A fall can shorten a person’s life. Preventing them with wearable robotics requires 2 steps: detecting fall-related conditions and using predictions to activate a wearable device. This team has proposed exactly such a detection-and-activation system:
AI HARDWARE
-
Condor Galaxy 1 Powers BTLM and Jais Open-Source LLMs
By
–
Condor Galaxy 1 (CG-1) was used to train BTLM, the top-performing open-source 3B parameter LLM. Now, CG-1 has been used to train Jais, the best open-source Arabic model. CG-1 is moving the open-source community forward. Contact us to train on CG-1 http://
cerebras.net/contact-us -
Groq LPU Processing Speed Increases to 240 Tokens Per Second
By
–
Our fast Language Processing Unit™ (LPU™) just got faster: 100 to 240 Tokens per second per user in 3 weeks! https://
youtu.be/hk4QGpQAvSY
Join us for #GroqDay on Sept 7 at 1:00 PT to learn more – http://
groq.com/groqday -
CodeLlama PR merged with 4-bit quantization inference benchmarks
By
–
The CodeLlama PR just got merged: https://
github.com/Lightning-AI/l
it-gpt/pull/472
… When I tried it with bnb's 4-bit Normal Float quantization, the 34B Instruct and Python variants used about 20 Gb for inference: -
Police Safety Device Eliminates Officer Risk During Traffic Stops
By
–
Los policías de todo el mundo ponen en riesgo su vida cuando detienen un vehículo para identificar al conductor o ponerle una multa. Pero eso está a punto de cambiar.
— Juan Merodio (@juanmerodio) 1 septembre 2023
Este dispositivo hará el trabajo policial mucho más seguro, evitando que el agente tenga que bajarse del… pic.twitter.com/2pUaKwc7cjLos policías de todo el mundo ponen en riesgo su vida cuando detienen un vehículo para identificar al conductor o ponerle una multa. Pero eso está a punto de cambiar. Este dispositivo hará el trabajo policial mucho más seguro, evitando que el agente tenga que bajarse del
-
Cerebras-GPT: First External Customer AI Model Achievement
By
–
Jais was trained on the CG-1 AI supercomputer. We are super proud of the work we did on Cerebras-GPT and BTLM. This is the first time an external customer used our system to produce a world class model – another huge milestone for Cerebras!
-
34B Parameter Models Now Run on Laptops, Massive Hardware Progress
By
–
Crazy how 34 billions parameters models seemed huge and unmanageable outside of a data center just maybe 1.5 years ago. Now it’s laptop stuff https://t.co/gXCg770BOM
— Thomas Wolf (@Thom_Wolf) 31 août 2023Crazy how 34 billions parameters models seemed huge and unmanageable outside of a data center just maybe 1.5 years ago. Now it’s laptop stuff
-
Google Cloud MultiSlice Scales ML Training Across TPU Pods
By
–
The @googlecloud MultiSlice system (see blog post in tweet below) is an external version of some internal software+hardware we've been using for a while to scale some of our largest ML training runs across multiple TPUv4 and TPUv5e pods. Scaling made easy!
-
H100 GPUs lack gaming hardware, potential for CAVE rendering demo
By
–
The H100 GPUs don’t have either rasterization or ray tracing hardware, so conventional game engines can’t run on them. I do think it would be a fun to do some kind of super high end 240 fps CAVE rendering demo with a cluster!
-
Jais Arabic LLM Outperforms LLaMa 13B Efficiently
By
–
(1/2) Jais is the world's best-performing open-source Arabic language model. And in English, it performs as well as LLaMa 13B despite using a fraction of the training data and a fraction of the electricity to train. Contact us to learn more: http://
cerebras.net/contact-us