Record architecture-specific runtime format audit and n-gram scope
Browse files- README.md +11 -0
- runtime_audit.json +37 -0
README.md
CHANGED
|
@@ -15,6 +15,17 @@ tags:
|
|
| 15 |
- development-preview
|
| 16 |
---
|
| 17 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
# AX-DeepSeek-OCR-2-CUDA-AXQ-NVFP4-W4A4
|
| 19 |
|
| 20 |
**Development preview with native FP4 execution.** Native AXQuant RTN converts
|
|
|
|
| 15 |
- development-preview
|
| 16 |
---
|
| 17 |
|
| 18 |
+
## Runtime format audit (2026-10-06)
|
| 19 |
+
|
| 20 |
+
No quantization-container correction was needed.
|
| 21 |
+
This family has no n-gram tensors; no n-gram file or declaration was added.
|
| 22 |
+
This CUDA pack is outside MLX/oMLX/MTPLX export scope.
|
| 23 |
+
|
| 24 |
+
See [runtime_audit.json](runtime_audit.json) for pinned config/index/header
|
| 25 |
+
bindings, architecture, physical-format findings, and applied corrections.
|
| 26 |
+
This is development evidence; no quality, MTP exactness, speed, or certification
|
| 27 |
+
claim is added. Historical evidence stays bound to its original revision.
|
| 28 |
+
|
| 29 |
# AX-DeepSeek-OCR-2-CUDA-AXQ-NVFP4-W4A4
|
| 30 |
|
| 31 |
**Development preview with native FP4 execution.** Native AXQuant RTN converts
|
runtime_audit.json
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"applied_config_corrections": [],
|
| 3 |
+
"current_config_sha256": "4e2c1b1a75b4da1bc730d6f8fdd3ac5d2ea91c4016796809e4d0d6288d941d08",
|
| 4 |
+
"date": "2026-10-06",
|
| 5 |
+
"header_sha256": {
|
| 6 |
+
"model-00001-of-000001.safetensors": "985e7d88819f706dca18560a65bebb850c68630fb2438cfe5100a9be55be9478"
|
| 7 |
+
},
|
| 8 |
+
"input_bindings": {
|
| 9 |
+
"config.json": "4e2c1b1a75b4da1bc730d6f8fdd3ac5d2ea91c4016796809e4d0d6288d941d08",
|
| 10 |
+
"model.safetensors.index.json": "1bde1f2ecd63fcfe16e72e81539418ea406e3a19b907520b1a072f634a4f8fe6"
|
| 11 |
+
},
|
| 12 |
+
"issues": [],
|
| 13 |
+
"model_type": "deepseek_vl_v2",
|
| 14 |
+
"mtp_files": [],
|
| 15 |
+
"ngram_action": "none; do not invent n-gram data",
|
| 16 |
+
"ngram_files": [],
|
| 17 |
+
"ngram_quantization": [],
|
| 18 |
+
"ngram_table_metadata": null,
|
| 19 |
+
"ngram_tensor_count": 0,
|
| 20 |
+
"quality_certified": false,
|
| 21 |
+
"quantization": {
|
| 22 |
+
"quantization_config": {
|
| 23 |
+
"quant_method": "compressed-tensors"
|
| 24 |
+
}
|
| 25 |
+
},
|
| 26 |
+
"removed_source_evidence": [],
|
| 27 |
+
"repo_id": "AutomatosX/AX-DeepSeek-OCR-2-CUDA-AXQ-NVFP4-W4A4",
|
| 28 |
+
"runtime_arch_id": null,
|
| 29 |
+
"runtime_verified": false,
|
| 30 |
+
"schema_version": "axquant.hub-runtime-audit.v1",
|
| 31 |
+
"scope": "Pinned remote config/index/Safetensors header audit; no runtime load or generation claim.",
|
| 32 |
+
"source_revision": "7e07d08981bfd89b1822fa6a32346ea3f7d55712",
|
| 33 |
+
"status": "outside-mlx-peer-scope",
|
| 34 |
+
"unchanged_weight_sha256": {
|
| 35 |
+
"model-00001-of-000001.safetensors": "183a0741f37efa8338a9fd897316ba7e29f89a54f334f58ca84eaa689e2ffa4c"
|
| 36 |
+
}
|
| 37 |
+
}
|