search

KimMeen / Time-LLM

[ICLR 2024] Official implementation of " 🦙 Time-LLM: Time Series Forecasting by Reprogramming Large Language Models"
https://arxiv.org/abs/2310.01728
Apache License 2.0
1.02k stars 179 forks source link

请问用RTX3090 24GB显存能跑动这个模型吗 #90

Closed NEUMX closed 1 month ago

JackKoLing commented 1 month ago

可以，llama的batchsize设为4可以跑起来，就是很慢

kwuking commented 1 month ago

可以的，缩小batch_size即可，或者减少层数，替换base model为gpt2或者bert，都可以有效减少显存开销

NEUMX commented 1 month ago

@kwuking 你好，请问能用这个显卡运行llama模型吗