Artificial IntelliTools

DeepSeek-OCR Isn’t About OCR , It’s About Token Compression DeepSeek can use just 100 vision tokens to represent what would normally require 1,000 text tokens, and then decode it back with 97% accuracy.

  🚀 Playing with DeepSeek R1 Distill Qwen 1.5B: So I tried out DeepSeek R1, the distilled 1.5B small version because resources. lol. Its a tiny yet powerful 1.5B parameter Non quantized model, using Group Relative Policy Optimization (GRPO) for reinforcement learning. All of DeepSeek’s models are open source, and DeepSeek has been making news lately about how they managed to pull off powerful models using little resources and even skipping steps everyone thought were necessary to develop powerful models. They did all this with the smallest budgets and not-so-powerful GPUs. And the fact that it’s open source is a whole other thing, because before all this, the only powerful open-source models were Meta’s lineup of LLaMA models. So, having a new player that’s just as powerful, costs a tenth of what industry leaders charge, and is open source is a really big deal. 🔍 Here are some of my takeaways from the hands-on experience: And remember, this is the Tiny Tiny version, only 1.5B. ...

  I Tried out DeepSeek's Janus Pro 7B, a text-to-image generation and visual understanding model: First, three reasons it's been making news: 1. Multimodal: Unlike models like DALL-E 3 that focus on image generation, Janus Pro can handle multiple tasks within a single framework. It can generate images from text prompts, analyze and interpret images, and handle text-based tasks. Basically it can not only generate images for you but you can ask it questions about the images you upload or generate as well. This range and versatility make it a more rounded AI tool. 2. Performance and efficiency: It reportedly outperforms OpenAI's DALL-E 3, Google's MU3 Gen, and Stability AI's SDXL on key AI benchmarks. But whats making all the buzz and fuss is that it achieves all this using less powerful Nvidia chips, raising questions about the necessity of expensive hardware in AI development. 3.Open -source: DeepSeek has made the model and code for Janus Pro available on Hugging Fac...