Files
unslothai--unsloth/studio/backend/assets/configs/inference_defaults.json
T
wehub-resource-sync e93507a09c
Lockfile supply-chain audit / lockfile supply-chain audit (push) Has been cancelled
Windows Studio GGUF CI / GPU prebuilt resolves without Visual Studio (push) Has been cancelled
Windows Studio GGUF CI / setup.ps1 unit tests (VS 2026 / CMake guard) (push) Has been cancelled
Windows Studio GGUF CI / real-VS detection (VS 2022) (push) Has been cancelled
Windows Studio GGUF CI / real-VS detection (VS 2026) (push) Has been cancelled
Windows Studio GGUF CI / VC++ runtime detect + install round-trip (windows-2025-vs2026) (push) Has been cancelled
Windows Studio GGUF CI / VC++ runtime detect + install round-trip (windows-latest) (push) Has been cancelled
Windows Studio Update CI / Studio Updating Tests (push) Has been cancelled
Wheel CI / Wheel build + content sanity + import smoke (push) Has been cancelled
Lint CI / Source lint (Python + shell + YAML + JSON + safety nets) (push) Has been cancelled
MLX CI on Mac M1 / dispatch (push) Has been cancelled
Security audit / advisory audit (pip + npm + cargo) (push) Has been cancelled
Security audit / pip scan-packages :: extras (push) Has been cancelled
Security audit / pip scan-packages :: studio (push) Has been cancelled
Security audit / pip scan-packages :: hf-stack (push) Has been cancelled
Security audit / npm scan-packages (Studio frontend tarballs) (push) Has been cancelled
Security audit / workflow-trigger lint (pull_request_target / cache-poisoning) (push) Has been cancelled
Security audit / pytest tests/security (push) Has been cancelled
Security audit / npm provenance + new install-script diff (push) Has been cancelled
Studio API CI / Studio API & Auth Tests (push) Has been cancelled
Backend CI / (Python 3.10) (push) Has been cancelled
Backend CI / (Python 3.11) (push) Has been cancelled
Backend CI / (Python 3.12) (push) Has been cancelled
Backend CI / (Python 3.13) (push) Has been cancelled
Backend CI / Repo tests (CPU) (push) Has been cancelled
Frontend CI / Frontend build + bundle sanity (push) Has been cancelled
Studio GGUF CI / OpenAI, Anthropic API tests (push) Has been cancelled
Studio GGUF CI / Tool calling Tests (push) Has been cancelled
Studio GGUF CI / JSON, images (push) Has been cancelled
Mac Studio GGUF CI / OpenAI, Anthropic API tests (push) Has been cancelled
Mac Studio GGUF CI / Tool calling Tests (push) Has been cancelled
Mac Studio GGUF CI / JSON, images (push) Has been cancelled
Mac Studio Install Matrix CI / Install + load (macos-14) (push) Has been cancelled
Mac Studio Install Matrix CI / Install + load (macos-15) (push) Has been cancelled
Mac Studio Install Matrix CI / Install + load (macos-26) (push) Has been cancelled
Mac Studio Install Matrix CI / Install + load (macos-15-intel) (push) Has been cancelled
Mac Studio API CI / Studio API & Auth Tests (push) Has been cancelled
Mac Studio Install Matrix CI / Install + load (macos-26-intel) (push) Has been cancelled
Mac Studio UI CI / Chat UI Tests (push) Has been cancelled
Studio Tauri CI / Tauri Linux debug build (no codesign) (push) Has been cancelled
Mac Studio Update CI / Studio Updating Tests (push) Has been cancelled
Studio UI CI / Chat UI Tests (push) Has been cancelled
Windows Studio API CI / Studio API & Auth Tests (push) Has been cancelled
Windows Studio UI CI / Chat UI Tests (push) Has been cancelled
Studio Update CI / Studio Updating Tests (push) Has been cancelled
Core / Core (HF=default + TRL=default) (push) Has been cancelled
Core / Core (HF=4.57.6 + TRL<1) (push) Has been cancelled
Core / Core (HF=latest + TRL=latest) (push) Has been cancelled
Core / llama.cpp build + smoke (push) Has been cancelled
Windows Studio GGUF CI / OpenAI, Anthropic API tests (push) Has been cancelled
Windows Studio GGUF CI / Tool calling Tests (push) Has been cancelled
Windows Studio GGUF CI / JSON, images (push) Has been cancelled
Windows Studio GGUF CI / Studio install + inference without Visual Studio (push) Has been cancelled
Studio export capability / capability (macos-latest) (push) Has been cancelled
Studio export capability / capability (ubuntu-latest) (push) Has been cancelled
Studio export capability / capability (windows-latest) (push) Has been cancelled
Cross-platform parity / parity (macos-latest) (push) Has been cancelled
Cross-platform parity / parity (windows-latest) (push) Has been cancelled
Scorecard supply-chain security / Scorecard analysis (push) Has been cancelled
Studio load-orchestrator CI / test (push) Has been cancelled
chore: import upstream snapshot with attribution
2026-07-13 12:59:56 +08:00

413 lines
8.7 KiB
JSON

{
"_comment": "Per-model-family inference parameter defaults. Sources: (1) Ollama params blobs, (2) Existing Unsloth Studio YAML configs. Patterns ordered longest-match-first.",
"families": {
"qwen3.6": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0,
"presence_penalty": 1.5
},
"qwen3.5": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0,
"presence_penalty": 1.5
},
"qwen3-coder": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwen3-next": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwen3-vl": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwen3": {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwen2.5-coder": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"qwen2.5-vl": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"qwen2.5-omni": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwen2.5-math": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwen2.5": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwen2-vl": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"qwen2": {
"temperature": 0.7,
"top_p": 0.8,
"top_k": 20,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"qwq": {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 40,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"gemma-4": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 64,
"min_p": 0.0,
"repetition_penalty": 1.0,
"presence_penalty": 0.0
},
"gemma-3n": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 64,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"gemma-3": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 64,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"medgemma": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 64,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"gemma-2": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 64,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"llama-4": {
"temperature": 1.0,
"top_p": 0.9,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"llama-3.3": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"llama-3.2": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"llama-3.1": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"llama-3": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"phi-4": {
"temperature": 0.8,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"phi-3": {
"temperature": 0.7,
"top_p": 0.9,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"mistral-nemo": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"mistral-small": {
"temperature": 0.15,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"mistral-large": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"magistral": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"ministral": {
"temperature": 0.15,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"devstral": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"pixtral": {
"temperature": 1.5,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.1,
"repetition_penalty": 1.0
},
"deepseek-v4": {
"temperature": 1.0,
"top_p": 1.0,
"top_k": -1,
"min_p": 0.0,
"repetition_penalty": 1.0
},
"deepseek-r1": {
"temperature": 0.6,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"deepseek-v3": {
"temperature": 0.6,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"deepseek-ocr": {
"temperature": 0.0,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"glm-5": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"glm-4": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"nemotron": {
"temperature": 1.0,
"top_p": 1.0,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"minimax-m2.7": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 40,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"minimax-m2.5": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 40,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"minimax": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": 40,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"gpt-oss": {
"temperature": 1.0,
"top_p": 1.0,
"top_k": 0,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"granite-4": {
"temperature": 0.0,
"top_p": 1.0,
"top_k": 0,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"kimi-k2": {
"temperature": 0.6,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"kimi": {
"temperature": 0.6,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"lfm2": {
"temperature": 0.1,
"top_p": 0.1,
"top_k": 50,
"min_p": 0.15,
"repetition_penalty": 1.05
},
"smollm": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"olmo": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"falcon": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"ernie": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"seed": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"grok": {
"temperature": 1.0,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
},
"mimo": {
"temperature": 0.7,
"top_p": 0.95,
"top_k": -1,
"min_p": 0.01,
"repetition_penalty": 1.0
}
},
"patterns": [
"qwen3.6", "qwen3.5",
"qwen3-coder", "qwen3-next", "qwen3-vl", "qwen3",
"qwen2.5-coder", "qwen2.5-vl", "qwen2.5-omni", "qwen2.5-math", "qwen2.5",
"qwen2-vl", "qwen2",
"qwq",
"gemma-4", "gemma-3n", "gemma-3", "medgemma", "gemma-2",
"llama-4", "llama-3.3", "llama-3.2", "llama-3.1", "llama-3",
"phi-4", "phi-3",
"mistral-nemo", "mistral-small", "mistral-large", "magistral", "ministral",
"devstral", "pixtral",
"deepseek-v4", "deepseek-r1", "deepseek-v3", "deepseek-ocr",
"glm-5", "glm-4",
"nemotron",
"minimax-m2.7", "minimax-m2.5", "minimax",
"gpt-oss", "granite-4",
"kimi-k2", "kimi",
"lfm2", "smollm", "olmo", "falcon", "ernie", "seed", "grok", "mimo"
]
}