>
G
Model Details
Domain:
Model Access:
Open weights (non-commercial)
Citations:
1197
Introduction
GLM-130B (ICLR 2023) is an open bilingual (English & Chinese) bidirectional dense model with 130 billion parameters, pre-trained using the General Language Model (GLM) algorithm1. It is designed to support inference tasks with the 130B parameters on a single A100 (40G * 8) or V100 (32G * 8) server. As of July 3rd, 2022, GLM-130B has been trained on over 400 billion text tokens (200B each for Chinese and English)
Benchmarking
FLOPs3.55e+23
Notes: "96 NVIDIA A100 (40G * 8) servers for 2 months" 312 TFLOPS/GPU * 96 servers * 8 GPU/server * 2 months * 32.5% utilization = 4.037e23 utilization rate - citation from the paper: "we report hardware FLOPs utilization (HFU) of 43.3% and model FLOPs utilization (MFU) of 32.5% due to re-materialization." Aligns pretty well with 6ND: 6 * 400B * 130B = 3.12E23 Geometric mean: sqrt(4.037e23 * 3.12e23) = 3.549e23
Training
Training Code Accessibilitynon commercial license. looks like inference but not training code: https://github.com/THUDM/GLM-130B/blob/main/MODEL_LICENSE
HardwareNVIDIA A100 SXM4 40 GB
Hardware Quantity768
Size Notes: 400B "We completed the 400B-token training and evaluation of GLM-130B in July, and subsequently released the model and pre-training details in August 2022. " from https://arxiv.org/pdf/2406.12793 "As of July 3rd, 2022, GLM-130B has been trained on over 400 billion text tokens (200B each for Chinese and English)"
Parameters
Parameters130000000000
Notes: Dense model
Authors
Aohan Zeng, Xiao Liu, Zhengxiao Du, Zihan Wang, Hanyu Lai, Ming Ding, Zhuoyi Yang, Yifan Xu, Wendi Zheng, Xiao Xia, Weng Lam Tam, Zixuan Ma, Yufei Xue, Jidong Zhai, Wenguang Chen, Peng Zhang, Yuxiao Dong, Jie Tang
