from noise into an image structure, and the VAE translates this latent vision into visible pixels.
A tensor is simply a multi-dimensional array of numbers, basically the data structure where the model stores all its calculations, like colors, intensities, and intermediate results of the image.
In this case, the VAE experienced a numerical instability: the latent tensor contained invalid values (NaNs), so the dreamed image could not be decoded correctly. In short: the model was still dreaming in the latent space, but the numbers “exploded” along the way (e.g., division by zero, overflow, or undefined operations).
#StableDiffusion #SDXL #AIArt #DiffusionModel #VAE #LatentSpace #Tensor #DigitalArt #CinematicArt #Kunst #KI #AI #DigitalIllustration #StilizedRealism #UrbanFantasy #Motion #DramaticLighting #FilmStill #AICommunity
Client Info
Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst