bleedingedge.ai
@bleedingedgeai
Followers
9K
Following
15
Media
51
Statuses
171
Joined October 2022
New open weights LLM from @MistralAI. params.json:.- hidden_dim / dim = 14336/4096 => 3.5X MLP expand.- n_heads / n_kv_heads = 32/8 => 4X multiquery.- "moe" => mixture of experts 8X top 2 👀. Likely related code: . Oddly absent: an over-rehearsed
0
0
7
mixtral-8x7b-32kseqlen from @MistralAI. Mixture of Experts? 🤔.
magnet:?xt=urn:btih:5546272da9065eddeb6fcd7ffddeef5b75be79a7&dn=mixtral-8x7b-32kseqlen&tr=udp%3A%2F%3A6969%2Fannounce&tr=http%3A%2F%3A80%2Fannounce. RELEASE a6bbd9affe0c2725c1b7410d66833e24.
1
2
5
NVIDIA announces model for high-resolution text to video generation. It can generate videos with resolution up to 1280 x 2048.
4
6
35
StabilityAI released StableLM: an open source LLM with 3b and 7b parameters. RLHF and larger models coming soon.
1
0
15
Meta releases DINOv2 the first method for training computer vision models that uses self-supervised learning (no labeling needed) to achieve industry standard results.
1
0
11
OpenAI released Consistency Models: a new family of generative models that support fast one-step generation of images.
1
1
19
Meta releases the Segment Anything Project: .- SA-1B dataset containing 1-Billion segmentation masks.- Segment Anything Model (SAM) for general image and video segmentation
1
0
18
Databricks open-sources Dolly: a 6b instruction-tuned LLM similar to ChatGPT. It's built off of EleutherAI's GPT-J-6B.
2
4
51
Damo releases an open-source text to video model with 1.7B parameters. The demo only requires 16GB CPU RAM and 16GB GPU RAM to run. Try it out on Hugging Face below:
2
3
19
Researchers introduce Video-P2P: Video Editing with.Cross-attention Control.
1
2
15
Microsoft introduces 365 Copilot - bringing AI assistance into Word, PowerPoint, Excel, Outlook, Teams, Viva, and Power Platform.
2
1
28