Feedback Transformer - Pytorch
Simple implementation of Feedback Transformer in Pytorch. They improve on Transformer-XL by having each token have access to the representations of all previous layers through time. This is achieved by aggregating the outputs of all layers into a shared memory, which each token across layers can attend to at each time step.
The main drawback is longer training time, due to its non-parallel nature. But I thought I'd build it to further exploration and research into this line of work.
I also took the liberty to add some various enhancements, including pre-normalization, GLU gated feedforwards, as well as simplified T5 relative positional embeddings.
Install
$ pip install feedback-transformer-pytorch
Usage
import torch
from feedback_transformer_pytorch import FeedbackTransformer
model = FeedbackTransformer(
num_tokens = 20000, # number of tokens
dim = 512, # dimension
depth = 6, # depth
seq_len = 2, # the sequence length of each segment or window
mem_len = 256, # length of the memory buffer
dim_head = 64, # dimension of each head
heads = 8, # number of heads
attn_dropout = 0.1, # attention dropout
ff_dropout = 0.1 # feedforward dropout
).cuda()
x = torch.randint(0, 20000, (2, 64)).cuda()
model(x) # (2, 64, 20000)
If you would like to have fine control over the memory (when to detach, etc), you can do it with some extra keyword arguments on .forward
import torch
from feedback_transformer_pytorch import FeedbackTransformer
model = FeedbackTransformer(
num_tokens = 20000,
dim = 512,
depth = 6,
seq_len = 32,
mem_len = 256
).cuda()
x1 = torch.randint(0, 20000, (2, 32)).cuda()
x2 = torch.randint(0, 20000, (2, 32)).cuda()
x3 = torch.randint(0, 20000, (2, 32)).cuda()
out1, mem1 = model(x1, return_memory = True)
out2, mem2 = model(x2, memory = mem1, return_memory = True)
out3, mem3 = model(x3, memory = mem2, return_memory = True) # (2, 32, 20000)
Citations
@misc{fan2021addressing,
title = {Addressing Some Limitations of Transformers with Feedback Memory},
author = {Angela Fan and Thibaut Lavril and Edouard Grave and Armand Joulin and Sainbayar Sukhbaatar},
year = {2021},
eprint = {2002.09402},
archivePrefix = {arXiv},
primaryClass = {cs.LG}
}
297 Dec 17, 2022
50 Nov 17, 2022
9.7k Jan 01, 2023
3.6k Jan 08, 2023
92 Dec 12, 2022
2 Dec 16, 2022
173 Dec 29, 2022
4 Jul 01, 2022
33 Dec 31, 2022
348 Dec 23, 2022
10 Oct 16, 2022
3 Feb 18, 2022
3k Jan 05, 2023
13 Dec 23, 2022
181 Dec 27, 2022
165 Dec 28, 2022
223 Dec 17, 2022
115 Dec 23, 2022
[email protected]">
4 Oct 14, 2022