AutomatosX commited on
Commit
bbaf496
·
verified ·
1 Parent(s): 4fe784f

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +1 -0
  2. README.md +176 -0
  3. axquant_manifest.json +406 -0
  4. axquant_mtp_sidecar_manifest.json +34 -0
  5. axquant_plan.json +3 -0
  6. axquant_quantizer_execution.json +0 -0
  7. axquant_runtime.json +58 -0
  8. config.json +0 -0
  9. generation_config.json +9 -0
  10. model-00001-of-00043.safetensors +3 -0
  11. model-00002-of-00043.safetensors +3 -0
  12. model-00003-of-00043.safetensors +3 -0
  13. model-00004-of-00043.safetensors +3 -0
  14. model-00005-of-00043.safetensors +3 -0
  15. model-00006-of-00043.safetensors +3 -0
  16. model-00007-of-00043.safetensors +3 -0
  17. model-00008-of-00043.safetensors +3 -0
  18. model-00009-of-00043.safetensors +3 -0
  19. model-00010-of-00043.safetensors +3 -0
  20. model-00011-of-00043.safetensors +3 -0
  21. model-00012-of-00043.safetensors +3 -0
  22. model-00013-of-00043.safetensors +3 -0
  23. model-00014-of-00043.safetensors +3 -0
  24. model-00015-of-00043.safetensors +3 -0
  25. model-00016-of-00043.safetensors +3 -0
  26. model-00017-of-00043.safetensors +3 -0
  27. model-00018-of-00043.safetensors +3 -0
  28. model-00019-of-00043.safetensors +3 -0
  29. model-00020-of-00043.safetensors +3 -0
  30. model-00021-of-00043.safetensors +3 -0
  31. model-00022-of-00043.safetensors +3 -0
  32. model-00023-of-00043.safetensors +3 -0
  33. model-00024-of-00043.safetensors +3 -0
  34. model-00025-of-00043.safetensors +3 -0
  35. model-00026-of-00043.safetensors +3 -0
  36. model-00027-of-00043.safetensors +3 -0
  37. model-00028-of-00043.safetensors +3 -0
  38. model-00029-of-00043.safetensors +3 -0
  39. model-00030-of-00043.safetensors +3 -0
  40. model-00031-of-00043.safetensors +3 -0
  41. model-00032-of-00043.safetensors +3 -0
  42. model-00033-of-00043.safetensors +3 -0
  43. model-00034-of-00043.safetensors +3 -0
  44. model-00035-of-00043.safetensors +3 -0
  45. model-00036-of-00043.safetensors +3 -0
  46. model-00037-of-00043.safetensors +3 -0
  47. model-00038-of-00043.safetensors +3 -0
  48. model-00039-of-00043.safetensors +3 -0
  49. model-00040-of-00043.safetensors +3 -0
  50. model-00041-of-00043.safetensors +3 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ axquant_plan.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,176 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: mlx
4
+ base_model: deepseek-ai/DeepSeek-V4-Flash
5
+ base_model_relation: quantized
6
+ pipeline_tag: text-generation
7
+ tags:
8
+ - mlx
9
+ - apple-silicon
10
+ - quantized
11
+ - mixed-precision
12
+ - axquant
13
+ - axq
14
+ - development
15
+ - deepseek-v4
16
+ - 4bit
17
+ - 4-bit
18
+ - mtp
19
+ ---
20
+
21
+ # AX-DeepSeek-V4-Flash-MLX-AXQ-4bit
22
+
23
+ An **AXQuant (AXQ)** mixed-precision MLX checkpoint for Apple Silicon, converted directly from
24
+ the BF16 source model. The language path is quantized while the multi-token-prediction (MTP) head are preserved at BF16 in the checkpoint (or a bound sidecar when present).
25
+
26
+ > **Development evidence — not a certified AXQuant release.** This package has conversion and
27
+ > artifact-integrity records, but it does not publish measured quality, long-context, kernel-speed,
28
+ > or MTP-speed evidence. Do not interpret the AXQ product label as a benchmark claim.
29
+
30
+
31
+ ## Model details
32
+
33
+ | Property | Value |
34
+ | --- | --- |
35
+ | Base model | [deepseek-ai/DeepSeek-V4-Flash](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash/tree/60d8d70770c6776ff598c94bb586a859a38244f1) |
36
+ | Source revision | `60d8d70770c6776ff598c94bb586a859a38244f1` |
37
+ | Product family | `deepseek-v4` |
38
+ | Source architecture | `DeepseekV4ForCausalLM` (mixture of experts (MoE)); text path optimized |
39
+ | Main-model parameters | 284.33B logical parameters |
40
+ | Quantizer | AXQuant `1.5.1` |
41
+ | Hub budget class | `4bit` |
42
+ | AXQuant base precision class | `4bit` |
43
+ | Planned storage-adjusted BPW | 4.8500 |
44
+ | Measured main-model BPW | 4.5929 |
45
+ | Measured total BPW, including MTP | **4.5874** |
46
+ | Safetensors weight size | 166.83 GB |
47
+ | Approximate complete download | 166.91 GB |
48
+ | Configured maximum context | 1,048,576 tokens; practical limits depend on unified memory |
49
+ | Primary MLX runtime | MLX-LM |
50
+ | AX Engine native execution | Not established; no validated native manifest is included |
51
+ | MTP present | `True` |
52
+ | Vision present | `False` |
53
+ | Audio present | `False` |
54
+
55
+ This repository contains MLX Safetensors. It does **not** contain PyTorch or GGUF weights.
56
+
57
+ ## Choosing an AXQ pack
58
+
59
+ AXQ names describe a **storage-budget product class**, not one uniform precision applied to every
60
+ tensor. Protected tensors remain at higher precision, so the exact measured BPW is authoritative.
61
+ In particular, a `6bit`-named mixed plan may retain `4bit` as its base precision while selecting
62
+ 6-bit, 8-bit, or BF16 for other tensors to meet an approximately 6-BPW total budget. Protection
63
+ floors can also raise a `4bit`-named pack close to (or above) a `6bit` budget on small or heavily
64
+ protected models. When that collapse happens, AutomatosX does **not** publish a separate
65
+ misleading `4bit` sibling for that base.
66
+
67
+
68
+ | Sibling | Intended trade-off |
69
+ | --- | --- |
70
+ | [4bit sibling](https://huggingface.co/AutomatosX/AX-DeepSeek-V4-Flash-MLX-AXQ-4bit) | Lower-storage AXQ budget; check its exact BPW |
71
+ | [6bit sibling](https://huggingface.co/AutomatosX/AX-DeepSeek-V4-Flash-MLX-AXQ-6bit) | Higher average precision near the 6-BPW budget |
72
+
73
+ See the [AutomatosX MLX model catalog](https://huggingface.co/collections/AutomatosX/automatosx-mlx-model-catalog)
74
+ for related MLX and OptiQ alternatives.
75
+
76
+ ## Download
77
+
78
+ ```bash
79
+ python -m pip install -U huggingface_hub
80
+ hf download AutomatosX/AX-DeepSeek-V4-Flash-MLX-AXQ-4bit --local-dir ./AX-DeepSeek-V4-Flash-MLX-AXQ-4bit
81
+ ```
82
+
83
+ Allow at least 166.91 GB of free disk space. Pin the resulting Hub commit in reproducible
84
+ deployments rather than relying indefinitely on `main`.
85
+
86
+ ## Run with MLX-LM
87
+
88
+ ```bash
89
+ python -m pip install -U mlx-lm
90
+ mlx_lm.generate \
91
+ --model AutomatosX/AX-DeepSeek-V4-Flash-MLX-AXQ-4bit \
92
+ --prompt "Explain mixed-precision quantization in three sentences." \
93
+ --max-tokens 128 \
94
+ --temp 0.0
95
+ ```
96
+
97
+ MLX-LM compatibility covers standard **text/backbone inference**. It may ignore AXQuant runtime
98
+ metadata and optional sidecars (`vision.safetensors`, `mtp.safetensors`); this command therefore
99
+ does not establish MTP acceleration or vision-language quality. The artifact records MLX
100
+ `0.32.0` and MLX-LM `0.31.3` from conversion.
101
+
102
+ ## AX Engine status
103
+
104
+ This package does **not** include a validated native `model-manifest.json`, so AX Engine execution
105
+ is not established by this release. The AX Engine fields in `axquant_runtime.json` describe the
106
+ intended compatibility contract, not observed runtime evidence. Use the architecture-specific MLX
107
+ runtime path above. The artifact records AX Engine version
108
+ `6.11.1`, but version discovery alone is not a runtime check.
109
+
110
+ ## Quantization layout
111
+
112
+ | Main-weight precision | Parameters | Share |
113
+ | --- | ---: | ---: |
114
+ | `4bit` | 281.75B | 96.84% |
115
+ | `8bit` | 529.53M | 0.18% |
116
+ | `bf16` | 8.67B | 2.98% |
117
+
118
+ - Quantization methods: `affine, bf16`.
119
+ - Group sizes used by quantized assignments: `32, 64`.
120
+ - MTP sidecar: 1575 tensors, 6.61B parameters, 3.59 GB, BF16, F32, F8_E4M3, F8_E8M0, I8.
121
+ - Vision sidecar: not included.
122
+ - Optimization scope: `text-path`.
123
+ - Support tier: `convertible`.
124
+
125
+ BF16 sidecars, when present, are included in total download size. Their presence does not by itself
126
+ establish MTP acceleration or vision-language quality.
127
+
128
+ ## Evidence and validation status
129
+
130
+ | Check | Status |
131
+ | --- | --- |
132
+ | Planning evidence | `architecture_prior` |
133
+ | Calibration | none; the allocation is based on architecture priors |
134
+ | Quantizer execution | 33492/33492 recorded module conversions succeeded; 0 fallbacks |
135
+ | AX Engine native manifest | not included |
136
+ | Quality versus BF16 or uniform baselines | Not published; no quality-retention claim |
137
+ | MTP acceptance and speed | not measured; no MTP speedup claim |
138
+ | AX Engine kernel evidence | `unmeasured` |
139
+ | Vision-language quality | Not applicable (no vision tower in this package) |
140
+ | Speech-recognition quality | Not applicable |
141
+ | Long-context quality | 1,048,576-token capacity is config metadata, not a validated claim |
142
+ | Release certification | **Not certified**; formal AXQuant M0-M8 gates are not closed |
143
+
144
+ ## Intended use and limitations
145
+
146
+ - Intended for local development and evaluation on Apple Silicon with MLX-compatible runtimes.
147
+ - No minimum unified-memory figure is claimed; loadability depends on model size, context length,
148
+ KV-cache policy, runtime buffers, and other processes using unified memory.
149
+ - Architecture-prior allocation is not measured sensitivity. It must not be presented as measured
150
+ model quality.
151
+ - MTP may be ignored outside AX Engine and its speedup is unmeasured for this exact checkpoint.
152
+ - The configured context window can require substantially more memory as the KV cache grows.
153
+ - AX Engine execution is not established because this package has no validated native manifest.
154
+
155
+ - Upstream capabilities, limitations, biases, and responsible-use guidance still apply.
156
+
157
+ ## Provenance and audit files
158
+
159
+ - [`axquant_manifest.json`](axquant_manifest.json): package identity, byte accounting, runtime
160
+ contract, software versions, and file checksums.
161
+ - [`axquant_plan.json`](axquant_plan.json): per-tensor precision decisions and planning evidence.
162
+ - [`axquant_quantizer_execution.json`](axquant_quantizer_execution.json): conversion coverage and
163
+ fallback records.
164
+ - [`axquant_runtime.json`](axquant_runtime.json): declared AX Engine and MLX compatibility metadata; runtime checks remain separate evidence.
165
+ - [`axquant_mtp_sidecar_manifest.json`](axquant_mtp_sidecar_manifest.json): MTP tensor provenance.
166
+
167
+ All published provenance uses repository-relative paths. Local source paths are stripped before
168
+ publication. The checkpoint was converted from BF16 rather than re-quantized from an OptiQ
169
+ artifact. If an OptiQ repository is published separately, it uses a different quantizer and
170
+ should not be assumed to have identical BPW or quality.
171
+
172
+ ## License
173
+
174
+ The checkpoint follows the upstream model license where applicable (often Apache License 2.0). See
175
+ the [deepseek-ai/DeepSeek-V4-Flash model card](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash/tree/60d8d70770c6776ff598c94bb586a859a38244f1) for license terms, model
176
+ limitations, and responsible-use guidance.
axquant_manifest.json ADDED
@@ -0,0 +1,406 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "axquant_version": "1.5.1",
3
+ "calibration": null,
4
+ "created_at": "2026-08-08T11:02:23.928237Z",
5
+ "effective_bpw": 4.849999823314483,
6
+ "files": [
7
+ {
8
+ "path": "README.md",
9
+ "sha256": "13fe7a4f78888af6434a44da7ac7cd27937cd126c41e67dd51e00d61cfc346dd",
10
+ "size_bytes": 8020
11
+ },
12
+ {
13
+ "path": "axquant_mtp_sidecar_manifest.json",
14
+ "sha256": "e438b603bf22bdee2d90ad00ee747c7b62e6e78ad6a082ae18bfcbc90e32d8b3",
15
+ "size_bytes": 938
16
+ },
17
+ {
18
+ "path": "axquant_plan.json",
19
+ "sha256": "f74baadfd219d0efbd1e33ba5810fa8588fbe65648dc05322b74bf728fd6500d",
20
+ "size_bytes": 61520034
21
+ },
22
+ {
23
+ "path": "axquant_quantizer_execution.json",
24
+ "sha256": "bdddb917d89cbf366d233df0735ec9cd508bccce2ac08bbf8df8d95a8cfec382",
25
+ "size_bytes": 8049291
26
+ },
27
+ {
28
+ "path": "axquant_runtime.json",
29
+ "sha256": "669d13662ec73b7dbe0b10fff7638464c9e80ace2307213771d6702337816633",
30
+ "size_bytes": 1676
31
+ },
32
+ {
33
+ "path": "config.json",
34
+ "sha256": "381a8059166bcd5cc1910d4ee29c042a587026f0bbfe4219f4a240834ada4765",
35
+ "size_bytes": 159869
36
+ },
37
+ {
38
+ "path": "generation_config.json",
39
+ "sha256": "5fccff80f55a4d455bbe516bdd552edf3e9623df95e99fbf2a3c3389fdf91af0",
40
+ "size_bytes": 170
41
+ },
42
+ {
43
+ "path": "model-00001-of-00043.safetensors",
44
+ "sha256": "742f1f8bb2a32fedd140f45dad06152c0ed02cddc9f9622305a152836d3762e7",
45
+ "size_bytes": 4431029612
46
+ },
47
+ {
48
+ "path": "model-00002-of-00043.safetensors",
49
+ "sha256": "e698cad9e6d6196e85e44fbe77030691c1698c031e645596cb80ca1887699d4e",
50
+ "size_bytes": 3765589702
51
+ },
52
+ {
53
+ "path": "model-00003-of-00043.safetensors",
54
+ "sha256": "c98f53e0aae2cf2c7769503884b23ad5375045be7cf2a727225346ecd67d5c7b",
55
+ "size_bytes": 3754322827
56
+ },
57
+ {
58
+ "path": "model-00004-of-00043.safetensors",
59
+ "sha256": "048a8eb15988f3a00e767449e5aeb094be244136dc001e31e28534bec1fd5f96",
60
+ "size_bytes": 3762487939
61
+ },
62
+ {
63
+ "path": "model-00005-of-00043.safetensors",
64
+ "sha256": "ed5f606fa477aa88c0c678b4b8cd6151c826a987078d6088904823eea662f3fa",
65
+ "size_bytes": 3754322759
66
+ },
67
+ {
68
+ "path": "model-00006-of-00043.safetensors",
69
+ "sha256": "1d26e6bdf304a7d8559212914cf5457a32a77b2f100e1adba5b4856f8be167ef",
70
+ "size_bytes": 3762488035
71
+ },
72
+ {
73
+ "path": "model-00007-of-00043.safetensors",
74
+ "sha256": "79432d69a2f36ffa2c73227476e2bb7c5c4cb677fb9f5a8f150f608568b62246",
75
+ "size_bytes": 3754322781
76
+ },
77
+ {
78
+ "path": "model-00008-of-00043.safetensors",
79
+ "sha256": "674f48d0940d32b95c1fbaa154e548359d318a2ac464239aadd4351e4474338e",
80
+ "size_bytes": 3762488011
81
+ },
82
+ {
83
+ "path": "model-00009-of-00043.safetensors",
84
+ "sha256": "e5a5197573b2bee8261c6d4e21597cdb3cc365549a887b1d112b6a14db91353c",
85
+ "size_bytes": 3754322811
86
+ },
87
+ {
88
+ "path": "model-00010-of-00043.safetensors",
89
+ "sha256": "ff658c2253fb2423d7c8c7d96812746536e4d28cee7870eacd181461fe20ffc4",
90
+ "size_bytes": 3762488143
91
+ },
92
+ {
93
+ "path": "model-00011-of-00043.safetensors",
94
+ "sha256": "8bc0725b30c4d3fddf68245005db62fb48daf12a4e2b7d6d3df50b77e791e977",
95
+ "size_bytes": 3754322848
96
+ },
97
+ {
98
+ "path": "model-00012-of-00043.safetensors",
99
+ "sha256": "b41bc36d4f05b875fcacc8378fd822a6a1d75e925e3ed535177037a37627c599",
100
+ "size_bytes": 3762488124
101
+ },
102
+ {
103
+ "path": "model-00013-of-00043.safetensors",
104
+ "sha256": "5127c2225a5ff804e6555aa953952c7eb130e4c8b9f030bae451843aa81fa98f",
105
+ "size_bytes": 3754322852
106
+ },
107
+ {
108
+ "path": "model-00014-of-00043.safetensors",
109
+ "sha256": "2f8ca4f39a5f987295b8ed8ead5ca3a3465b420b023db92f1b609552d378fb3a",
110
+ "size_bytes": 3762488018
111
+ },
112
+ {
113
+ "path": "model-00015-of-00043.safetensors",
114
+ "sha256": "7698907af61989fa7fc7998f73634c896ea67a1562dcaf160b2f88ca37be1183",
115
+ "size_bytes": 3754322830
116
+ },
117
+ {
118
+ "path": "model-00016-of-00043.safetensors",
119
+ "sha256": "e11fe74ee18f31272e8e773672aa3fd224ef6750cbece508e434539d2398ea98",
120
+ "size_bytes": 3762488000
121
+ },
122
+ {
123
+ "path": "model-00017-of-00043.safetensors",
124
+ "sha256": "7b0f9385a83e985b43f7314aa00ed4c0819f3b1d7fb8beb02fbc994934370839",
125
+ "size_bytes": 3754322864
126
+ },
127
+ {
128
+ "path": "model-00018-of-00043.safetensors",
129
+ "sha256": "516e8eece65d5457d2e51cdc4a9ab91a6889d784fd0aab1f1717cab659b3f6fd",
130
+ "size_bytes": 3762488048
131
+ },
132
+ {
133
+ "path": "model-00019-of-00043.safetensors",
134
+ "sha256": "4b918d316e8de10d0ff0221e7a6b364af67c3178c11a9a2518cae4f6f1605add",
135
+ "size_bytes": 3754322836
136
+ },
137
+ {
138
+ "path": "model-00020-of-00043.safetensors",
139
+ "sha256": "4a41a0073e9b93cb4a1d023a67a8b74112f9364d82ce238f298ef026e871ffc6",
140
+ "size_bytes": 3762488044
141
+ },
142
+ {
143
+ "path": "model-00021-of-00043.safetensors",
144
+ "sha256": "b431b563d1b562d5e349a66f87efa5cb6acabf94cc7eeb79b48d1327e41b41fa",
145
+ "size_bytes": 3754322824
146
+ },
147
+ {
148
+ "path": "model-00022-of-00043.safetensors",
149
+ "sha256": "9a8cffc20003b0a466e718950325b6203dbc5aa06df38ea3ab36e59a4efc6d76",
150
+ "size_bytes": 3762488128
151
+ },
152
+ {
153
+ "path": "model-00023-of-00043.safetensors",
154
+ "sha256": "7bc9dca364a2f46e121250786d525e010115471af584fd871e8adb101a39a70c",
155
+ "size_bytes": 3754322792
156
+ },
157
+ {
158
+ "path": "model-00024-of-00043.safetensors",
159
+ "sha256": "93e89943eae8f18ff099bfd513dba11f44f4fc5ed4f4814d13cb178eaa107b60",
160
+ "size_bytes": 3762488048
161
+ },
162
+ {
163
+ "path": "model-00025-of-00043.safetensors",
164
+ "sha256": "ee48f604126102c06ab29d92ee098f588a0aac76afdb40068fdef066ea6faa1f",
165
+ "size_bytes": 3754322744
166
+ },
167
+ {
168
+ "path": "model-00026-of-00043.safetensors",
169
+ "sha256": "c20905ded3a08f2db8431a0f61e769fff8ed76f4f0d7d531d266d855e081e5f5",
170
+ "size_bytes": 3762488052
171
+ },
172
+ {
173
+ "path": "model-00027-of-00043.safetensors",
174
+ "sha256": "3942636fa9c626fb6e134bec18064fc8db288be6c141f4d1d32ba024bfc404bd",
175
+ "size_bytes": 3754322804
176
+ },
177
+ {
178
+ "path": "model-00028-of-00043.safetensors",
179
+ "sha256": "f4b91262c19519818d4f564389ac8c3f2c2534840e9e5d0f0d4c67715e88bbca",
180
+ "size_bytes": 3762488132
181
+ },
182
+ {
183
+ "path": "model-00029-of-00043.safetensors",
184
+ "sha256": "5bdfc3cca2683f82108efff3caf343be58104372d5b8fee2b50f4e540b1c2466",
185
+ "size_bytes": 3754322782
186
+ },
187
+ {
188
+ "path": "model-00030-of-00043.safetensors",
189
+ "sha256": "57eddb8006ff551cce8b24fcd0be41a6f0ca2852d4fd48b9db9f31f92984bb15",
190
+ "size_bytes": 3762488028
191
+ },
192
+ {
193
+ "path": "model-00031-of-00043.safetensors",
194
+ "sha256": "8dfdf2f7a59bce1aceebaeb2135ee741a42a74d0018e9b19a578f130b5f022c1",
195
+ "size_bytes": 3754322862
196
+ },
197
+ {
198
+ "path": "model-00032-of-00043.safetensors",
199
+ "sha256": "1f6d4f3c948fe68a069afed83471f56e1f9e7d24a494f192e3d027595a4f468f",
200
+ "size_bytes": 3762488134
201
+ },
202
+ {
203
+ "path": "model-00033-of-00043.safetensors",
204
+ "sha256": "0868eae3afd7aec958cef5781126aab2dabfae8cf843247be58d1ee51067f521",
205
+ "size_bytes": 3754322784
206
+ },
207
+ {
208
+ "path": "model-00034-of-00043.safetensors",
209
+ "sha256": "5764abeebddd52d4f4d2ed7a8f481c05b5429a68e8c6b99bcf150071da456cc7",
210
+ "size_bytes": 3762488088
211
+ },
212
+ {
213
+ "path": "model-00035-of-00043.safetensors",
214
+ "sha256": "19ab452e4caec7a8185d613c2d2603a90215c091cc34cdfe7a7c993433251191",
215
+ "size_bytes": 3754322844
216
+ },
217
+ {
218
+ "path": "model-00036-of-00043.safetensors",
219
+ "sha256": "5033ac63eb6a6f12f89adcc3630b6c6a0d29ef75740f6fe04b87aa9cae23e6ac",
220
+ "size_bytes": 3762488076
221
+ },
222
+ {
223
+ "path": "model-00037-of-00043.safetensors",
224
+ "sha256": "4d3a14ba390d7f72fb1450120aa2214f0f66995cf8c5d158f245aa013f48d595",
225
+ "size_bytes": 3754322900
226
+ },
227
+ {
228
+ "path": "model-00038-of-00043.safetensors",
229
+ "sha256": "edc8c7af75fc10ad60586cfd8b3c3772b48ac4c153489f340e7ee0164a277623",
230
+ "size_bytes": 3762504476
231
+ },
232
+ {
233
+ "path": "model-00039-of-00043.safetensors",
234
+ "sha256": "284b99aa85fcf9d2ea04735318ff5b60b4898e4b0fc3551ab241e7e73ac52a34",
235
+ "size_bytes": 3754322854
236
+ },
237
+ {
238
+ "path": "model-00040-of-00043.safetensors",
239
+ "sha256": "d44157863f4854c8a13314af03e63008859cba6c663c9b5e7d4ca57b5b23de59",
240
+ "size_bytes": 3763290916
241
+ },
242
+ {
243
+ "path": "model-00041-of-00043.safetensors",
244
+ "sha256": "ac7219977429ab499f88684a14a2b2244ce1684d15f54f70eaa772bc6fa5e421",
245
+ "size_bytes": 3754978186
246
+ },
247
+ {
248
+ "path": "model-00042-of-00043.safetensors",
249
+ "sha256": "131e1c4316448e437fb2e0fad704233301986cca111965f66dcdb5c6c957582a",
250
+ "size_bytes": 3768271634
251
+ },
252
+ {
253
+ "path": "model-00043-of-00043.safetensors",
254
+ "sha256": "0edae881ae9221b8552d9cd79dcc6b31fc3ec682671b2c1e9fe51f2bd652e591",
255
+ "size_bytes": 4700532087
256
+ },
257
+ {
258
+ "path": "model.safetensors.index.json",
259
+ "sha256": "7e2c5b001b61d263b838b84ff2514fb72633ba39f200dafef83354ee806eea5f",
260
+ "size_bytes": 209031
261
+ },
262
+ {
263
+ "path": "mtp.safetensors",
264
+ "sha256": "445e2f72d140a344ab3429b0a073ba2dc3f4198b40b0b7fe7ccc5bc41531806c",
265
+ "size_bytes": 3593958116
266
+ },
267
+ {
268
+ "path": "mtplx_runtime.json",
269
+ "sha256": "119b930ed4b470a08008aa2a51b4a83162907a85b8455f559cffa03e33236ed6",
270
+ "size_bytes": 108
271
+ },
272
+ {
273
+ "path": "tokenizer.json",
274
+ "sha256": "8421075aceec0741baf15179b6c337f443f9029e53ea88d8c1e909dd0d69b70b",
275
+ "size_bytes": 10071664
276
+ },
277
+ {
278
+ "path": "tokenizer_config.json",
279
+ "sha256": "21aa165d554e2c2f252a49730b71962edc15c36bbfa4b090ab7bd3340a2fa549",
280
+ "size_bytes": 397
281
+ }
282
+ ],
283
+ "format": "mlx",
284
+ "logical_parameters": 290944616402,
285
+ "main_logical_parameters": 284334567511,
286
+ "main_weight_file_size_bytes": 163240627259,
287
+ "measured_main_bpw": 4.592916821559088,
288
+ "measured_total_bpw": 4.587390890766196,
289
+ "mtp_acceptance_retention": null,
290
+ "mtp_distribution": {
291
+ "bf16": {
292
+ "fraction": 1.0,
293
+ "parameters": 6610048891
294
+ }
295
+ },
296
+ "mtp_measured_speedup": null,
297
+ "mtp_policy": {
298
+ "candidate_bits": [
299
+ 8,
300
+ 16
301
+ ],
302
+ "min_bits": 8,
303
+ "mode": "protected",
304
+ "optimize_for_acceptance": true,
305
+ "preserve_external_sidecar": true,
306
+ "protect_norms": true,
307
+ "protect_output_head": true
308
+ },
309
+ "mtp_present": true,
310
+ "mtp_weight_file_size_bytes": 3593958116,
311
+ "plan_sha256": "e32b61ad057fa2d9d3cd8b1c37765c1a018c38b29a0cc1f240f617ded8fbd678",
312
+ "profile": "general",
313
+ "protected_weight_file_size_bytes": 0,
314
+ "quantizer": "axquant",
315
+ "runtime": {
316
+ "ax_engine": {
317
+ "decode_kernel": null,
318
+ "fused_mtp": null,
319
+ "kernel_evidence": "unmeasured",
320
+ "model_manifest": "model-manifest.json",
321
+ "preferred_group_size": 32
322
+ },
323
+ "compatible_runtimes": [
324
+ {
325
+ "compatibility_level": "B",
326
+ "manifest": "config.json",
327
+ "mtp_support": "runtime-dependent",
328
+ "name": "mlx-lm",
329
+ "notes": [
330
+ "Standard backbone inference is the compatibility target.",
331
+ "AXQuant MTP metadata may be ignored by MLX-LM."
332
+ ],
333
+ "standard_inference": true,
334
+ "standard_mlx_weights": true,
335
+ "support_level": "standard-inference"
336
+ }
337
+ ],
338
+ "created_at": "2026-08-08T10:58:22.231133Z",
339
+ "kv_cache": null,
340
+ "memory_policy": {
341
+ "kv_cache_precision": "runtime-default",
342
+ "mtp_buffers": "preallocate-when-enabled",
343
+ "prefix_cache": "runtime-managed",
344
+ "unified_memory_safety_margin": "benchmark-required"
345
+ },
346
+ "mtp": {
347
+ "acceptance_retention": null,
348
+ "detected": true,
349
+ "draft_tokens": 1,
350
+ "enabled_by_default": true,
351
+ "head_precision": null,
352
+ "measured_speedup": null,
353
+ "optimized": false,
354
+ "recommended_temperature_max": null,
355
+ "sidecar_file": "mtp.safetensors",
356
+ "verification_mode": "runtime-default"
357
+ },
358
+ "optimization_scope": "text-path",
359
+ "primary_runtime": {
360
+ "compatibility_level": "A",
361
+ "manifest": "model-manifest.json",
362
+ "mtp_support": "native",
363
+ "name": "ax-engine",
364
+ "notes": [
365
+ "Runtime claims require a passing AX Engine doctor and benchmark report."
366
+ ],
367
+ "standard_inference": true,
368
+ "standard_mlx_weights": true,
369
+ "support_level": "optimized"
370
+ },
371
+ "schema_version": "axquant.runtime.v1"
372
+ },
373
+ "schema_version": "axquant.artifact.v2",
374
+ "software_versions": {
375
+ "ax_engine": "6.11.1",
376
+ "axquant": "1.5.1",
377
+ "mlx": "0.32.0",
378
+ "mlx_lm": "0.31.3",
379
+ "pydantic": "2.13.4",
380
+ "python": "3.13.15",
381
+ "safetensors": "0.8.0"
382
+ },
383
+ "source_model": {
384
+ "architecture": "DeepseekV4ForCausalLM",
385
+ "format": "mlx",
386
+ "local_path": null,
387
+ "model_id": "deepseek-ai/DeepSeek-V4-Flash",
388
+ "revision": "60d8d70770c6776ff598c94bb586a859a38244f1"
389
+ },
390
+ "target_class": "4bit",
391
+ "weight_distribution": {
392
+ "4bit": {
393
+ "fraction": 0.9683956042915912,
394
+ "parameters": 281749487616
395
+ },
396
+ "8bit": {
397
+ "fraction": 0.0018200401387332902,
398
+ "parameters": 529530880
399
+ },
400
+ "bf16": {
401
+ "fraction": 0.029784355569675465,
402
+ "parameters": 8665597906
403
+ }
404
+ },
405
+ "weight_file_size_bytes": 166834585375
406
+ }
axquant_mtp_sidecar_manifest.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "created_at": "2026-08-08T10:58:21.496927Z",
3
+ "dtypes": [
4
+ "BF16",
5
+ "F32",
6
+ "F8_E4M3",
7
+ "F8_E8M0",
8
+ "I8"
9
+ ],
10
+ "output": {
11
+ "path": "mtp.safetensors",
12
+ "sha256": "445e2f72d140a344ab3429b0a073ba2dc3f4198b40b0b7fe7ccc5bc41531806c",
13
+ "size_bytes": 3593958116
14
+ },
15
+ "parameters": 6610048891,
16
+ "role": "mtp",
17
+ "schema_version": "axquant.protected-tensor-sidecar.v1",
18
+ "source_files": [
19
+ {
20
+ "path": "model-00046-of-00046.safetensors",
21
+ "sha256": "f58f722893a6148216a2155cee4a57fe691cea4d3b323135c433a936b932055d",
22
+ "size_bytes": 3593956092
23
+ }
24
+ ],
25
+ "source_model": {
26
+ "architecture": "DeepseekV4ForCausalLM",
27
+ "format": "mlx",
28
+ "local_path": null,
29
+ "model_id": "deepseek-ai/DeepSeek-V4-Flash",
30
+ "revision": "60d8d70770c6776ff598c94bb586a859a38244f1"
31
+ },
32
+ "tensor_count": 1575,
33
+ "tensor_names_sha256": "5004dd518bc51fa41daefec6f8f5145b78bf8a96f788a17a7af255c484dd2777"
34
+ }
axquant_plan.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f74baadfd219d0efbd1e33ba5810fa8588fbe65648dc05322b74bf728fd6500d
3
+ size 61520034
axquant_quantizer_execution.json ADDED
The diff for this file is too large to render. See raw diff
 
axquant_runtime.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ax_engine": {
3
+ "decode_kernel": null,
4
+ "fused_mtp": null,
5
+ "kernel_evidence": "unmeasured",
6
+ "model_manifest": "model-manifest.json",
7
+ "preferred_group_size": 32
8
+ },
9
+ "compatible_runtimes": [
10
+ {
11
+ "compatibility_level": "B",
12
+ "manifest": "config.json",
13
+ "mtp_support": "runtime-dependent",
14
+ "name": "mlx-lm",
15
+ "notes": [
16
+ "Standard backbone inference is the compatibility target.",
17
+ "AXQuant MTP metadata may be ignored by MLX-LM."
18
+ ],
19
+ "standard_inference": true,
20
+ "standard_mlx_weights": true,
21
+ "support_level": "standard-inference"
22
+ }
23
+ ],
24
+ "created_at": "2026-08-08T10:58:22.231133Z",
25
+ "kv_cache": null,
26
+ "memory_policy": {
27
+ "kv_cache_precision": "runtime-default",
28
+ "mtp_buffers": "preallocate-when-enabled",
29
+ "prefix_cache": "runtime-managed",
30
+ "unified_memory_safety_margin": "benchmark-required"
31
+ },
32
+ "mtp": {
33
+ "acceptance_retention": null,
34
+ "detected": true,
35
+ "draft_tokens": 1,
36
+ "enabled_by_default": true,
37
+ "head_precision": null,
38
+ "measured_speedup": null,
39
+ "optimized": false,
40
+ "recommended_temperature_max": null,
41
+ "sidecar_file": "mtp.safetensors",
42
+ "verification_mode": "runtime-default"
43
+ },
44
+ "optimization_scope": "text-path",
45
+ "primary_runtime": {
46
+ "compatibility_level": "A",
47
+ "manifest": "model-manifest.json",
48
+ "mtp_support": "native",
49
+ "name": "ax-engine",
50
+ "notes": [
51
+ "Runtime claims require a passing AX Engine doctor and benchmark report."
52
+ ],
53
+ "standard_inference": true,
54
+ "standard_mlx_weights": true,
55
+ "support_level": "optimized"
56
+ },
57
+ "schema_version": "axquant.runtime.v1"
58
+ }
config.json ADDED
The diff for this file is too large to render. See raw diff
 
generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 1,
5
+ "do_sample": true,
6
+ "temperature": 1.0,
7
+ "top_p": 1.0,
8
+ "transformers_version": "4.46.3"
9
+ }
model-00001-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:742f1f8bb2a32fedd140f45dad06152c0ed02cddc9f9622305a152836d3762e7
3
+ size 4431029612
model-00002-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e698cad9e6d6196e85e44fbe77030691c1698c031e645596cb80ca1887699d4e
3
+ size 3765589702
model-00003-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c98f53e0aae2cf2c7769503884b23ad5375045be7cf2a727225346ecd67d5c7b
3
+ size 3754322827
model-00004-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:048a8eb15988f3a00e767449e5aeb094be244136dc001e31e28534bec1fd5f96
3
+ size 3762487939
model-00005-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed5f606fa477aa88c0c678b4b8cd6151c826a987078d6088904823eea662f3fa
3
+ size 3754322759
model-00006-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d26e6bdf304a7d8559212914cf5457a32a77b2f100e1adba5b4856f8be167ef
3
+ size 3762488035
model-00007-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79432d69a2f36ffa2c73227476e2bb7c5c4cb677fb9f5a8f150f608568b62246
3
+ size 3754322781
model-00008-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:674f48d0940d32b95c1fbaa154e548359d318a2ac464239aadd4351e4474338e
3
+ size 3762488011
model-00009-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5a5197573b2bee8261c6d4e21597cdb3cc365549a887b1d112b6a14db91353c
3
+ size 3754322811
model-00010-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff658c2253fb2423d7c8c7d96812746536e4d28cee7870eacd181461fe20ffc4
3
+ size 3762488143
model-00011-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8bc0725b30c4d3fddf68245005db62fb48daf12a4e2b7d6d3df50b77e791e977
3
+ size 3754322848
model-00012-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b41bc36d4f05b875fcacc8378fd822a6a1d75e925e3ed535177037a37627c599
3
+ size 3762488124
model-00013-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5127c2225a5ff804e6555aa953952c7eb130e4c8b9f030bae451843aa81fa98f
3
+ size 3754322852
model-00014-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f8ca4f39a5f987295b8ed8ead5ca3a3465b420b023db92f1b609552d378fb3a
3
+ size 3762488018
model-00015-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7698907af61989fa7fc7998f73634c896ea67a1562dcaf160b2f88ca37be1183
3
+ size 3754322830
model-00016-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e11fe74ee18f31272e8e773672aa3fd224ef6750cbece508e434539d2398ea98
3
+ size 3762488000
model-00017-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b0f9385a83e985b43f7314aa00ed4c0819f3b1d7fb8beb02fbc994934370839
3
+ size 3754322864
model-00018-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:516e8eece65d5457d2e51cdc4a9ab91a6889d784fd0aab1f1717cab659b3f6fd
3
+ size 3762488048
model-00019-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b918d316e8de10d0ff0221e7a6b364af67c3178c11a9a2518cae4f6f1605add
3
+ size 3754322836
model-00020-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a41a0073e9b93cb4a1d023a67a8b74112f9364d82ce238f298ef026e871ffc6
3
+ size 3762488044
model-00021-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b431b563d1b562d5e349a66f87efa5cb6acabf94cc7eeb79b48d1327e41b41fa
3
+ size 3754322824
model-00022-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a8cffc20003b0a466e718950325b6203dbc5aa06df38ea3ab36e59a4efc6d76
3
+ size 3762488128
model-00023-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bc9dca364a2f46e121250786d525e010115471af584fd871e8adb101a39a70c
3
+ size 3754322792
model-00024-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93e89943eae8f18ff099bfd513dba11f44f4fc5ed4f4814d13cb178eaa107b60
3
+ size 3762488048
model-00025-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee48f604126102c06ab29d92ee098f588a0aac76afdb40068fdef066ea6faa1f
3
+ size 3754322744
model-00026-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c20905ded3a08f2db8431a0f61e769fff8ed76f4f0d7d531d266d855e081e5f5
3
+ size 3762488052
model-00027-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3942636fa9c626fb6e134bec18064fc8db288be6c141f4d1d32ba024bfc404bd
3
+ size 3754322804
model-00028-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4b91262c19519818d4f564389ac8c3f2c2534840e9e5d0f0d4c67715e88bbca
3
+ size 3762488132
model-00029-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5bdfc3cca2683f82108efff3caf343be58104372d5b8fee2b50f4e540b1c2466
3
+ size 3754322782
model-00030-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57eddb8006ff551cce8b24fcd0be41a6f0ca2852d4fd48b9db9f31f92984bb15
3
+ size 3762488028
model-00031-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dfdf2f7a59bce1aceebaeb2135ee741a42a74d0018e9b19a578f130b5f022c1
3
+ size 3754322862
model-00032-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f6d4f3c948fe68a069afed83471f56e1f9e7d24a494f192e3d027595a4f468f
3
+ size 3762488134
model-00033-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0868eae3afd7aec958cef5781126aab2dabfae8cf843247be58d1ee51067f521
3
+ size 3754322784
model-00034-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5764abeebddd52d4f4d2ed7a8f481c05b5429a68e8c6b99bcf150071da456cc7
3
+ size 3762488088
model-00035-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19ab452e4caec7a8185d613c2d2603a90215c091cc34cdfe7a7c993433251191
3
+ size 3754322844
model-00036-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5033ac63eb6a6f12f89adcc3630b6c6a0d29ef75740f6fe04b87aa9cae23e6ac
3
+ size 3762488076
model-00037-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d3a14ba390d7f72fb1450120aa2214f0f66995cf8c5d158f245aa013f48d595
3
+ size 3754322900
model-00038-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edc8c7af75fc10ad60586cfd8b3c3772b48ac4c153489f340e7ee0164a277623
3
+ size 3762504476
model-00039-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:284b99aa85fcf9d2ea04735318ff5b60b4898e4b0fc3551ab241e7e73ac52a34
3
+ size 3754322854
model-00040-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d44157863f4854c8a13314af03e63008859cba6c663c9b5e7d4ca57b5b23de59
3
+ size 3763290916
model-00041-of-00043.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac7219977429ab499f88684a14a2b2244ce1684d15f54f70eaa772bc6fa5e421
3
+ size 3754978186