facebookresearch xformers issues

facebookresearch / xformers

Hackable and optimized Transformers building blocks, supporting a composable construction.

https://facebookresearch.github.io/xformers/

Other

8.67k stars 618 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

visual studio 2022,cuda 12.4 ,but it still can't build the wheel, and return this

#1161 zslefour opened 11 hours ago
1
[refactor] Generalized SwiGLU python code

#1160 warpuv opened 20 hours ago
0
Incompatibility Between xformers FA3 Torch Custom op Wrapper and recent `flashattn_hopper_cuda`

#1159 ohwi opened 20 hours ago
1
Generalize SwiGLU related python code

#1158 warpuv opened 21 hours ago
0
[FA3] Link to cuda library to fix the FA3 extension build

#1157 xuzhao9 closed 1 day ago
2
Triton Module Not Available: xformers Optimizations Failing on Windows with CUDA 12.4

#1156 BasimBashir closed 2 days ago
1
need to release xformers-0.0.28.post3.whl with manylinux2014_x86_64 os version.

#1155 controlRun opened 3 days ago
1
ValueError: not enough values to unpack (expected 2, got 1)

#1154 algorithmconquer closed 1 week ago
1
RuntimeError: Found an unsupported argument type c10::SymInt in the JIT tracer. File a bug report.

#1153 J4Q8 closed 2 days ago
2
[fix] Fix activation checkpointing of SwiGLU when AMP is enabled.

#1152 warpuv closed 1 week ago
1
Activation checkpointing on fused SwiGLU is not working when AMP is enabled.

#1151 warpuv closed 1 week ago
0
TORCH_CUDA_ARCH_LIST: 8.0+PTX not being detected properly

#1150 HighSec-org closed 1 week ago
1
fix: [N, 1] pattern should take N to create 1d causal mask

#1149 davidqqq opened 1 week ago
2
Can't install xformers

#1148 GUEST-1001 closed 1 week ago
1
Customization of BlockDiagonalMask or LowerTriangularMask

#1147 kc334 opened 1 week ago
0
Can I use xformers with torch-2.4.1?

#1146 LukeLIN-web closed 1 week ago
1
build without flash attn?

#1145 sipie800 opened 2 weeks ago
1
RuntimeError: CUDA error: CUBLAS_STATUS_NOT_INITIALIZED when calling `cublasCreate(handle)`

#1144 antony-frolov closed 2 weeks ago
1
# ❓ Questions and Help NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs:

#1143 Was-11-wendy closed 2 weeks ago
2
How to install to aarch64

#1142 gongchangsui opened 2 weeks ago
1
[refactor] Generalization of dual_gemm_silu_identity_mul

#1141 warpuv closed 2 weeks ago
3
Generalization of dual_gemm_silu_identity_mul to use custom activation function, not only SiLU

#1140 warpuv closed 2 weeks ago
0
Cann't install xformers==0.0.28.post2

#1139 Wiselnn570 closed 3 weeks ago
1
vllm 0.6.3 createLLM error TypeError: autotune() got an unexpected keyword argument 'use_cuda_graph' on windows

#1138 xiezhipeng-git opened 3 weeks ago
4
torch2.5.1 support

#1137 luohao123 closed 3 weeks ago
4
Update xFormers MinGPT notebook reference

#1136 emmanuel-ferdman closed 1 week ago
1
incorrect causal mask in global attention

#1135 davidqqq opened 1 month ago
1
timm.models.layers is deprecated/resume_download is deprecated

#1134 zd391 closed 4 weeks ago
1
[fix] Add back BlockDiagonalMask import

#1133 tanvitiwari-meta closed 3 weeks ago
2
Cant use powershell command to build for Nightly PyTorch on Win - filename too long

#1132 Mescalamba opened 1 month ago
5
Is there an efficient way to use memory_efficient_attention with a causal mask that has a small rectangle of zeros?

#1131 arilato opened 1 month ago
1
AMD64 version of Xformers

#1130 pandayummy closed 1 month ago
3
AttributeError: module 'xformers.ops' has no attribute 'AttentionOpDispatch'

#1129 LarsDoorenbos closed 1 month ago
1
xformers 0.0.20 memoery efficient attent cutalss backward non deterministic

#1128 Bavesh-B opened 1 month ago
0
[fix] Fix the activation checkpointing when using SwiGLUPackedFusedOp

#1127 warpuv closed 3 weeks ago
8
Activation checkpointing is not working on SwiGLU

#1126 warpuv closed 3 weeks ago
0
xformers.sparse.utils._coo_to_csr is incorrect when n > m

#1125 francois-rozet opened 1 month ago
0
Incorrect attention output with SparseCS mask

#1124 francois-rozet opened 1 month ago
1
Add support for CUDNN attention via CUDNN_FRONTEND Python API?

#1123 Skylion007 opened 1 month ago
0
[Bug] Unexpected behavior of `memory_efficient_attention` with `BlockDiagonalMask`

#1122 xiangxu-google opened 1 month ago
1
Are there any version for Torch 2.5 including dev? like 0.0.29.dev921?

#1121 FurkanGozukara closed 1 week ago
6
Does memory efficient attention cutlass kernel support various seq len inputs for q/k/v + tensor bias?

#1120 ShijunK opened 1 month ago
0
Documentation update FMHA __init__.py

#1119 lhallee opened 1 month ago
2
Why xformers 0.0.28.post1 doesn't have pre-compiled wheel for Windows?

#1118 FurkanGozukara opened 1 month ago
7
I tried upgrade visual studio,cuda 12.4 and 12.6,but it still can't build the wheel,and return this

#1117 neutronslime opened 1 month ago
5
I cannot run xformers on an AMD Radeon RX 6600 with torch 2.4.1

#1116 renich closed 1 month ago
2
scaled_dot_product_attention output is different from memory_efficient_attention

#1114 aenoca opened 1 month ago
1
Enabling softcap option

#1113 SpyrosMouselinos closed 1 month ago
1
CUTLASS Fused multi head attention

#1112 yoon5862 opened 1 month ago
2
what version about torch in FLUX bf16？

#1111 Zhuangvictor0 closed 1 month ago
1