Skip to content

Phi4 Mini Instruct QCOM recipie#475

Draft
rM-planet wants to merge 16 commits into
microsoft:mainfrom
CodeLinaro:mlperf_qnn_phi4
Draft

Phi4 Mini Instruct QCOM recipie#475
rM-planet wants to merge 16 commits into
microsoft:mainfrom
CodeLinaro:mlperf_qnn_phi4

Conversation

@rM-planet

Copy link
Copy Markdown
Contributor

No description provided.

Ronak Mahawar and others added 3 commits June 4, 2026 12:21
Trim genai_lib, llm_utils, quantizer_utils, and utilities to only the
files actually imported by phi4.py. Removes 67 files that belonged to
other model families (LVM, Llama, Mistral, Qwen, Baichuan, etc.) or
were unused infrastructure. Adds utilities/nsptargets.py copied from
the Llama recipe which the script requires.
qti-kromero and others added 13 commits June 5, 2026 20:42
…ve engine_config_overrides

Switch input_model from QairtPreparedModel to HFModel with trust_remote_code=true
to bypass the rope_scaling validation error in transformers 4.46. Remove
engine_config_overrides from qe pass as this SDK version does not support
EngineConfig in LLMContainer.export().
…icrosoft#434)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: hualxie <hualxie@microsoft.com>
Co-authored-by: hualxie <hualxie@microsoft.com>
Co-authored-by: hualxie <hualxie@microsoft.com>
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Akshay Sonawane <111780983+apsonawane@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants