cs-fan-101

8 ms·

1.
Cerebras Inference now runs Llama 3.1-70B at 2100 tokens/s(cerebras.ai)
6 points by cs-fan-101 2y ago|0 comments
2.
Cerebras Launches the Fastest AI Inference(inference.cerebras.ai)
13 points by cs-fan-101 2y ago|1 comments
3.
Sparse Maximal Update Parameterization(arxiv.org)
2 points by cs-fan-101 2y ago|0 comments
4.
Cerebras Unveils Fastest AI Chip with Whopping 4T Transistors(cerebras.net)
6 points by cs-fan-101 3y ago|0 comments
5.
Cerebras launches PyTorch-based library for sparse training(cerebras.net)
2 points by cs-fan-101 3y ago|0 comments
6.
Jais – the world’s most advanced Arabic large language model
4 points by cs-fan-101 3y ago|0 comments
7.
Opentensor and Cerebras announce BTLM-3B-8K, a leading 3B param. language model(huggingface.co)
9 points by cs-fan-101 3y ago|2 comments
8.
by cs-fan-101 3y ago
Cerebras and Opentensor are pleased to announce BTLM-3B-8K (Bittensor Language Model), a new state-of-the-art 3 billion parameter open-source language model that achieves breakthrough accuracy across a dozen AI benchmarks. BTLM-3B-8K Highli
9.
by cs-fan-101 3y ago
[Cerebras employee here] Condor Galaxy 1 can support beyond 600 billion parameters. In standard config its 600B but it can scale to train upwards of 100T parameter models
10.
by cs-fan-101 3y ago
Cerebras announced today that it has built and sold a 4 exaFLOPS AI Supercomputer, named Condor Galaxy 1 (CG-1), to its strategic partner G42, the Abu Dhabi-based AI pioneer. Located in Santa Clara, CA, CG-1 is the first of nine interconnec
11.
Cerebras and G42 Unveil Condor Galaxy 1, a 4 ExaFLOPS AI Supercomputer(cerebras.net)
7 points by cs-fan-101 3y ago|4 comments
12.
Cerebras-GPT: Open Compute-Optimal Language Models Trained on Cerebras Cluster(arxiv.org)
97 points by cs-fan-101 3y ago|12 comments
13.
by cs-fan-101 3y ago
Recently, we announced in this post ( https://news.ycombinator.com/item?id=35343763#35345980 ) the release of Cerebras-GPT — a family of open-source GPT models trained on the Pile dataset using the Chinchilla formula. Today,
14.
by cs-fan-101 3y ago
Simply focusing on the "better in every regard" part of the comment. One example where Cerebras systems perform well is when a user is interested in training models that require long sequence lengths or high-resolution images. One
15.
by cs-fan-101 3y ago
Someone posted this repost from the Cerebras Discord earlier, but sharing for visibility - "We chose to train these models to 20 tokens per param to fit a scaling law to the Pile data set. These models are optimal for a fixed compute b