nemoguard_8b_topic_control_HNPU / v81 /nemoguard-topic-8b.json
sanmonga22's picture
Prune QHexRT HNPU bundle to runtime-minimum artifacts
8ebd849 verified
Raw
History Blame Contribute Delete
1.72 kB
{
"schema_version": 1,
"model": {
"name": "nemoguard-topic-8b",
"family": "llm",
"dsp_arch": "v81",
"tokenizer_pre": "llama3"
},
"params": {
"hidden": 4096,
"vocab": 128256,
"n_layers": 32,
"max_ctx": 512,
"kv_dim": 1024,
"head_dim": 128,
"rope_theta": 500000.0,
"eos_token_id": 128001
},
"artifacts": {
"contexts": {
"decode_p0": {
"bin": "nemoguardtopic8b_decode_p0_w8.bin"
},
"decode_p1": {
"bin": "nemoguardtopic8b_decode_p1_w8.bin"
},
"decode_p2": {
"bin": "nemoguardtopic8b_decode_p2_w8.bin"
},
"decode_p3": {
"bin": "nemoguardtopic8b_decode_p3_w8.bin"
},
"decode_p4": {
"bin": "nemoguardtopic8b_decode_p4_w8.bin"
},
"decode_p5": {
"bin": "nemoguardtopic8b_decode_p5_w8.bin"
},
"decode_p6": {
"bin": "nemoguardtopic8b_decode_p6_w8.bin"
},
"decode_p7": {
"bin": "nemoguardtopic8b_decode_p7_w8.bin"
},
"lmhead": {
"bin": "nemoguardtopic8b_lmhead_w8.bin"
}
},
"embed": "nemoguardtopic8b_embed_f16.bin",
"tokenizer": "tokenizer.json"
},
"plan": {
"steps": [
{
"host": "llama_generate_sharded",
"params": {
"decode_parts": "nemoguardtopic8b_decode_p0_w8,nemoguardtopic8b_decode_p1_w8,nemoguardtopic8b_decode_p2_w8,nemoguardtopic8b_decode_p3_w8,nemoguardtopic8b_decode_p4_w8,nemoguardtopic8b_decode_p5_w8,nemoguardtopic8b_decode_p6_w8,nemoguardtopic8b_decode_p7_w8",
"lmhead": "nemoguardtopic8b_lmhead_w8",
"max_new": 64,
"neg": -50000,
"rope_factor": 8.0
}
}
]
}
}