Add Core ML package (prefill on ANE + stateful GPU decode), tokenizer, host config

48abc2c verified 2 days ago

393 Bytes

	{
	"base_model": "Qwen/Qwen2.5-Coder-0.5B-Instruct",
	"format": "coreml",
	"package": "qwen2_5_coder_0_5b.mlpackage",
	"prefillLength": 448,
	"cacheLength": 1024,
	"layers": 24,
	"kvHeads": 2,
	"headDim": 64,
	"padID": 151643,
	"stopIDs": [
	151645,
	151643
	],
	"maxNewTokens": 512,
	"systemPrompt": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."
	}