AMAImedia/Nemotron-Orchestrator-8B-Qwen3-BF16-NOESIS
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: AMAImedia/Nemotron-Orchestrator-8B-Qwen3-BF16-NOESIS
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-04-20 09:19:59 +08:00
commit 0c4e1c7daf
16 changed files with 1204 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
noesis_provenance.json Normal file
View File

@@ -0,0 +1,21 @@
{
"noesis_version": "v14.1",
"framework": "DHCF-FNO",
"founder": "Ilia Bolotnikov",
"organization": "AMAImedia.com",
"source_model": "nvidia/Nemotron-Orchestrator-8B",
"source_format": "FP32 safetensors",
"source_license": "NVIDIA Open Model License (research and development only)",
"base_model": "Qwen/Qwen3-8B",
"architecture": "Qwen3ForCausalLM (dense decoder-only, NO MoE)",
"vocab_size": 151936,
"language": "en",
"conversion": {
"operation": "FP32 -> BF16 cast",
"method": "torch.Tensor.to(dtype=torch.bfloat16)",
"rounding": "IEEE 754 round-to-nearest-even (PyTorch default)",
"lossless_for_inference": true,
"reason": "BF16 has same 8-bit exponent as FP32; 7-bit mantissa sufficient for weight storage"
},
"purpose": "Bandwidth-friendly BF16 reference checkpoint for downstream quantization and inference"
}