MrYxJ/calculate-flops.pytorch

Issues

• The FLOPS of scalar dot products in Attention has not been considered？

  #46 opened 4 months ago by XIronMan0220
  1

• Can this support StableDiffusion？

  #23 opened a year ago by sunzhe09
  1

• Library `diffusers` is not supported yet

  #12 opened a year ago by XianweiC
  3

• RuntimeError: A view was created in no_grad mode and is being modified inplace with grad mode enabled. Given that this use case is ambiguous and error-prone, it is forbidden. You can clarify your code by moving both the view and the inplace either both inside the no_grad block (if you don't want the inplace to be tracked) or both outside (if you want the inplace to be tracked).

  #44 opened 4 months ago by JokeQ666
  1

• Why the unit of fwd FLOPs is FLOPS?

  #16 opened a year ago by Fibonacci43
  1

• Compute the FLOPs for each transformer layer of an LLM

  #43 opened 5 months ago by jeremyzhangsq
  0

• Ran the example code but got wrong results

  #42 opened 6 months ago by nameongithub
  2

• AttributeError: 'list' object has no attribute 'shape' in function _einsum_flops_compute()

  #40 opened 6 months ago by Necolizer
  1

• Why are `torch.mm` and `torch.bmm` deleted in pytorch_ops.py

  #36 opened 8 months ago by wzk1015
  2

• 你好，比如llama这些LLM模型，我看统计的结果好像没有包括attention矩阵的Q@K.T那部分计算量。

  #26 opened 10 months ago by ywh187
  5

• gated models do not work

  #31 opened 9 months ago by mohbay
  1

• How to test models deployed on multiple GPUs？

  #39 opened 7 months ago by May2064
  0

• packaging.version.InvalidVersion: Invalid version: '0.10.1,<0.11'

  #38 opened 7 months ago by ymmm-4
  0

• Flops for GNN

  #37 opened 7 months ago by ritamartinspereira
  1

• Only nn.modules are supported for FLOP analysis?

  #35 opened 8 months ago by avisinghal6
  0

• Add support of non-linear operations

  #34 opened 9 months ago by Ther-nullptr
  0

• MACs for ConvTranspose2d

  #33 opened 9 months ago by qiuzidian
  0

• How to calculate the FLOPs when the model has some freezed variables, or using LoRA?

  #30 opened 10 months ago by Ledzy
  0

• What is the difference between the 'forward' mode and the 'generate' mode?

  #20 opened a year ago by AboveParadise
  1

• On the calculation of FLOPs for SAM/MobileSAM

  #29 opened 10 months ago by BeliverK
  0

• 评价Gemma的时候参数量为0

  #28 opened 10 months ago by JackeyWang777
  1

• Why is the proportion of FLOPs in each submodule always half of the proportion of MACS

  #7 opened a year ago by wulinunu
  1

• 关于百分比的问题

  #27 opened 10 months ago by lzd-1230
  2

• Please make a CITATION file

  #19 opened 10 months ago by Livioni
  1

• Cannot calculate EfficientNet

  #11 opened 10 months ago by QingyuLiu
  1

• Library `gpt-neox` is not supported yet

  #18 opened 10 months ago by melisa-writer
  1

• flops percent wrong

  #24 opened 10 months ago by Lix1993
  1

• Is it considering output length of transformer models?

  #25 opened 10 months ago by liut89
  1

• [Bug] Upsample flops not being assigned after calculation

  #22 opened a year ago by Ridhamz-nd
  0

• How to calculate model with more than one input?

  #17 opened a year ago by yuanzhi-zhu
  0

• error with the example code

  #14 opened a year ago by tranlm
  1

• 访存量的计算

  #10 opened a year ago by Amanda-Barbara
  2

• about 3D object detection

  #13 opened a year ago by ouyangziyao
  0

• This project can be used on large visual language models?

  #5 opened a year ago by jiabao-wang
  4

• ImportError: cannot import name 'is_timm_available' from 'accelerate.utils'

  #6 opened a year ago by jiabao-wang
  2

• why I can not find transformers.modeling_outputs,Can you tell me your vision of transformers

  #9 opened a year ago by kingofengineer
  1

• why seq_len of llm is 128?

  #8 opened a year ago by LinglingGreat
  1

• 大模型微调

  #1 opened 2 years ago by Pobby321
  1

• ImportError at 'from calflops import calculate_flops': cannot import name 'is_timm_available' from 'accelerate.utils'

  #2 opened 2 years ago by biplob1ly
  3

• 给的示例无法运行

  #3 opened 2 years ago by Arcmoon-Hu
  1