Code

merge transformers into main#8

Code

Merged

plutonium-239 merged 39 commits intomainplutonium-239/memsave_torch:mainfrom transformersplutonium-239/memsave_torch:transformersCopy head branch name to clipboardAug 22, 2024

+572-95

Commits

Commits on Aug 22, 2024

• add llm trial code

  Show description for 3893c6c
  plutonium-239

  committed
  3893c6c

• transformers.Conv1D

  Show description for f970dfe
  plutonium-239

  committed
  f970dfe

• fix inputs for transformers

  Show description for 9a9801d
  plutonium-239

  committed
  9a9801d

• minor

  Show description for 5a34b84
  plutonium-239

  committed
  5a34b84

• directly pass input embeddings

  Show description for e3e5476
  plutonium-239

  committed
  e3e5476

• format

  Show description for 281866f
  plutonium-239

  committed
  281866f

• replace transformers.Conv1D with MemSaveLinear

  Show description for f0aa31f
  plutonium-239

  committed
  f0aa31f

• replace transformers.Conv1D with MemSaveLinear

  Show description for 63a4f7c
  plutonium-239

  committed
  63a4f7c

• add vit, torch vanilla transformer to experiment models

  Show description for 80d4270
  plutonium-239

  committed
  80d4270

• change estimate - add transformer as architecture

  Show description for e5f728d
  plutonium-239

  committed
  e5f728d

• improve transformers config handling

  Show description for c6f8c8e
  plutonium-239

  committed
  c6f8c8e

• minor

  Show description for c471abc
  plutonium-239

  committed
  c471abc

• various fixes + formatting

  Show description for b470f16
  plutonium-239

  committed
  b470f16

• add many hf models

  Show description for 0641fe9
  plutonium-239

  committed
  0641fe9

• demo add transformers

  Show description for d55c3db
  plutonium-239

  committed
  d55c3db

• change from lambdas inside loop to functool.partials

  Show description for f30f8e2
  plutonium-239

  committed
  f30f8e2

• fix for encoder-decoder models + avoid loading transformers in memsave_torch.nn

  Show description for b92ee08
  plutonium-239

  committed
  b92ee08

• format + minor

  Show description for 30f8ed6
  plutonium-239

  committed
  30f8ed6

• fix "Do not use mutable data structures for argument defaults"

  Show description for 420c027
  plutonium-239

  committed
  420c027

• fix

  Show description for d519e34
  plutonium-239

  committed
  d519e34

• minor

  Show description for f690d13
  plutonium-239

  committed
  f690d13

• lm_head is non trainable

  Show description for d7622d1
  plutonium-239

  committed
  d7622d1

• misc fixes, code for making latex table

  Show description for 02ed7fa
  plutonium-239

  committed
  02ed7fa

• qol

  Show description for 13897da
  plutonium-239

  committed
  13897da

• fix for different model dtypes

  Show description for 9b8e7f5
  plutonium-239

  committed
  9b8e7f5

• adjust batch size for mistral-7b

  Show description for b9d31cc
  plutonium-239

  committed
  b9d31cc

• minor: correct cache_kws

  Show description for 8c281f6
  plutonium-239

  committed
  8c281f6

• minor printing changes

  Show description for 69d2bd1
  plutonium-239

  committed
  69d2bd1

• update optional deps

  Show description for d12a2d9
  plutonium-239

  committed
  d12a2d9

• minor

  Show description for f976b61
  plutonium-239

  committed
  f976b61

• new implementation of dropout

  Show description for b057db2
  plutonium-239

  committed
  b057db2

• adding llama3 and phi3

  Show description for f256f93
  plutonium-239

  committed
  f256f93

• refactor cases to refer to a central location

  Show description for 7cde54b
  plutonium-239

  committed
  7cde54b

• format + prep for merge

  Show description for 51537c6
  plutonium-239

  committed
  51537c6

• minor fixes

  Show description for 9ae8507
  plutonium-239

  committed
  9ae8507

• add VLM architecture in estimate, refactor cases to refer to a central location

  Show description for b36baa6
  plutonium-239

  committed
  b36baa6

• format + prep for merge

  Show description for 3d5ec71
  plutonium-239

  committed
  3d5ec71

• Merge branch 'transformers' into vlms

  plutonium-239

  authored
  b78d341

• Merge pull request #7 from plutonium-239/vlms

  Show description for 4d489fa
  plutonium-239

  authored
  4d489fa