[AMD] Add RyzenAI 8D + MLPerf recipes (Qwen3-8B, Phi-4, Llama-3.1-8B)

[Pooja Ganesh (poganesh)]

• Aligns RyzenAI recipes with the AMD - Microsoft 8D drop.
• Adds MLPerf 2.0 recipes (full_fusion / hybrid) for Qwen3-8B, Phi-4-mini, and Llama-3.1-8B.
• Updates gpt-oss inputs.
• Updates AMD recipe requirements with: OGA == v0.14.0, quark == 0.12.post1, onnx==1.22.0, numpy == 2.5.1
• Includes README updates for the updated models.

Note: on-device runtime validation is pending an internal fix to transaction-binary filtering; the recipes generate correctly.

[Copilot]

Pull request overview

This PR updates and expands the AMD RyzenAI recipe set to align with the “8D” drop, adding new RyzenAI configurations (including MLPerf-oriented full_fusion and hybrid variants) and updating RyzenAI dependency pins across multiple existing model recipes.

Changes:

• Added RyzenAI recipe configs + documentation for Qwen/Qwen3-8B, including MLPerf full_fusion and hybrid variants and a new info.yaml.
• Added MLPerf full_fusion and hybrid RyzenAI recipes (plus README/info updates) for microsoft/Phi-4-mini-instruct and meta-llama/Llama-3.1-8B-Instruct.
• Updated many existing RyzenAI requirements_ryzenai_llm.txt files to the new pinned versions (e.g., Olive 1.8.0-8D index, amd-quark==0.12.post1, onnx==1.22.0, numpy==2.5.1, onnxruntime-genai==0.14.0) and updated several token_fusion configs to include max_seq_len in extra_options.

Reviewed changes

Copilot reviewed 81 out of 81 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
Qwen-Qwen3-8B/RyzenAI/requirements_ryzenai_llm.txt	New RyzenAI dependency set for Qwen3-8B aligned to 8D pins.
Qwen-Qwen3-8B/RyzenAI/README.md	New/updated setup + run instructions for Qwen3-8B (token/full/hybrid).
Qwen-Qwen3-8B/RyzenAI/Qwen3-8B_quark_ryzenai_llm.json	New Qwen3-8B token_fusion RyzenAI recipe.
Qwen-Qwen3-8B/RyzenAI/Qwen3-8B_quark_ryzenai_llm_hybrid.json	New Qwen3-8B MLPerf hybrid RyzenAI recipe.
Qwen-Qwen3-8B/RyzenAI/Qwen3-8B_quark_ryzenai_llm_full_fusion.json	New Qwen3-8B MLPerf full_fusion RyzenAI recipe.
Qwen-Qwen3-8B/RyzenAI/info.yaml	New recipe registration for Qwen3-8B RyzenAI variants.
Qwen-Qwen2.5-Coder-7B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-Coder-7B-Instruct/RyzenAI/Qwen2.5-Coder-7B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2.5-Coder-1.5B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-Coder-1.5B-Instruct/RyzenAI/Qwen2.5-Coder-1.5B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2.5-Coder-0.5B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-Coder-0.5B-Instruct/RyzenAI/Qwen2.5-Coder-0.5B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2.5-7B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-7B-Instruct/RyzenAI/Qwen2.5-7B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2.5-3B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-3B/RyzenAI/Qwen2.5-3B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2.5-3B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-3B-Instruct/RyzenAI/Qwen2.5-3B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2.5-1.5B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-1.5B-Instruct/RyzenAI/Qwen2.5-1.5B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2.5-0.5B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2.5-0.5B-Instruct/RyzenAI/Qwen2.5-0.5B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2-7B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2-7B/RyzenAI/Qwen2-7B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen2-1.5B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen2-1.5B/RyzenAI/Qwen2-1.5B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
Qwen-Qwen1.5-7B-Chat/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
Qwen-Qwen1.5-7B-Chat/RyzenAI/Qwen1.5-7B-Chat_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
mistralai-Mistral-7B-v0.3/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
mistralai-Mistral-7B-v0.3/RyzenAI/Mistral-7B-v0.3_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
mistralai-Mistral-7B-Instruct-v0.3/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
mistralai-Mistral-7B-Instruct-v0.3/RyzenAI/Mistral-7B-Instruct-v0.3_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
mistralai-Mistral-7B-Instruct-v0.2/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
mistralai-Mistral-7B-Instruct-v0.2/RyzenAI/Mistral-7B-Instruct-v0.2_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
mistralai-Mistral-7B-Instruct-v0.1/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
mistralai-Mistral-7B-Instruct-v0.1/RyzenAI/Mistral-7B-Instruct-v0.1_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
microsoft-Phi-4-mini-reasoning/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
microsoft-Phi-4-mini-reasoning/RyzenAI/Phi-4-mini-reasoning_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
microsoft-Phi-4-mini-instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
microsoft-Phi-4-mini-instruct/RyzenAI/README.md	Updated docs to include MLPerf full_fusion and hybrid recipe variants.
microsoft-Phi-4-mini-instruct/RyzenAI/Phi-4-mini-instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
microsoft-Phi-4-mini-instruct/RyzenAI/Phi-4-mini-instruct_quark_ryzenai_llm_hybrid.json	Added new MLPerf hybrid RyzenAI recipe.
microsoft-Phi-4-mini-instruct/RyzenAI/Phi-4-mini-instruct_quark_ryzenai_llm_full_fusion.json	Added new MLPerf full_fusion RyzenAI recipe.
microsoft-Phi-4-mini-instruct/RyzenAI/info.yaml	Registered new Phi-4-mini-instruct RyzenAI MLPerf variants.
microsoft-Phi-3.5-mini-instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
microsoft-Phi-3.5-mini-instruct/RyzenAI/Phi-3.5-mini-instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
microsoft-Phi-3-mini-4k-instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
microsoft-Phi-3-mini-4k-instruct/RyzenAI/Phi-3-mini-4k-instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
microsoft-Phi-3-mini-128k-instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
microsoft-Phi-3-mini-128k-instruct/RyzenAI/Phi-3-mini-128k-instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Meta-Llama-3-8B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Meta-Llama-3-8B/RyzenAI/Meta-Llama-3-8B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Llama-3.2-3B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Llama-3.2-3B/RyzenAI/Llama-3.2-3B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Llama-3.2-3B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Llama-3.2-3B-Instruct/RyzenAI/Llama-3.2-3B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Llama-3.2-1B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Llama-3.2-1B/RyzenAI/Llama-3.2-1B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Llama-3.2-1B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Llama-3.2-1B-Instruct/RyzenAI/Llama-3.2-1B-Instruct_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Llama-3.1-8B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Llama-3.1-8B/RyzenAI/Meta-Llama-3.1-8B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Llama-3.1-8B-Instruct/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Llama-3.1-8B-Instruct/RyzenAI/README.md	Updated docs to include MLPerf full_fusion and hybrid recipe variants.
meta-llama-Llama-3.1-8B-Instruct/RyzenAI/Llama-3.1-8B-Instruct_quark_ryzenai_llm.json	Updated token_fusion recipe with RyzenGenerateModelLLM extra_options (model_type).
meta-llama-Llama-3.1-8B-Instruct/RyzenAI/Llama-3.1-8B-Instruct_quark_ryzenai_llm_hybrid.json	Added new MLPerf hybrid RyzenAI recipe.
meta-llama-Llama-3.1-8B-Instruct/RyzenAI/Llama-3.1-8B-Instruct_quark_ryzenai_llm_full_fusion.json	Added new MLPerf full_fusion RyzenAI recipe.
meta-llama-Llama-3.1-8B-Instruct/RyzenAI/info.yaml	Registered new Llama-3.1-8B-Instruct RyzenAI MLPerf variants.
meta-llama-Llama-2-7b-hf/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
meta-llama-Llama-2-7b-hf/RyzenAI/Llama-2-7b-hf_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
meta-llama-Llama-2-7b-chat-hf/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
gpt-oss-20b/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
gpt-oss-20b/RyzenAI/README.md	Cleaned up incorrect/irrelevant comment in run instructions.
deepseek-ai-DeepSeek-R1-Distill-Qwen-7B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
deepseek-ai-DeepSeek-R1-Distill-Qwen-7B/RyzenAI/DeepSeek-R1-Distill-Qwen-7B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
deepseek-ai-DeepSeek-R1-Distill-Qwen-1.5B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
deepseek-ai-DeepSeek-R1-Distill-Qwen-1.5B/RyzenAI/DeepSeek-R1-Distill-Qwen-1.5B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
deepseek-ai-DeepSeek-R1-Distill-Llama-8B/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
deepseek-ai-DeepSeek-R1-Distill-Llama-8B/RyzenAI/DeepSeek-R1-Distill-Llama-8B_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.
codellama-CodeLlama-7b-Instruct-hf/RyzenAI/requirements_ryzenai_llm.txt	Updated RyzenAI dependency pins to 8D versions.
codellama-CodeLlama-7b-Instruct-hf/RyzenAI/CodeLlama-7b-Instruct-hf_quark_ryzenai_llm.json	Added max_seq_len extra option to token_fusion recipe.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.