linuzj/graph-data-quantum
Viewer
•
Updated
•
14.5k
•
123
•
2
Cong Yu
Benyucong
AI & ML interests
Machine Learning Systems
Recent Activity
liked
a dataset
26 days ago
linuzj/graph-data-quantum
liked
a model
about 1 month ago
Skywork/Skywork-R1V-38B
liked
a model
6 months ago
meta-llama/Llama-3.1-8B-Instruct
Organizations
Benyucong's activity
liked
a
dataset
26 days ago
liked
a
model
about 1 month ago
liked
a
model
6 months ago
liked
2
models
7 months ago
upvoted
an
article
10 months ago
Article
Introduction to 3D Gaussian Splatting
•
61
liked
a
model
12 months ago
upvoted
a
paper
about 1 year ago
reacted to
akhaliq's
post with ❤️
about 1 year ago
Post
2313
Jamba
A Hybrid Transformer-Mamba Language Model
Jamba: A Hybrid Transformer-Mamba Language Model (2403.19887)
We present Jamba, a new base large language model based on a novel hybrid Transformer-Mamba mixture-of-experts (MoE) architecture. Specifically, Jamba interleaves blocks of Transformer and Mamba layers, enjoying the benefits of both model families. MoE is added in some of these layers to increase model capacity while keeping active parameter usage manageable. This flexible architecture allows resource- and objective-specific configurations. In the particular configuration we have implemented, we end up with a powerful model that fits in a single 80GB GPU. Built at large scale, Jamba provides high throughput and small memory footprint compared to vanilla Transformers, and at the same time state-of-the-art performance on standard language model benchmarks and long-context evaluations. Remarkably, the model presents strong results for up to 256K tokens context length. We study various architectural decisions, such as how to combine Transformer and Mamba layers, and how to mix experts, and show that some of them are crucial in large scale modeling. We also describe several interesting properties of these architectures which the training and evaluation of Jamba have revealed, and plan to release checkpoints from various ablation runs, to encourage further exploration of this novel architecture. We make the weights of our implementation of Jamba publicly available under a permissive license.
A Hybrid Transformer-Mamba Language Model
Jamba: A Hybrid Transformer-Mamba Language Model (2403.19887)
We present Jamba, a new base large language model based on a novel hybrid Transformer-Mamba mixture-of-experts (MoE) architecture. Specifically, Jamba interleaves blocks of Transformer and Mamba layers, enjoying the benefits of both model families. MoE is added in some of these layers to increase model capacity while keeping active parameter usage manageable. This flexible architecture allows resource- and objective-specific configurations. In the particular configuration we have implemented, we end up with a powerful model that fits in a single 80GB GPU. Built at large scale, Jamba provides high throughput and small memory footprint compared to vanilla Transformers, and at the same time state-of-the-art performance on standard language model benchmarks and long-context evaluations. Remarkably, the model presents strong results for up to 256K tokens context length. We study various architectural decisions, such as how to combine Transformer and Mamba layers, and how to mix experts, and show that some of them are crucial in large scale modeling. We also describe several interesting properties of these architectures which the training and evaluation of Jamba have revealed, and plan to release checkpoints from various ablation runs, to encourage further exploration of this novel architecture. We make the weights of our implementation of Jamba publicly available under a permissive license.
upvoted
a
paper
over 1 year ago
liked
a
model
over 1 year ago
