clef-text-0.6b-coreml / config.json
alexwengg's picture
clef-text-0.6b Core ML: fp16 decoder (L256-L2048) + head, GPU / Neural Engine
3c61a60 verified
Raw History Blame Contribute Delete
406 Bytes
{
"model": "clef-text-0.6b",
"base_model": "Qwen/Qwen3-0.6B",
"teacher": "Cloudflare/clef-flash",
"hidden_size": 1024,
"head_dim": 128,
"rope_theta": 1000000,
"pad_id": 151643,
"vocab_size": 151936,
"buckets": [
256,
512,
1024,
2048
],
"max_questions": 16,
"max_options": 96,
"decoder": "Decoder.mlpackage",
"head": "Head.mlpackage",
"tied_embeddings": true
}