LLaDA-8B-dLLM-CoreAI / macos /metadata.json
mlboydaisuke's picture
Swap to S=256 bundle (no answer truncation)
680ce68 verified
Raw
History Blame Contribute Delete
641 Bytes
{
"metadata_version": "0.2",
"kind": "dllm",
"name": "LLaDA-8B d3LLM (diffusion)",
"assets": { "main": "llada8b_int4hu_s256.aimodel" },
"language": {
"tokenizer": "GSAI-ML/LLaDA-8B-Instruct",
"vocab_size": 126464,
"max_context_length": 256,
"embedded_tokenizer": true,
"function_map": { "main": ["main"] }
},
"diffusion": { "seq": 256, "block_size": 32, "threshold": 1.0, "mask_token_id": 126336, "eos_token_id": 126081 },
"source": { "model_definition": "torch", "hf_model_id": "d3LLM/d3LLM_LLaDA" },
"compression": "int4 per-block-32 body + int8 head",
"compilation": { "targets": ["macos-gpu"] }
}