microsoft/TransformerCompression

Issues

• Convert Layernorm to RMSnorm

  #110 opened 9 months ago by lihuang258
  1

• Issues with LLAMA 3-8B-Instruct model

  #177 opened 3 months ago by madhusrivatsav
  0

• NaN weights after rotating and slicing

  #176 opened 3 months ago by sriyachakravarthy
  0

• How to export entire model?

  #114 opened 9 months ago by YuanzeSun
  9

• need help

  #175 opened 4 months ago by jiajunsun68
  0

• NotImplementedError: xx is neither a Hugging Face model nor a supported local model.

  #138 opened 7 months ago by qxpBlog
  2

• Quarot: DeepSeek-V2 Support

  #174 opened 4 months ago by RanchiZhao
  0

• How to evaluate the sliced model

  #173 opened 5 months ago by 1250826219
  0

• What is the number of parameters afrer slicing?

  #165 opened 5 months ago by yaya-sy
  2

• `run_benchmark.py` runs error when using `--distribute-model`

  #122 opened 8 months ago by zhaoyang-star
  1

• How to finetune with multi-gpus under data parallel setting?

  #167 opened 6 months ago by kriskrisliu
  0

• can't reproduce result

  #127 opened 8 months ago by MrGGLS
  8

• How to load the tuned slicemodel

  #140 opened 6 months ago by qxpBlog
  1

• can't install slicegpt using "pip install -e." on CPU platform

  #126 opened 6 months ago by JCDemon
  2

• Command R and R+ support

  #133 opened 7 months ago by Steel-skull
  1

• A problem about the PPL value after sliced model fine-tuning

  #141 opened 6 months ago by qxpBlog
  2

• PHI包导入问题

  #164 opened 6 months ago by 1250826219
  2

• model inference

  #160 opened 6 months ago by ChrisXULC
  0

• error when Fine-tuning a sliced model llama 3

  #155 opened 6 months ago by ChrisXULC
  0

• How to run with llama-1?

  #128 opened 8 months ago by liuxiaozhu01
  4

• Question of `RMSNorm`'s `forward` function

  #131 opened 7 months ago by zhaoyang-star
  2

• Layer fusion with Llama

  #129 opened 8 months ago by kiucho
  2

• why the generation speed of the pruned model by SliceGPT is slower than the original model？

  #115 opened 8 months ago by joyce0105-ops
  11

• question about the cal_dataset

  #113 opened 8 months ago by tu2022
  1

• Proof of Equation 2

  #121 opened 8 months ago by kiucho
  1

• Recovery the drop in accuracy

  #119 opened 8 months ago by zhaoyang-star
  3

• run_finetuning.py issue on non-sliced private models

  #123 opened 8 months ago by taaviv0
  2

• failures when evaluating the model

  #117 opened 8 months ago by YuanzeSun
  2

• Is there any inference demo for sliced model?

  #116 opened 8 months ago by zhaoyang-star
  0

• Question about perplexity results shown on the paper

  #118 opened 8 months ago by moonlightian
  1

• Reproducing the speed up results in table 2

  #112 opened 9 months ago by Ahmed-Roushdy
  1

• Finetuning fails with seg fault unless CUDA_VISIBLE_DEVICES is set

  #106 opened 9 months ago by nailimixaM
  0

• c4 dataset download fails

  #56 opened 10 months ago by nailimixaM
  1

• loading models is painful and not HF compatible

  #73 opened 10 months ago by jameshensman
  0

• parameter count is innacurate

  #72 opened 9 months ago by jameshensman
  1

• Replace <pad> with <eos> for Llama2 and Phi2

  #60 opened 10 months ago by nailimixaM
  0

• Mistral Support

  #81 opened 10 months ago by fakerybakery
  5

• Update README with links to arxiv paper (when uploaded)

  #64 opened 10 months ago by nailimixaM
  0

• Scrub keys from commit history

  #76 opened 10 months ago by nailimixaM
  0

• --distribute fails for calculating perplexity

  #66 opened 10 months ago by nailimixaM
  0

• Slicing with batches of varying sizes and seqlens fails

  #48 opened 10 months ago by nailimixaM
  0

• hf token should be retrivable from env

  #58 opened 10 months ago by jameshensman
  1

• llamadecoder in transformers 4.36

  #49 opened 10 months ago by jameshensman
  1

• Common command-line args

  #52 opened a year ago by nailimixaM
  0

• Action required: migrate or opt-out of migration to GitHub inside Microsoft

  #30 opened a year ago by microsoft-github-policy-service
  6