Code

Add tests for examples#149

Code

Merged

mgoin merged 32 commits intovllm-project:mainvllm-project/llm-compressor:mainfrom dbarbuzzi:add-examples-testsdbarbuzzi/llm-compressor:add-examples-testsCopy head branch name to clipboardSep 27, 2024

Commits

Commits on Sep 20, 2024

• Add tests for examples

  dbarbuzzi

  committed
  906018d

  Copy full SHA for 906018d

• Ignore examples tests by default

  dbarbuzzi

  committed
  0ef1325

  Copy full SHA for 0ef1325

• Trailing comma

  dbarbuzzi

  committed
  0fbb4ed

  Copy full SHA for 0fbb4ed

• Add test for "quantizing_moe_fp8" example

  dbarbuzzi

  committed
  817a311

  Copy full SHA for 817a311

• Update "quantizing_moe" example tests

  dbarbuzzi

  committed
  3d15824

  Copy full SHA for 3d15824

• Add test for "compressed_inference" example folder

  dbarbuzzi

  committed
  81b7c95

  Copy full SHA for 81b7c95

• Remove unused import

  dbarbuzzi

  committed
  35a6903

  Copy full SHA for 35a6903

• Add new dependency

  dbarbuzzi

  committed
  da8eaaa

  Copy full SHA for da8eaaa

• Different approach for "flash_attn"

  dbarbuzzi

  committed
  95dea5a

  Copy full SHA for 95dea5a

• Add comment about flash_attn requirement

  dbarbuzzi

  committed
  92be6d0

  Copy full SHA for 92be6d0

• Test additional quantizing_moe example script

  dbarbuzzi

  committed
  0e99d6a

  Copy full SHA for 0e99d6a

• Add decorator to skip based on available VRAM

  dbarbuzzi

  committed
  e75399c

  Copy full SHA for e75399c

• Limit GPU usage in 'cpu_offloading' example

  dbarbuzzi

  committed
  95e950a

  Copy full SHA for 95e950a

• Add optional pytest-xdist parallelization

  dbarbuzzi

  committed
  534a613

  Copy full SHA for 534a613

• Reduce persistent /tmp usage

  dbarbuzzi

  committed
  4880d4b

  Copy full SHA for 4880d4b

• Fix parametrization in big_models

  dbarbuzzi

  committed
  f8a97ce

  Copy full SHA for f8a97ce

• Add pytest mark for GPU count requirement

  dbarbuzzi

  committed
  34a5f34

  Copy full SHA for 34a5f34

• Add 'multi_gpu' pytest marker

  dbarbuzzi

  committed
  7ddbb49

  Copy full SHA for 7ddbb49

• Skip 'deepseek_moe_w4a16.py' by default

  dbarbuzzi

  committed
  af5eeaf

  Copy full SHA for af5eeaf

• Fix skip mark

  dbarbuzzi

  committed
  a2b961e

  Copy full SHA for a2b961e

Commits on Sep 21, 2024

• Mark 'ex_trl_distillation.py' as multi_gpu

  dbarbuzzi

  committed
  1f4ec50

  Copy full SHA for 1f4ec50

Commits on Sep 23, 2024

• Abstract command copy/run to helper functions

  dbarbuzzi

  committed
  28abda4

  Copy full SHA for 28abda4

• Update for MoE examples PR vllm-project#192

  dbarbuzzi

  committed
  5d27928

  Copy full SHA for 5d27928

• Merge remote-tracking branch 'upstream/main' into add-examples-tests

  dbarbuzzi

  committed
  789bd5a

  Copy full SHA for 789bd5a

• Reduce test marker/decorator redundancy

  dbarbuzzi

  committed
  e5cacca

  Copy full SHA for e5cacca

• style fixes

  dbarbuzzi

  committed
  a50ce6c

  Copy full SHA for a50ce6c

• Rip out unused run parallelization

  dbarbuzzi

  committed
  42121f8

  Copy full SHA for 42121f8

• Exclude 'deepseek_moe_w8a8_fp8' from multi-GPU

  dbarbuzzi

  committed
  465ac92

  Copy full SHA for 465ac92

Commits on Sep 25, 2024

• Merge branch 'main' into add-examples-tests

  dbarbuzzi

  authored
  9ba702f

  Copy full SHA for 9ba702f

• Use variable for repeated string literal

  dbarbuzzi

  committed
  7283d80

  Copy full SHA for 7283d80

Commits on Sep 27, 2024

• Use requires_gpu_count over requires_gpu

  dbarbuzzi

  committed
  da2f10f

  Copy full SHA for da2f10f

• Merge branch 'main' into add-examples-tests

  dbarbuzzi

  authored
  a21ce33

  Copy full SHA for a21ce33