Sunny Madra’s Post

In case you missed, it Groq achieved over 40k input tokens PER SECOND. You maybe asking why input speed is so important, as context window limits are expanding (128k for GPT4o, and 1Mil in the case of Gemni 1.5) these speeds will be increasingly important.

Q2_K quantized model meets its FP32 base

This is so true for at least most of out of box quantized model i.e from TheBloke on HF. Still quantization makes these LLM more accessible but I will never use them in production deployment, the result is unpredictable

Check out another example of Davinci in action: https://lnkd.in/eHj2fwQk Using information from a datasheet Davinci will automatically populate and link components.

weekend updates: 1. complete CAN node circuits ✅ 2. write code for sending and receiving CAN frames ✅ 3. end 2 end simulations for the circuit ✅ 4. create a demo video ✅ the next steps will be plugin the sniffer into the car and reading its data.

@ Feature Pyramid Network, it actually considers the high level feature map with lower level feature maps to actually create a new set of feature maps, then why do we use an extension of FPN ? it gives you a depth knowledge about one part of v4....

SW100X Allseas: hull mould completed. Think, plan, act!

Fping, which stands for "fast ping" can run ping tests to multiple hosts in parallel and really fast (even with 1 millisecond intervals!). https://hubs.ly/Q02szHkP0 #Fping

Want to Achieve FP32 Accuracy for INT8 Inference Using Quantization Aware Training . Explore https://lnkd.in/gH6xqgss

There are really neat applications of Wolfram System modeller in Engineering and Environmental Modeling / Green Economy. #Wolfram #greeneconomy #circulareconomy