GPT - gMLP
This repository will attempt to crack long context autoregressive language modeling (GPT) using variations of gMLPs. Specifically, it will contain a variant that does gMLP for local sliding windows. The hope is to be able to stretch a single GPU to be able to train context lengths of 4096 and above efficiently and well.
GPT is technically a misnomer now, since there will be no attention (transformer) at all contained in the architecture.
Install
$ pip install g-mlp-gpt
Usage
import torch
from g_mlp_gpt import gMLPGPT
model = gMLPGPT(
num_tokens = 20000,
dim = 512,
depth = 4,
seq_len = 1024,
window_size = (128, 256, 512, 1024) # window sizes for each depth
)
x = torch.randint(0, 20000, (1, 1000))
logits = model(x) # (1, 1000, 20000)
16k context length
import torch
from g_mlp_gpt import gMLPGPT
model = gMLPGPT(
num_tokens = 20000,
dim = 512,
seq_len = 16384,
depth = 8,
reversible = True,
window = (128, 128, 256, 512, 1024, 1024, 2048, 2048, 4096, 4096, 8192, 8192),
axial = (1, 1, 1, 1, 1, 1, 2, 2, 4, 4, 8, 8)
).cuda()
x = torch.randint(0, 20000, (1, 16384)).cuda()
logits = model(x) # (1, 16384, 20000)
Citations
@misc{liu2021pay,
title = {Pay Attention to MLPs},
author = {Hanxiao Liu and Zihang Dai and David R. So and Quoc V. Le},
year = {2021},
eprint = {2105.08050},
archivePrefix = {arXiv},
primaryClass = {cs.LG}
}
9 Nov 16, 2022
24 Oct 26, 2022
331 Jan 3, 2023
74 Dec 5, 2022
28 Feb 20, 2022
81 Nov 26, 2022
39 Nov 30, 2022
49 Dec 22, 2022
120 Dec 12, 2022
12 Dec 23, 2022
40 Oct 13, 2022
13 Jul 12, 2022
1.4k Jan 01, 2023
1 Jan 13, 2022
35 Sep 19, 2022
39 Jan 07, 2023
138 Apr 15, 2022
87 Jan 03, 2023
79 Dec 30, 2022
1 Mar 06, 2022
13 Dec 19, 2022
360 Dec 27, 2022
30 Nov 21, 2022
2 Jan 24, 2022
127 Jan 04, 2023
387 Jan 08, 2023
7 Mar 08, 2022
4 Dec 14, 2022
14 Dec 20, 2022