huggingface.co/sensenova/Se...
- 9B
- Apache 2.0
- 56 layer executions via weight sharing
- Up to 256K context
- Hybrid attention + two stream mHC
huggingface.co/sensenova/Se...
- 9B
- Apache 2.0
- 56 layer executions via weight sharing
- Up to 256K context
- Hybrid attention + two stream mHC
0.6B & 9B, text + images, and you can index with 9B while serving queries with 0.6B.
A thread on how they work 🧵
huggingface.co/collections/...
0.6B & 9B, text + images, and you can index with 9B while serving queries with 0.6B.
A thread on how they work 🧵
huggingface.co/collections/...
SenseTime released SenseNova Loop-9B-Preview, an open-source 9B multimodal model that doubles its processing depth by reusing a shared block of layers.
SenseTime released SenseNova Loop-9B-Preview, an open-source 9B multimodal model that doubles its processing depth by reusing a shared block of layers.
📩 https://www.wearequeeraf.com/understand-the-lgbtqia-news-a-combination-of-biological-psychological-and-social-factors-how-generative-ai-describes-trans-people/<a href="/hashtag/%25F0%259F%2592%259B-intersex-open-letter" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#%F0%9F%92%9B-intersex-open-letter
📩 https://www.wearequeeraf.com/understand-the-lgbtqia-news-a-combination-of-biological-psychological-and-social-factors-how-generative-ai-describes-trans-people/<a href="/hashtag/%25F0%259F%2592%259B-intersex-open-letter" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#%F0%9F%92%9B-intersex-open-letter
xfolio.jp/portfolio/9b...
xfolio.jp/portfolio/9b...
en.defence-ua.com/industries/c...
en.defence-ua.com/industries/c...
youtu.be/uIVr09MoFqo
youtu.be/uIVr09MoFqo
That lets you encode a corpus with 9B, then search it with 0.6B queries. Normalize after truncating 9B vectors to the shared width.
Both sizes:
huggingface.co/collections/...
That lets you encode a corpus with 9B, then search it with 0.6B queries. Normalize after truncating 9B vectors to the shared width.
Both sizes:
huggingface.co/collections/...
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
Parece que empieza el invierno y el frío, pero eso es lo único que hace es que mis pezones se pongan más duros y mi mente busque calor
Deseando conocer a nuevos chicos con los que divertirme
Parece que empieza el invierno y el frío, pero eso es lo único que hace es que mis pezones se pongan más duros y mi mente busque calor
Deseando conocer a nuevos chicos con los que divertirme
It stacks layers instead of parameters. The decoder has 32 layers, but layers 4 through 27 run twice with shared weights, so a forward pass does 56 layers of work while the checkpoint stays at 9B.
It stacks layers instead of parameters. The decoder has 32 layers, but layers 4 through 27 run twice with shared weights, so a forward pass does 56 layers of work while the checkpoint stays at 9B.
𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗦𝗲𝗻𝘀𝗲𝗡𝗼𝘃𝗮 𝗟𝗼𝗼𝗽-𝟵𝗕-𝗣𝗿𝗲𝘃𝗶𝗲𝘄, 𝗻𝗼𝘄 𝗼𝗽𝗲𝗻 𝘀𝗼𝘂𝗿𝗰𝗲!
𝗦𝗰𝗮𝗹𝗶𝗻𝗴 𝗰𝗼𝗺𝗽𝘂𝘁𝗲 𝗱𝗲𝗽𝘁𝗵, 𝗻𝗼𝘁 𝗷𝘂𝘀𝘁 𝗺𝗼𝗱𝗲𝗹 𝘀𝗶𝘇𝗲: our 9B Looped Transformer uses parameter-shared loop...
https://x.com/i/web/status/2116151078219858346
𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗦𝗲𝗻𝘀𝗲𝗡𝗼𝘃𝗮 𝗟𝗼𝗼𝗽-𝟵𝗕-𝗣𝗿𝗲𝘃𝗶𝗲𝘄, 𝗻𝗼𝘄 𝗼𝗽𝗲𝗻 𝘀𝗼𝘂𝗿𝗰𝗲!
𝗦𝗰𝗮𝗹𝗶𝗻𝗴 𝗰𝗼𝗺𝗽𝘂𝘁𝗲 𝗱𝗲𝗽𝘁𝗵, 𝗻𝗼𝘁 𝗷𝘂𝘀𝘁 𝗺𝗼𝗱𝗲𝗹 𝘀𝗶𝘇𝗲: our 9B Looped Transformer uses parameter-shared loop...
https://x.com/i/web/status/2116151078219858346