issues
search
lucidrains
/
rotary-embedding-torch
Implementation of Rotary Embeddings, from the Roformer paper, in Pytorch
MIT License
573
stars
44
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Unused params if cache_if_possible=True when multiple rotary dimensions are used
#37
lukasschmit
closed
1 month ago
1
Slower than absolute positional embeddings?
#36
umarbutler
opened
2 months ago
4
Explicit casting instead of autocasting
#35
lminer
opened
2 months ago
0
Fix chi scale multiplication
#34
TimFelixBeyer
closed
2 months ago
1
Add sequence position interpolation to axial RoPE
#33
tasansal
closed
2 months ago
1
Fine-tuning Axial RoPE with frequency scaling?
#32
tasansal
opened
2 months ago
0
apply_rotary_emb - remove inplace operation
#31
blasscoc
closed
3 months ago
1
RoPE embeddings
#30
PRamoneda
opened
3 months ago
1
Request for YaRN
#29
VarunGumma
opened
3 months ago
0
Lastest commit incompatible with local_attention
#28
MarcusLoppe
closed
4 months ago
3
xPOS embeddings during inference
#27
VarunGumma
closed
4 months ago
2
LieRE: Generalizing Rotary Position Encodings. Beats RoPE-mixed by large margin and is much faster (compute-wise)
#26
kabachuha
closed
1 month ago
28
RoPE-Mixed: Improvement over Axial for n-D
#25
tasansal
opened
4 months ago
1
nan loss when training in fp8 with transformer engine
#24
saurabh-kataria
closed
4 months ago
1
Repeat order.
#23
AliYoussef97
closed
4 months ago
0
added indexing
#22
AlxSp
closed
8 months ago
0
Request for permission to publish a Rust port of this python module
#21
Mekadrom
closed
8 months ago
1
`torch.cat` failes in `apply_rotary_emb` when `freqs.shape[-1] == t.shape[-1]`, and `start_index = 0`
#20
mattaltberg
opened
8 months ago
1
RoPE on Images
#19
aaprasad
closed
9 months ago
1
caching frequency results in RuntimeError: Trying to backward through the graph a second time
#18
wren93
closed
6 months ago
3
Is 'broadcat' part of the API?
#17
rsxdalv
closed
11 months ago
5
Error caused by tensor-type seq_len
#16
cmunna0052
closed
11 months ago
1
Model hangs on eval
#15
GarrettMerz
opened
11 months ago
18
implementing on vision tranformers
#14
mukvnd
closed
1 year ago
2
Does this library support 2D RoPE embeddings?
#13
logicchains
closed
1 year ago
2
Support for sequence length ordering
#12
iiSeymour
closed
1 year ago
7
Bug in cache
#11
N0r9st
closed
1 year ago
1
Usage with x-transformers
#10
sonovice
opened
1 year ago
4
Bfloat16 support for use_xpos=True
#9
rostro36
closed
1 year ago
8
Using with xpos causes NaNs after rotating Q, K
#8
andersonbcdefg
closed
1 year ago
2
AttributeError: 'NoneType' object has no attribute 'to'
#7
yingzhao27
closed
1 year ago
3
freqs reference
#6
biirving
opened
1 year ago
0
Tricks for training with RoPE? Specific initialisers for QK projections?
#5
thorinf
opened
1 year ago
0
why dim of q be different from dim of RotaryEmbedding
#4
HiSultryMan
opened
1 year ago
2
Length Extrapolatable Rotary Embeddings
#3
hugofloresgarcia
opened
1 year ago
2
Custom position offset when rotating queries or keys
#2
krasserm
closed
1 year ago
1
about axial rotary embeddings
#1
raindrop313
opened
2 years ago
0