update qnn gpu

[xieofxie]

model \ ep	8D
phi	Tokens per second: 10.61
llama3.2-1b	Tokens per second: 29.39
llama3.1-8b	Tokens per second: 6.85
qwen2.5-1.5b	Tokens per second: 25.33

[Copilot]

🟡 Changes recommended

Moderate findings remain in WebGPU comparison ignore patterns for nested and removed s fields.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

Updates AITK QNN GPU and WebGPU recipes with Olive support, revised conversion passes, and model-specific comparison settings.

Changes:

• Adds Olive runtime dependencies and feature flags.
• Updates QNN GPU conversion and context settings.
• Adds WebGPU references and improves sanitizer diagnostics.

File summaries

File	Summary
Qwen-Qwen2.5-1.5B-Instruct/aitk/qwen2_5_webgpu.json.config	Moderate (2 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
Qwen-Qwen2.5-1.5B-Instruct/aitk/qwen2_5_qnn_gpu_config.json.config	Adds Olive runtime feature requirements.
Qwen-Qwen2.5-1.5B-Instruct/aitk/qwen2_5_qnn_gpu_config.json	Updates QNN conversion passes.
Qwen-Qwen2.5-1.5B-Instruct/aitk/info.yml	Moderate (3 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
microsoft-Phi-3.5-mini-instruct/aitk/phi3_5_webgpu.json.config	Moderate (2 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
microsoft-Phi-3.5-mini-instruct/aitk/phi3_5_qnn_gpu_config.json.config	Adds Olive runtime feature requirements.
microsoft-Phi-3.5-mini-instruct/aitk/phi3_5_qnn_gpu_config.json	Updates QNN conversion passes.
microsoft-Phi-3.5-mini-instruct/aitk/info.yml	Moderate (3 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
meta-llama-Llama-3.2-1B-Instruct/aitk/llama3_2_webgpu.json.config	Moderate (2 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
meta-llama-Llama-3.2-1B-Instruct/aitk/llama3_2_qnn_gpu_config.json.config	Adds Olive runtime feature requirements.
meta-llama-Llama-3.2-1B-Instruct/aitk/llama3_2_qnn_gpu_config.json	Updates QNN conversion passes.
meta-llama-Llama-3.2-1B-Instruct/aitk/info.yml	Moderate (3 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
meta-llama-Llama-3.1-8B-Instruct/aitk/llama3_1_webgpu.json.config	Moderate (2 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
meta-llama-Llama-3.1-8B-Instruct/aitk/llama3_1_qnn_gpu.json.config	Adds Olive runtime feature requirements.
meta-llama-Llama-3.1-8B-Instruct/aitk/llama3_1_qnn_gpu.json	Updates QNN conversion passes.
meta-llama-Llama-3.1-8B-Instruct/aitk/info.yml	Moderate (3 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
deepseek-ai-DeepSeek-R1-Distill-Qwen-1.5B/aitk/info.yml	Moderate (3 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
deepseek-ai-DeepSeek-R1-Distill-Qwen-1.5B/aitk/deepseek_webgpu.json.config	Moderate (2 votes): The s ignore pattern does not match nested or removed paths; use a subtree prefix and ignoreRemoved.
.aitk/scripts/sanitize/model_parameter.py	Improves Olive reference path diagnostics.
.aitk/requirements/General/CUDA_py3.12.9-olive.txt	Adds pinned Olive dependencies.
.aitk/configs/checks.json	Updates validation counts.

Review details

• Files reviewed: 26/26 changed files
• Comments generated: 10
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.