Code

[NPU] Example & Quickstart updates#12650

Code

Merged

Oscilloscope98 merged 29 commits intointel:mainintel/ipex-llm:mainfrom Oscilloscope98:npu-example-quickstart-updatesOscilloscope98/ipex-llm:npu-example-quickstart-updatesCopy head branch name to clipboardJan 7, 2025

+314-495

Commits

Commits on Jan 3, 2025

• Remove model with optimize_model=False in NPU verified models tables, and remove related example

  Oscilloscope98

  committed
  633a52d

  Copy full SHA for 633a52d

• Remove experimental in run optimized model section title

  Oscilloscope98

  committed
  6a6396b

  Copy full SHA for 6a6396b

• Unify model table order & example cmd

  Oscilloscope98

  committed
  9cd894d

  Copy full SHA for 9cd894d

• Move embedding example to separate folder & update quickstart example link

  Oscilloscope98

  committed
  6b0acfe

  Copy full SHA for 6b0acfe

• Add Quickstart reference in main NPU readme

  Oscilloscope98

  committed
  9a613fd

  Copy full SHA for 9a613fd

• Small fix

  Oscilloscope98

  committed
  95b668e

  Copy full SHA for 95b668e

Commits on Jan 6, 2025

• Small fix

  Oscilloscope98

  committed
  ed49fb5

  Copy full SHA for ed49fb5

• Move save/load examples under NPU/HF-Transformers-AutoModels

  Oscilloscope98

  committed
  885a336

  Copy full SHA for 885a336

• Add low-bit and polish arguments for LLM Python examples

  Oscilloscope98

  committed
  7421361

  Copy full SHA for 7421361

• Small fix

  Oscilloscope98

  committed
  50d30c6

  Copy full SHA for 50d30c6

• Add low-bit and polish arguments for Multi-Model examples

  Oscilloscope98

  committed
  a285861

  Copy full SHA for a285861

• Polish argument for Embedding models

  Oscilloscope98

  committed
  e1f2339

  Copy full SHA for e1f2339

• Polish argument for LLM CPP examples

  Oscilloscope98

  committed
  9d7d8b3

  Copy full SHA for 9d7d8b3

• Add low-bit and polish argument for Save-Load examples

  Oscilloscope98

  committed
  b9e5831

  Copy full SHA for b9e5831

• Add accuracy tuning tips for examples

  Oscilloscope98

  committed
  0e82867

  Copy full SHA for 0e82867

• Update NPU qucikstart accuracy tuning with low-bit optimizations

  Oscilloscope98

  committed
  de61245

  Copy full SHA for de61245

• Add save/load section to qucikstart

  Oscilloscope98

  committed
  890cdd9

  Copy full SHA for 890cdd9

• Update CPP example sample output to EN

  Oscilloscope98

  committed
  e65934c

  Copy full SHA for e65934c

• Add installation regarding cmake for CPP examples

  Oscilloscope98

  committed
  a8e5426

  Copy full SHA for a8e5426

• Small fix

  Oscilloscope98

  committed
  a7799ed

  Copy full SHA for a7799ed

• Small fix

  Oscilloscope98

  committed
  6009534

  Copy full SHA for 6009534

• Small fix

  Oscilloscope98

  committed
  a161de3

  Copy full SHA for a161de3

• Small fix

  Oscilloscope98

  committed
  809605f

  Copy full SHA for 809605f

• Small fix

  Oscilloscope98

  committed
  1a61689

  Copy full SHA for 1a61689

• Small fix

  Oscilloscope98

  committed
  37986b4

  Copy full SHA for 37986b4

Commits on Jan 7, 2025

• Unify max prompt length to 512

  Oscilloscope98

  committed
  1a729ce

  Copy full SHA for 1a729ce

• Change recommended low-bit for Qwen2.5-3B-Instruct to asym_int4

  Oscilloscope98

  committed
  281b5af

  Copy full SHA for 281b5af

• Update based on comments

  Oscilloscope98

  committed
  8aead4d

  Copy full SHA for 8aead4d

• Small fix

  Oscilloscope98

  committed
  e9bc6c3

  Copy full SHA for e9bc6c3