chener

Bruh, what kind of magic did DeepSeek pull off this time?

🔴Radar Görmedi, İDA Dibine Kadar Girdi! ABD merkezli Seasats'a ait Lightfish insansız deniz aracı, Filipinler açıklarında seyreden Çin Donanması'na ait Type 052D sınıfı muhribe fark edilmeden yaklaşarak görüntü aldı. Yaklaşık 240 bin dolarlık küçük bir İDA'nın milyar dolarlık bir savaş gemisine tespit edilmeden yaklaşabilmesi, düşük maliyetli insansız sistemlerin deniz savaşlarında oluşturduğu tehdidi gözler önüne serdi. Lightfish; elektro-optik ve kızılötesi kameralar, aktif-pasif sonar sistemleri ve elektronik harp görev yükleri taşıyabiliyor.

The Kimi K3 architecture figure for yesterday's big open-weight model release, along with some observations and thoughts. 1. Yes, it looks relatively complicated, but it's essentially a scaled-up production version of their Kimi Linear model they released last year (scaled up from 48B -> 2.8T; K3 is by far the biggest open-weight model right now) 2. The one new component compared to Kimi Linear is the LatentMoE. I omitted it in the figure below since it's already very crowded, but that's essentially the same LatentMoE as in Nemotron 3 Ultra (you can find it in my LLM Architecture Gallery if you are curious). The idea here is to compress (down-project) large linear layers similar to multi-head latent attention. 3. Kimi K3's overall trend (similar to Nemotron 3, DeepSeek V4, and others) is also towards better inference efficiency. That is, there are many components that replace existing components with efficiency-tweaked versions. I.e., MoE -> LatentMoE, regular attention -> multi-head latent attention and Kimi Delta Attention. (I also have short tutorials and write-ups in my gallery if you are curious about additional details). 4. The one component change that is not an efficiency tweak is attention residuals. Like DeepSeek V4 improved the residual path with mHC (manifold-constrained Hyper-Connections), attention residuals are a way to improve the residual path, but it works a bit differently. I.e., mHC made the residual path wider. Attention residuals (also already part of Kimi Linear) connect the residuals across layers; the connection itself uses an attention score for an important/contribution weight. According to the report, it improves the validation loss and downstream performance (a bit) consistently and adds about 4% in training cost and 2% in inference cost. 5. Interestingly, Kimi K3 got rid of all RoPE layers and uses NoPE (No Positional Embeddings) everywhere instead. (Again, this is inherited from Kimi Linear). In other architectures, the recent trend was towards RoPE in local attention layers (like sliding window attention) and NoPE in the global layers. There were a few architectures that only used NoPE everywhere, but this is the first frontier-level one as far as I know. 6. Kimi K3 now also has native multimodal support, which is great! There are several other interesting training tidbits in the technical report, but that's it from the architecture front so far. A really great release overall.

Kimi K3 已经如期开源，各位估计明天就能在各大平台的 Token Plan 使用 Kimi K3 了