Tag: #dit
Writing on GPUs, LLMs, MLOps, Kubernetes — and mindset · 5 posts
5 posts are tagged #dit; the most recent was published on 2026-05-16.
Often tagged together: #diffusion-transformer 3 #2026-03 2 #ai-papers 2 #deep-dive 2 #english 2
Most read in the last 90 days:
- HunyuanVideo and LTX-2 Complete Analysis: Architecture, Performance, and Practical Guide to Open-Source Video Generation Models
- AI Video Generation 2026 — Sora 2 / Veo 3 / Kling 2 / Hailuo / Runway Gen-4 / Luma Ray 2 / HunyuanVideo Deep Dive
- Foundation Model Architectures 2026 — Beyond the Transformer / Mamba 2 / Hyena / RWKV / RetNet / Griffin / Jamba / xLSTM / TTT / DiT / MoE / Flash Attention 3 Deep Dive
Foundation Model Architectures 2026 — Beyond the Transformer / Mamba 2 / Hyena / RWKV / RetNet / Griffin / Jamba / xLSTM / TTT / DiT / MoE / Flash Attention 3 Deep Dive ♪ Listenable
In 2026 the foundation-model world is no longer Transformer-only. Vaswani 2017 "Attention is All You Need" remains the standard, but next to it stand state-space models (Mamba, Mamba 2), the linear-RNN renaissance (RWKV,
2026-05-16 · 22 min read #foundation-models#transformer#attention-is-all-you-need#vaswani#mambaAI Video Generation 2026 — Sora 2 / Veo 3 / Kling 2 / Hailuo / Runway Gen-4 / Luma Ray 2 / HunyuanVideo Deep Dive ♪ Listenable
Two years after Sora 1 stunned the film industry in Feb 2024, the AI video generation landscape of May 2026 has hardened around three camps — closed-source SOTA (Sora 2, Veo 3, Kling 2, Hailuo), industry-standard video t
2026-05-15 · 27 min read #ai-video#video-generation#sora#veo#klingDiffusion Model Paper Survey: Evolution of Image Generation from DDPM to Stable Diffusion, DiT, and SDXL
A comprehensive survey covering the DDPM/DDIM forward-reverse diffusion theory, score-based models, Latent Diffusion (Stable Diffusion) VAE+U-Net architecture, Classifier-free Guidance, DiT (Diffusion Transformer) adaLN-
2026-03-12 · 18 min read #ai-papers#diffusion-model#ddpm#stable-diffusion#ditDiffusion Transformer (DiT) Architecture Analysis: The Shift from U-Net to Transformer
An analysis of the Scalable Diffusion Models with Transformers (DiT) paper. We cover the motivations behind transitioning from U-Net backbones to Transformers, adaLN-Zero conditioning, scaling laws, and the downstream im
2026-03-03 · 10 min read #ai-papers#diffusion-transformer#dit#generative-ai#image-generationHunyuanVideo and LTX-2 Complete Analysis: Architecture, Performance, and Practical Guide to Open-Source Video Generation Models ♪ Listenable
A deep analysis of Tencent HunyuanVideo (13B) and Lightricks LTX-2 (19B) architectures, training methodologies, and performance benchmarks, along with a comprehensive comparison of the open-source video generation ecosys
2026-03-01 · 30 min read #hunyuan-video#ltx-video#ltx2#text-to-video#image-to-video