Hugging Face's logo

anemll
/
anemll-quantized-qwen3.8-27b-for-CoreAI

Text Generation
Safetensors
qwen3_5
coreai
apple-neural-engine
ane
quantized
gptq
vector-quantization
lut
speculative-decoding
dflash2
Model card Files
xet
 Community
anemll-quantized-qwen3.8-27b-for-CoreAI
File size: 3,522 Bytes
29ab1ad
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
{
  "upstream_model": {
    "id": "ProCreations/Ternary-Bonsai-2-27B-DFlash2",
    "revision": "4cfb6ad03268fed0f60ca96c1a659c0b1c77e50b",
    "license": "apache-2.0",
    "url": "https://huggingface.co/ProCreations/Ternary-Bonsai-2-27B-DFlash2/tree/4cfb6ad03268fed0f60ca96c1a659c0b1c77e50b"
  },
  "upstream_donors": [
    {
      "id": "z-lab/Qwen3.8-27B-DFlash2",
      "revision": "50307d4c4cde6860d4eee73e2547cd786fe8e8a4"
    },
    {
      "id": "prism-ml/Ternary-Bonsai-2-27B-gguf",
      "revision": "6ed5e12bf84b7a63069882c91dd9e9218647d17b"
    }
  ],
  "checkpoint": {
    "bytes": 3848817896,
    "sha256": "708e141bf4a33b09620c4f86263f1c7edcec7882f5dd39448e83e0c0defdc230",
    "fully_rehashed_and_matched_upstream_lfs": true
  },
  "runtime": {
    "model": "dflash2_lut4_gptq.aimodel",
    "source_package_files": [
      {
        "path": "main.hash",
        "bytes": 32,
        "sha256": "85e795ea77330297f1a04ddce4b2b2bbc89adfa3a5837805d480862a320b3cf4"
      },
      {
        "path": "main.mlirb",
        "bytes": 1636296986,
        "sha256": "dd523222671fa19f0df2a789c9a17ec5d396586d3bf25622aeb6e0c22108ec1b"
      },
      {
        "path": "metadata.json",
        "bytes": 105,
        "sha256": "7d39390ddb77548b2899763741dfd9134d23dcace92ff550a7bbcf1d46fc8b68"
      }
    ],
    "source_sidecar_sha256": "5db961b5797a7619dc7cf091067cae220dad9a5da748d315a2aac17a1e776bed",
    "target_export": "mix25in_mixr_lr64mix",
    "head_export": "mix25in_mixr_lr64mix/lm_head.safetensors",
    "calibration_export": "drafter_lut4_gptq_q7_cal",
    "mask_scale": 0.7,
    "context_ring_rows": 2048,
    "block_rows": 8,
    "proposals_per_cycle": 7,
    "builder": "coreai-core 1.0.0b2",
    "creation_date": "20260928T201801Z"
  },
  "selector_extraction": {
    "file": "selector.safetensors",
    "tensors": [
      {
        "name": "candidate_selector.predecessor_codebook",
        "dtype": "BF16",
        "shape": [
          248320,
          256
        ],
        "raw_tensor_sha256": "12fed59a4c90bedecaa0d4ddfb555a627d6054fd444ad2f11accac06929a43cc"
      },
      {
        "name": "candidate_selector.successor_codebook",
        "dtype": "BF16",
        "shape": [
          248320,
          256
        ],
        "raw_tensor_sha256": "09827c1de69d1acb474ac318aceeab9b020c2502dbf759e26a4941f0b1765545"
      }
    ],
    "only_required_keys": true,
    "tensor_values_bit_identical": true
  },
  "upstream_notices": [
    {
      "path": "LICENSE",
      "bytes": 10174,
      "sha256": "69849221bfb90053de2134ef5e6d540287b4b98062326492f1f96f5da685524b",
      "copied_without_modification": true
    },
    {
      "path": "NOTICE",
      "bytes": 446,
      "sha256": "48da4ef105620c8a5ed6d94853efc4588d914bbd436072ec3117aad7d7729db3",
      "copied_without_modification": true
    }
  ],
  "modifications": [
    "LUT4 GPTQ Core AI body with paired mixr target head and mask-row scale 0.7.",
    "Compact selector file extracts two unmodified BF16 tables from the pinned source.",
    "Sidecar local paths replaced with portable identifiers and a modification notice."
  ],
  "verification_gaps": [
    "Core AI body tensors have not been independently reconstructed from the pinned BF16 checkpoint and GPTQ export.",
    "Head/target identity and q7_cal calibration linkage are recorded by the source sidecar/export metadata, not a new tensor-level reconstruction.",
    "Speculative acceptance, generation, performance and ANE placement were not rerun during packaging."
  ]
}