Code

Quantization Compressor Support #2260

Code

Merged

Satrat merged 70 commits intomainneuralmagic/sparseml:mainfrom sa/compressorsneuralmagic/sparseml:sa/compressorsCopy head branch name to clipboardMay 9, 2024

+327-98

Commits

Commits on Apr 8, 2024

• initial commit

  dbogunowicz

  committed
  097bd79

• update setup.py

  dbogunowicz

  committed
  76970e3

• Update setup.py

  dbogunowicz

  authored
  bbf4b39

• fix setup.py

  dbogunowicz

  committed
  a272a30

Commits on Apr 10, 2024

• move all config to sparsetensors

  

  Sara Adkins

  committed
  c0d3ead

• Merge branch 'main' into feature/damian/sparsetensors

  

  Sara Adkins

  committed
  b3f7ff3

• cleanup class name and comments

  

  Sara Adkins

  committed
  a75f8da

Commits on Apr 16, 2024

• Merge branch 'main' into feature/damian/sparsetensors

  

  Sara Adkins

  authored
  c5b897e

• initial implementation untested

  

  Sara Adkins

  committed
  2c72ab1

• fixing issues

  

  Sara Adkins

  committed
  9174c1d

Commits on Apr 17, 2024

• add test script

  

  Sara Adkins

  committed
  aa17e77

• update perplexity test

  

  Sara Adkins

  committed
  f1f114c

Commits on Apr 18, 2024

• refactor to compressed-tensors

  dbogunowicz

  committed
  bbbdcb9

• Merge branch 'main' into feature/damian/sparsetensors

  dbogunowicz

  authored
  5d9c7dd

• rename sparsetensors

  

  Sara Adkins

  committed
  7a9f9e5

• update setup

  

  Sara Adkins

  committed
  fa43088

Commits on Apr 19, 2024

• Sa/model reload (#2250)

  Show description for 63266d8
  

  Sara Adkins

  authored
  63266d8

Commits on Apr 22, 2024

• Merge branch 'main' into sa/quant_mod_refactor

  

  Sara Adkins

  authored
  b0f0fc9

• Merge branch 'main' into feature/damian/sparsetensors

  

  Sara Adkins

  authored
  dfa41fb

• Merge branch 'feature/damian/sparsetensors' into sa/quant_mod_refactor

  

  Sara Adkins

  committed
  4af4852

• cleanup

  

  Sara Adkins

  committed
  55976c5

• refactor tests

  

  Sara Adkins

  committed
  38f4f77

• only run oneshot once

  

  Sara Adkins

  committed
  6574874

Commits on Apr 23, 2024

• all tests passing

  dbogunowicz

  committed
  7f5babf

• remove unused config

  dbogunowicz

  committed
  c0d6cb9

• reset models on each parameterize

  

  Sara Adkins

  committed
  a59e2af

• Merge branch 'feature/damian/sparsetensors' into sa/quant_mod_refactor

  

  Sara Adkins

  committed
  cba7c27

• style

  

  Sara Adkins

  committed
  2a6b0f2

Commits on Apr 24, 2024

• Merge branch 'main' into feature/damian/sparsetensors

  dbogunowicz

  authored
  1e7ee94

• bring back SparsityConfigMetadata

  dbogunowicz

  committed
  a4e0575

• Merge branch 'feature/damian/sparsetensors' of github.com:neuralmagic/sparseml into feature/damian/sparsetensors

  dbogunowicz

  committed
  06d4554

• Merge remote-tracking branch 'origin/feature/damian/sparsetensors' into sa/quant_mod_refactor

  dbogunowicz

  committed
  644da53

• Update setup.py

  Show description for 8ac18e7
  

  dbogunowicz

  and

  rahul-tuli

  authored
  8ac18e7

Commits on Apr 25, 2024

• add more comparisons, tighten threshold

  

  Sara Adkins

  committed
  de78247

• use wikitext for perplexity

  

  Sara Adkins

  committed
  4041f2e

• Merge branch 'main' into feature/damian/sparsetensors

  dbogunowicz

  authored
  f5adc4e

• update setup

  dbogunowicz

  committed
  c220772

• fix import problem

  dbogunowicz

  committed
  2fe554e

• fix clearml test

  dbogunowicz

  committed
  4e0413e

• compressed-tensors are transformers dep

  dbogunowicz

  committed
  a98a193

• Merge branch 'feature/damian/sparsetensors' into sa/quant_mod_refactor

  

  Sara Adkins

  committed
  b9b684c

• address PR comments

  

  Sara Adkins

  committed
  f4362cf

Commits on Apr 26, 2024

• can't repeat freeze

  

  Sara Adkins

  committed
  ca91c4f

• UX pr comments

  

  Sara Adkins

  committed
  c894305

Commits on Apr 29, 2024

• initial commit

  

  Sara Adkins

  committed
  604c4ef

• style

  

  Sara Adkins

  committed
  82d3dd8

Commits on Apr 30, 2024

• skipping unit tests

  

  Sara Adkins

  committed
  b650a8c

• tests for quantization

  

  Sara Adkins

  committed
  6a12295

• reloading unit tests

  

  Sara Adkins

  committed
  c1e0379

• backwards compat

  

  Sara Adkins

  committed
  ba397fc

• test updates

  

  Sara Adkins

  committed
  0a0ef06

• update format

  

  Sara Adkins

  committed
  b03d138

Commits on May 1, 2024

• fix inferring

  

  Sara Adkins

  committed
  2931d8e

• Merge branch 'main' into sa/quant_mod_refactor

  

  Sara Adkins

  authored
  1c3b31b

• quality

  

  Sara Adkins

  committed
  90795bd

• Merge branch 'sa/quant_mod_refactor' into sa/compressors

  

  Sara Adkins

  authored
  c287c05

• shape consistency

  horheynm

  committed
  bf7d0f6

• Merge branch 'sa/quant_mod_refactor' of github.com:neuralmagic/sparseml into sa/quant_mod_refactor

  horheynm

  committed
  579d201

Commits on May 2, 2024

• address PR comments

  

  Sara Adkins

  committed
  2432cf4

Commits on May 3, 2024

• PR comments

  

  Sara Adkins

  committed
  24437c7

• Merge branch 'sa/quant_mod_refactor' into sa/compressors

  

  Sara Adkins

  committed
  399087f

• fixing some things

  

  Sara Adkins

  committed
  e8bc021

Commits on May 7, 2024

• Merge branch 'main' into sa/compressors

  

  Sara Adkins

  committed
  3ca0298

• style

  

  Sara Adkins

  committed
  061de67

Commits on May 8, 2024

• Merge branch 'main' into sa/compressors

  

  Sara Adkins

  authored
  633d5a5

• pull from cp main

  

  Sara Adkins

  committed
  6e0f1bc

• postmerge too

  

  Sara Adkins

  committed
  3ff4dc8

• Merge branch 'main' into sa/compressors

  

  Sara Adkins

  committed
  6f4379c

• export needs it too

  

  Sara Adkins

  committed
  29a2186

Commits on May 9, 2024

• Update src/sparseml/modifiers/obcq/utils/sgpt_wrapper.py

  Show description for e93257f
  

  Sara Adkins

  and

  rahul-tuli

  authored
  e93257f