reformer-pytorch
Reformer, the Efficient Transformer, in Pytorch
It includes LSH attention, reversible network, and chunking. It has been validated with an auto-regressive task (enwik8).
Install
pip install reformer_pytorch
Usage
A simple Reformer language model
# should fit in ~ 5gb - 8k tokens
import torch
from reformer_pytorch import ReformerLM
model = ReformerLM(
num_tokens= 20000,
emb = 512,
depth = 12,
max_seq_len = 8192,
heads = 8,
lsh_dropout = 0.1,
causal = True, # auto-regressive or not
bucket_size = 64, # average size of qk per bucket, 64 was recommended in paper
n_hashes = 4, # 4 is permissible per author, 8 is the best but slower
ff_chunks = 200, # number of chunks for feedforward layer, make higher if there are memory issues
weight_tie = False, # tie parameters of each layer for no memory per additional depth
attn_chunks = 8, # process lsh attention in chunks, only way for memory to fit when scaling to 16k tokens
use_full_attn = False # use full self attention, for comparison
).cuda()
x = torch.randint(0, 20000, (1, 8192)).long().cuda()
y = model(x) # (1, 8192, 20000)
The Reformer (just a stack of reversible LSH attention)
# should fit in ~ 5gb - 8k embeddings
import torch
from reformer_pytorch import Reformer
model = Reformer(
emb = 512,
depth = 12,
max_seq_len = 8192,
heads = 8,
lsh_dropout = 0.1,
causal = True
).cuda()
x = torch.randn(1, 8192, 512).cuda()
y = model(x) # (1, 8192, 512)
Self Attention with LSH
import torch
from reformer_pytorch import LSHSelfAttention
attn = LSHSelfAttention(
emb = 128,
heads = 8,
bucket_size = 64,
n_hashes = 8,
causal = False
)
x = torch.randn(10, 1024, 128)
y = attn(x) # (10, 1024, 128)
LSH (locality sensitive hashing) Attention
import torch
from reformer_pytorch import LSHAttention
attn = LSHAttention(
bucket_size = 64,
n_hashes = 16,
causal = True
)
qk = torch.randn(10, 1024, 128)
v = torch.randn(10, 1024, 128)
attn_out, buckets = attn(qk, v) # (10, 1024, 128)
# buckets will contain the bucket number (post-argmax) of each token of each batch
Todo
- Make it so Reformer can be used as decoder where queries only attend to fed key/values
- Recurrence like Transformer XL
- All-attention learned memory key values