WaveCut commited on
Commit
39ef596
·
verified ·
1 Parent(s): 555a7e8

Requantize W2A4 with mixed-width low-bit protection

Browse files
README.md CHANGED
@@ -19,20 +19,9 @@ OrbitQuant is a calibration-free post-training quantization method for image and
19
  Install OrbitQuant and the Hugging Face runtime dependencies:
20
 
21
  ```bash
22
- pip install "orbitquant[hf,kernels]>=0.6.0"
23
  ```
24
 
25
- The optimized native kernel package is provisioned automatically at first
26
- model load: OrbitQuant derives the runtime variant (torch minor and CUDA
27
- version for CUDA, the torch stable ABI for CPU, plus OS and architecture),
28
- downloads the matching prebuilt wheel from the
29
- [OrbitQuant kernels release](https://github.com/iamwavecut/OrbitQuant/releases/tag/kernels-v1)
30
- with checksum verification, and caches it under `~/.cache/orbitquant/kernels`.
31
- Without a matching variant, CUDA inference falls back to the Triton packed
32
- path. Run `orbitquant kernels-install` to provision explicitly. For ComfyUI,
33
- the [ComfyUI-OrbitQuant](https://github.com/iamwavecut/ComfyUI-OrbitQuant)
34
- node pack provisions the same kernels from its install hook.
35
-
36
  Download this model repository as an OrbitQuant artifact, then load the source Diffusers pipeline with the quantized component patched in:
37
 
38
  ```python
@@ -158,6 +147,7 @@ The matrix uses all ten prompts from `image_visual_v2` at `1024x1024`. BF16 and
158
 
159
  - Method: `orbitquant`
160
  - Bits: `W2A4`
 
161
  - Runtime mode: `auto_fused`
162
  - Activation kernel backend: `auto`
163
  - Activation normalization epsilon: `1e-10`
 
19
  Install OrbitQuant and the Hugging Face runtime dependencies:
20
 
21
  ```bash
22
+ pip install "orbitquant[hf,kernels]>=0.4.0"
23
  ```
24
 
 
 
 
 
 
 
 
 
 
 
 
25
  Download this model repository as an OrbitQuant artifact, then load the source Diffusers pipeline with the quantized component patched in:
26
 
27
  ```python
 
147
 
148
  - Method: `orbitquant`
149
  - Bits: `W2A4`
150
+ - Module weight-width counts: `W2: 44, W3: 110, W4: 84`
151
  - Runtime mode: `auto_fused`
152
  - Activation kernel backend: `auto`
153
  - Activation normalization epsilon: `1e-10`
SHA256SUMS CHANGED
@@ -1,10 +1,10 @@
1
- 2b45a3bb9875d629aaaa4c5607aff6e41895fd879fd61a0f878453ec3e32cd98 README.md
2
- b92d0cacf0d8c4a7224dbde6bc3736d3dac2fbf231d465f8b992de6c7de16179 assets/image_generation_comparison_matrix.webp
3
- 3637c05001dfc37110277c582656a29310a0ceca8667007229c5f05a9fa0ab32 benchmark/summary.json
4
- aa8e04810221025dc262f66632975bdbf8466b341ad19db61ba4ad4f174f402f model.safetensors
5
- 33663f26f6c41269f71b4300ba1e92d57e426e52128854b079c25fec0e9080c3 model_index.json
6
- 9ee09b3fcc9615269db665df2a76b9a94099aea67fb5c9ce1523f5c53f0e7723 orbitquant_codebooks.safetensors
7
- f34306825b4e8a698cacfd5903648b21da8ac3d46182b9ae335095fca6305d56 orbitquant_manifest.json
8
  e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825 orbitquant_rotations.safetensors
9
- 0aee42e64abd66ff9293ec140df94e62bf7836eb96a21528a728282a0278c6ed prompts.json
10
- 6cac0dabb20f4fa9cceda63aa86ba4ca5df66902aeec52b98ab133f0c381bc71 quantization_config.json
 
1
+ 277f70f35835e37957a97fb915af255be5361340450f9a7310a093eff5633d8b README.md
2
+ 2ed2129a41b99b6c0f7f1222c590c97a34e5bccac1a2736b360c21dae22fe45e assets/image_generation_comparison_matrix.webp
3
+ 2268eba92998556016340e097a7113ecb2b249f3690cf4621f67963cae7c4380 benchmark/summary.json
4
+ 667324a633b4e4d92931ca523943c093da46c24472d3f7dc315782ca57847585 model.safetensors
5
+ a954b766fd9af50a2bcb5326e28b48f1f449b21eb2c9bafe21e9ce287192d15e model_index.json
6
+ e7737934089114db3107ba8c16e8c6a882f259fc97d02352cf5f4fb1b0656fe9 orbitquant_codebooks.safetensors
7
+ 2c4356aa0a97b7988049d4883b8efa18a44b5aba95a80c0c04090fc72e6f7d7d orbitquant_manifest.json
8
  e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825 orbitquant_rotations.safetensors
9
+ f48aca6d4d681d1ed10dcce8ad43fedb965b4a6a4b11003eb5e37ea53e4eb526 prompts.json
10
+ b80a7d0bbb5779a442d0ac5d6a318caf1a4dc76f76063c2acbd05b1d5bb06cc0 quantization_config.json
assets/image_generation_comparison_matrix.webp CHANGED

Git LFS Details

  • SHA256: b92d0cacf0d8c4a7224dbde6bc3736d3dac2fbf231d465f8b992de6c7de16179
  • Pointer size: 133 Bytes
  • Size of remote file: 23.8 MB

Git LFS Details

  • SHA256: 2ed2129a41b99b6c0f7f1222c590c97a34e5bccac1a2736b360c21dae22fe45e
  • Pointer size: 132 Bytes
  • Size of remote file: 5.5 MB
benchmark/summary.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "status": "native_validation_complete",
3
  "source_model_id": "Tongyi-MAI/Z-Image-Turbo",
4
  "source_revision": "f332072aa78be7aecdf3ee76d5c247082da564a6",
5
  "weight_bits": 2,
@@ -11,59 +11,75 @@
11
  "activation_eps": 1e-10,
12
  "quantization_device": "cuda",
13
  "weight_quantization_backend": "triton_cuda",
14
- "quantization_staging_mode": "streaming",
15
  "synchronize_per_module": false,
16
- "quantization_elapsed_seconds": 4.982273037079722,
17
- "orbitquant_seconds": 3.547152883838862,
18
- "adaln_seconds": 0.024637432768940926,
19
- "device_transfer_seconds": 1.3974662092514336,
20
- "module_device_transfer_count": 270,
21
- "metrics": {},
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
  "native_smoke": {
23
  "proof_format": "orbitquant-native-smoke-v1",
24
- "proof_source": "native_visual_pack_v2",
25
  "comparison_asset_path": "assets/image_generation_comparison_matrix.webp",
26
  "paired_prompt_seed_count": 10,
27
  "paired_prompt_seed_keys": [
28
  [
29
  "z-image-native",
30
  "0",
31
- "simple-object"
32
  ],
33
  [
34
  "z-image-native",
35
  "0",
36
- "two-object-composition"
37
  ],
38
  [
39
  "z-image-native",
40
  "0",
41
- "counting"
42
  ],
43
  [
44
  "z-image-native",
45
  "0",
46
- "color-binding"
47
  ],
48
  [
49
  "z-image-native",
50
  "0",
51
- "spatial-relationship"
52
  ],
53
  [
54
  "z-image-native",
55
  "0",
56
- "long-prompt"
57
  ],
58
  [
59
  "z-image-native",
60
  "0",
61
- "english-text-rendering"
62
  ],
63
  [
64
  "z-image-native",
65
  "0",
66
- "cyrillic-text-rendering"
67
  ],
68
  [
69
  "z-image-native",
@@ -73,7 +89,7 @@
73
  [
74
  "z-image-native",
75
  "0",
76
- "occlusion-reflection"
77
  ]
78
  ],
79
  "splits": {
@@ -86,57 +102,57 @@
86
  "0"
87
  ],
88
  "prompt_ids": [
89
- "simple-object",
90
- "two-object-composition",
91
- "counting",
92
  "color-binding",
93
- "spatial-relationship",
94
- "long-prompt",
95
- "english-text-rendering",
96
  "cyrillic-text-rendering",
 
 
 
 
 
97
  "style-heavy",
98
- "occlusion-reflection"
99
  ],
100
  "pair_keys": [
101
  [
102
  "z-image-native",
103
  "0",
104
- "simple-object"
105
  ],
106
  [
107
  "z-image-native",
108
  "0",
109
- "two-object-composition"
110
  ],
111
  [
112
  "z-image-native",
113
  "0",
114
- "counting"
115
  ],
116
  [
117
  "z-image-native",
118
  "0",
119
- "color-binding"
120
  ],
121
  [
122
  "z-image-native",
123
  "0",
124
- "spatial-relationship"
125
  ],
126
  [
127
  "z-image-native",
128
  "0",
129
- "long-prompt"
130
  ],
131
  [
132
  "z-image-native",
133
  "0",
134
- "english-text-rendering"
135
  ],
136
  [
137
  "z-image-native",
138
  "0",
139
- "cyrillic-text-rendering"
140
  ],
141
  [
142
  "z-image-native",
@@ -146,7 +162,7 @@
146
  [
147
  "z-image-native",
148
  "0",
149
- "occlusion-reflection"
150
  ]
151
  ],
152
  "native_settings": [
@@ -169,57 +185,57 @@
169
  "0"
170
  ],
171
  "prompt_ids": [
172
- "simple-object",
173
- "two-object-composition",
174
- "counting",
175
  "color-binding",
176
- "spatial-relationship",
177
- "long-prompt",
178
- "english-text-rendering",
179
  "cyrillic-text-rendering",
 
 
 
 
 
180
  "style-heavy",
181
- "occlusion-reflection"
182
  ],
183
  "pair_keys": [
184
  [
185
  "z-image-native",
186
  "0",
187
- "simple-object"
188
  ],
189
  [
190
  "z-image-native",
191
  "0",
192
- "two-object-composition"
193
  ],
194
  [
195
  "z-image-native",
196
  "0",
197
- "counting"
198
  ],
199
  [
200
  "z-image-native",
201
  "0",
202
- "color-binding"
203
  ],
204
  [
205
  "z-image-native",
206
  "0",
207
- "spatial-relationship"
208
  ],
209
  [
210
  "z-image-native",
211
  "0",
212
- "long-prompt"
213
  ],
214
  [
215
  "z-image-native",
216
  "0",
217
- "english-text-rendering"
218
  ],
219
  [
220
  "z-image-native",
221
  "0",
222
- "cyrillic-text-rendering"
223
  ],
224
  [
225
  "z-image-native",
@@ -229,7 +245,7 @@
229
  [
230
  "z-image-native",
231
  "0",
232
- "occlusion-reflection"
233
  ]
234
  ],
235
  "native_settings": [
@@ -246,6 +262,5 @@
246
  }
247
  },
248
  "published_summary": "compact",
249
- "raw_generation_records": "local-only",
250
- "comparison_prompt_pack": "image_visual_v2"
251
  }
 
1
  {
2
+ "status": "metrics_recorded",
3
  "source_model_id": "Tongyi-MAI/Z-Image-Turbo",
4
  "source_revision": "f332072aa78be7aecdf3ee76d5c247082da564a6",
5
  "weight_bits": 2,
 
11
  "activation_eps": 1e-10,
12
  "quantization_device": "cuda",
13
  "weight_quantization_backend": "triton_cuda",
14
+ "quantization_staging_mode": "component",
15
  "synchronize_per_module": false,
16
+ "quantization_elapsed_seconds": 4.950456083752215,
17
+ "orbitquant_seconds": 3.266821008641273,
18
+ "adaln_seconds": 0.28703642217442393,
19
+ "device_transfer_seconds": 1.381439802236855,
20
+ "module_device_transfer_count": 1,
21
+ "metrics": {
22
+ "original": {
23
+ "records": 10,
24
+ "latest_metrics": {
25
+ "generated_samples": 1,
26
+ "wall_time_seconds": 6.092661592178047,
27
+ "peak_vram_bytes": 23267429888
28
+ }
29
+ },
30
+ "orbitquant": {
31
+ "records": 10,
32
+ "latest_metrics": {
33
+ "generated_samples": 1,
34
+ "wall_time_seconds": 6.2424984918907285,
35
+ "peak_vram_bytes": 13385301504
36
+ }
37
+ }
38
+ },
39
  "native_smoke": {
40
  "proof_format": "orbitquant-native-smoke-v1",
 
41
  "comparison_asset_path": "assets/image_generation_comparison_matrix.webp",
42
  "paired_prompt_seed_count": 10,
43
  "paired_prompt_seed_keys": [
44
  [
45
  "z-image-native",
46
  "0",
47
+ "color-binding"
48
  ],
49
  [
50
  "z-image-native",
51
  "0",
52
+ "counting"
53
  ],
54
  [
55
  "z-image-native",
56
  "0",
57
+ "cyrillic-text-rendering"
58
  ],
59
  [
60
  "z-image-native",
61
  "0",
62
+ "english-text-rendering"
63
  ],
64
  [
65
  "z-image-native",
66
  "0",
67
+ "long-prompt"
68
  ],
69
  [
70
  "z-image-native",
71
  "0",
72
+ "occlusion-reflection"
73
  ],
74
  [
75
  "z-image-native",
76
  "0",
77
+ "simple-object"
78
  ],
79
  [
80
  "z-image-native",
81
  "0",
82
+ "spatial-relationship"
83
  ],
84
  [
85
  "z-image-native",
 
89
  [
90
  "z-image-native",
91
  "0",
92
+ "two-object-composition"
93
  ]
94
  ],
95
  "splits": {
 
102
  "0"
103
  ],
104
  "prompt_ids": [
 
 
 
105
  "color-binding",
106
+ "counting",
 
 
107
  "cyrillic-text-rendering",
108
+ "english-text-rendering",
109
+ "long-prompt",
110
+ "occlusion-reflection",
111
+ "simple-object",
112
+ "spatial-relationship",
113
  "style-heavy",
114
+ "two-object-composition"
115
  ],
116
  "pair_keys": [
117
  [
118
  "z-image-native",
119
  "0",
120
+ "color-binding"
121
  ],
122
  [
123
  "z-image-native",
124
  "0",
125
+ "counting"
126
  ],
127
  [
128
  "z-image-native",
129
  "0",
130
+ "cyrillic-text-rendering"
131
  ],
132
  [
133
  "z-image-native",
134
  "0",
135
+ "english-text-rendering"
136
  ],
137
  [
138
  "z-image-native",
139
  "0",
140
+ "long-prompt"
141
  ],
142
  [
143
  "z-image-native",
144
  "0",
145
+ "occlusion-reflection"
146
  ],
147
  [
148
  "z-image-native",
149
  "0",
150
+ "simple-object"
151
  ],
152
  [
153
  "z-image-native",
154
  "0",
155
+ "spatial-relationship"
156
  ],
157
  [
158
  "z-image-native",
 
162
  [
163
  "z-image-native",
164
  "0",
165
+ "two-object-composition"
166
  ]
167
  ],
168
  "native_settings": [
 
185
  "0"
186
  ],
187
  "prompt_ids": [
 
 
 
188
  "color-binding",
189
+ "counting",
 
 
190
  "cyrillic-text-rendering",
191
+ "english-text-rendering",
192
+ "long-prompt",
193
+ "occlusion-reflection",
194
+ "simple-object",
195
+ "spatial-relationship",
196
  "style-heavy",
197
+ "two-object-composition"
198
  ],
199
  "pair_keys": [
200
  [
201
  "z-image-native",
202
  "0",
203
+ "color-binding"
204
  ],
205
  [
206
  "z-image-native",
207
  "0",
208
+ "counting"
209
  ],
210
  [
211
  "z-image-native",
212
  "0",
213
+ "cyrillic-text-rendering"
214
  ],
215
  [
216
  "z-image-native",
217
  "0",
218
+ "english-text-rendering"
219
  ],
220
  [
221
  "z-image-native",
222
  "0",
223
+ "long-prompt"
224
  ],
225
  [
226
  "z-image-native",
227
  "0",
228
+ "occlusion-reflection"
229
  ],
230
  [
231
  "z-image-native",
232
  "0",
233
+ "simple-object"
234
  ],
235
  [
236
  "z-image-native",
237
  "0",
238
+ "spatial-relationship"
239
  ],
240
  [
241
  "z-image-native",
 
245
  [
246
  "z-image-native",
247
  "0",
248
+ "two-object-composition"
249
  ]
250
  ],
251
  "native_settings": [
 
262
  }
263
  },
264
  "published_summary": "compact",
265
+ "raw_generation_records": "local-only"
 
266
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:aa8e04810221025dc262f66632975bdbf8466b341ad19db61ba4ad4f174f402f
3
- size 1599432256
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:667324a633b4e4d92931ca523943c093da46c24472d3f7dc315782ca57847585
3
+ size 2535778088
model_index.json CHANGED
@@ -20,5 +20,5 @@
20
  "activation_eps": 1e-10,
21
  "quantization_device": "cuda",
22
  "weight_quantization_backend": "triton_cuda",
23
- "quantization_staging_mode": "streaming"
24
  }
 
20
  "activation_eps": 1e-10,
21
  "quantization_device": "cuda",
22
  "weight_quantization_backend": "triton_cuda",
23
+ "quantization_staging_mode": "component"
24
  }
orbitquant_codebooks.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9ee09b3fcc9615269db665df2a76b9a94099aea67fb5c9ce1523f5c53f0e7723
3
- size 952
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7737934089114db3107ba8c16e8c6a882f259fc97d02352cf5f4fb1b0656fe9
3
+ size 1204
orbitquant_manifest.json CHANGED
@@ -20,7 +20,7 @@
20
  "adaln_group_size": 64,
21
  "quantization_device": "cuda",
22
  "weight_quantization_backend": "triton_cuda",
23
- "quantization_staging_mode": "streaming",
24
  "target_policy": "z_image",
25
  "adaln_policy": "int4_rtn_group64_bf16_activation",
26
  "quantized_modules": [
@@ -342,43 +342,43 @@
342
  128
343
  ],
344
  "noise_refiner.0.attention.to_q.packed_weight_indices": [
345
- 3686400
346
  ],
347
  "noise_refiner.0.attention.to_q.row_norms": [
348
  3840
349
  ],
350
  "noise_refiner.0.attention.to_k.packed_weight_indices": [
351
- 3686400
352
  ],
353
  "noise_refiner.0.attention.to_k.row_norms": [
354
  3840
355
  ],
356
  "noise_refiner.0.attention.to_v.packed_weight_indices": [
357
- 3686400
358
  ],
359
  "noise_refiner.0.attention.to_v.row_norms": [
360
  3840
361
  ],
362
  "noise_refiner.0.attention.to_out.0.packed_weight_indices": [
363
- 3686400
364
  ],
365
  "noise_refiner.0.attention.to_out.0.row_norms": [
366
  3840
367
  ],
368
  "noise_refiner.0.feed_forward.w1.packed_weight_indices": [
369
- 9830400
370
  ],
371
  "noise_refiner.0.feed_forward.w1.row_norms": [
372
  10240
373
  ],
374
  "noise_refiner.0.feed_forward.w2.packed_weight_indices": [
375
- 9830400
376
  ],
377
  "noise_refiner.0.feed_forward.w2.row_norms": [
378
  3840
379
  ],
380
  "noise_refiner.0.feed_forward.w3.packed_weight_indices": [
381
- 9830400
382
  ],
383
  "noise_refiner.0.feed_forward.w3.row_norms": [
384
  10240
@@ -395,9 +395,6 @@
395
  "noise_refiner.0.ffn_norm2.weight": [
396
  3840
397
  ],
398
- "noise_refiner.0.adaLN_modulation.0.bias": [
399
- 15360
400
- ],
401
  "noise_refiner.0.adaLN_modulation.0.packed_weight": [
402
  1966080
403
  ],
@@ -405,6 +402,9 @@
405
  15360,
406
  4
407
  ],
 
 
 
408
  "noise_refiner.1.attention.norm_q.weight": [
409
  128
410
  ],
@@ -412,43 +412,43 @@
412
  128
413
  ],
414
  "noise_refiner.1.attention.to_q.packed_weight_indices": [
415
- 3686400
416
  ],
417
  "noise_refiner.1.attention.to_q.row_norms": [
418
  3840
419
  ],
420
  "noise_refiner.1.attention.to_k.packed_weight_indices": [
421
- 3686400
422
  ],
423
  "noise_refiner.1.attention.to_k.row_norms": [
424
  3840
425
  ],
426
  "noise_refiner.1.attention.to_v.packed_weight_indices": [
427
- 3686400
428
  ],
429
  "noise_refiner.1.attention.to_v.row_norms": [
430
  3840
431
  ],
432
  "noise_refiner.1.attention.to_out.0.packed_weight_indices": [
433
- 3686400
434
  ],
435
  "noise_refiner.1.attention.to_out.0.row_norms": [
436
  3840
437
  ],
438
  "noise_refiner.1.feed_forward.w1.packed_weight_indices": [
439
- 9830400
440
  ],
441
  "noise_refiner.1.feed_forward.w1.row_norms": [
442
  10240
443
  ],
444
  "noise_refiner.1.feed_forward.w2.packed_weight_indices": [
445
- 9830400
446
  ],
447
  "noise_refiner.1.feed_forward.w2.row_norms": [
448
  3840
449
  ],
450
  "noise_refiner.1.feed_forward.w3.packed_weight_indices": [
451
- 9830400
452
  ],
453
  "noise_refiner.1.feed_forward.w3.row_norms": [
454
  10240
@@ -465,9 +465,6 @@
465
  "noise_refiner.1.ffn_norm2.weight": [
466
  3840
467
  ],
468
- "noise_refiner.1.adaLN_modulation.0.bias": [
469
- 15360
470
- ],
471
  "noise_refiner.1.adaLN_modulation.0.packed_weight": [
472
  1966080
473
  ],
@@ -475,6 +472,9 @@
475
  15360,
476
  4
477
  ],
 
 
 
478
  "context_refiner.0.attention.norm_q.weight": [
479
  128
480
  ],
@@ -482,43 +482,43 @@
482
  128
483
  ],
484
  "context_refiner.0.attention.to_q.packed_weight_indices": [
485
- 3686400
486
  ],
487
  "context_refiner.0.attention.to_q.row_norms": [
488
  3840
489
  ],
490
  "context_refiner.0.attention.to_k.packed_weight_indices": [
491
- 3686400
492
  ],
493
  "context_refiner.0.attention.to_k.row_norms": [
494
  3840
495
  ],
496
  "context_refiner.0.attention.to_v.packed_weight_indices": [
497
- 3686400
498
  ],
499
  "context_refiner.0.attention.to_v.row_norms": [
500
  3840
501
  ],
502
  "context_refiner.0.attention.to_out.0.packed_weight_indices": [
503
- 3686400
504
  ],
505
  "context_refiner.0.attention.to_out.0.row_norms": [
506
  3840
507
  ],
508
  "context_refiner.0.feed_forward.w1.packed_weight_indices": [
509
- 9830400
510
  ],
511
  "context_refiner.0.feed_forward.w1.row_norms": [
512
  10240
513
  ],
514
  "context_refiner.0.feed_forward.w2.packed_weight_indices": [
515
- 9830400
516
  ],
517
  "context_refiner.0.feed_forward.w2.row_norms": [
518
  3840
519
  ],
520
  "context_refiner.0.feed_forward.w3.packed_weight_indices": [
521
- 9830400
522
  ],
523
  "context_refiner.0.feed_forward.w3.row_norms": [
524
  10240
@@ -542,43 +542,43 @@
542
  128
543
  ],
544
  "context_refiner.1.attention.to_q.packed_weight_indices": [
545
- 3686400
546
  ],
547
  "context_refiner.1.attention.to_q.row_norms": [
548
  3840
549
  ],
550
  "context_refiner.1.attention.to_k.packed_weight_indices": [
551
- 3686400
552
  ],
553
  "context_refiner.1.attention.to_k.row_norms": [
554
  3840
555
  ],
556
  "context_refiner.1.attention.to_v.packed_weight_indices": [
557
- 3686400
558
  ],
559
  "context_refiner.1.attention.to_v.row_norms": [
560
  3840
561
  ],
562
  "context_refiner.1.attention.to_out.0.packed_weight_indices": [
563
- 3686400
564
  ],
565
  "context_refiner.1.attention.to_out.0.row_norms": [
566
  3840
567
  ],
568
  "context_refiner.1.feed_forward.w1.packed_weight_indices": [
569
- 9830400
570
  ],
571
  "context_refiner.1.feed_forward.w1.row_norms": [
572
  10240
573
  ],
574
  "context_refiner.1.feed_forward.w2.packed_weight_indices": [
575
- 9830400
576
  ],
577
  "context_refiner.1.feed_forward.w2.row_norms": [
578
  3840
579
  ],
580
  "context_refiner.1.feed_forward.w3.packed_weight_indices": [
581
- 9830400
582
  ],
583
  "context_refiner.1.feed_forward.w3.row_norms": [
584
  10240
@@ -626,43 +626,43 @@
626
  128
627
  ],
628
  "layers.0.attention.to_q.packed_weight_indices": [
629
- 3686400
630
  ],
631
  "layers.0.attention.to_q.row_norms": [
632
  3840
633
  ],
634
  "layers.0.attention.to_k.packed_weight_indices": [
635
- 3686400
636
  ],
637
  "layers.0.attention.to_k.row_norms": [
638
  3840
639
  ],
640
  "layers.0.attention.to_v.packed_weight_indices": [
641
- 3686400
642
  ],
643
  "layers.0.attention.to_v.row_norms": [
644
  3840
645
  ],
646
  "layers.0.attention.to_out.0.packed_weight_indices": [
647
- 3686400
648
  ],
649
  "layers.0.attention.to_out.0.row_norms": [
650
  3840
651
  ],
652
  "layers.0.feed_forward.w1.packed_weight_indices": [
653
- 9830400
654
  ],
655
  "layers.0.feed_forward.w1.row_norms": [
656
  10240
657
  ],
658
  "layers.0.feed_forward.w2.packed_weight_indices": [
659
- 9830400
660
  ],
661
  "layers.0.feed_forward.w2.row_norms": [
662
  3840
663
  ],
664
  "layers.0.feed_forward.w3.packed_weight_indices": [
665
- 9830400
666
  ],
667
  "layers.0.feed_forward.w3.row_norms": [
668
  10240
@@ -679,9 +679,6 @@
679
  "layers.0.ffn_norm2.weight": [
680
  3840
681
  ],
682
- "layers.0.adaLN_modulation.0.bias": [
683
- 15360
684
- ],
685
  "layers.0.adaLN_modulation.0.packed_weight": [
686
  1966080
687
  ],
@@ -689,6 +686,9 @@
689
  15360,
690
  4
691
  ],
 
 
 
692
  "layers.1.attention.norm_q.weight": [
693
  128
694
  ],
@@ -696,43 +696,43 @@
696
  128
697
  ],
698
  "layers.1.attention.to_q.packed_weight_indices": [
699
- 3686400
700
  ],
701
  "layers.1.attention.to_q.row_norms": [
702
  3840
703
  ],
704
  "layers.1.attention.to_k.packed_weight_indices": [
705
- 3686400
706
  ],
707
  "layers.1.attention.to_k.row_norms": [
708
  3840
709
  ],
710
  "layers.1.attention.to_v.packed_weight_indices": [
711
- 3686400
712
  ],
713
  "layers.1.attention.to_v.row_norms": [
714
  3840
715
  ],
716
  "layers.1.attention.to_out.0.packed_weight_indices": [
717
- 3686400
718
  ],
719
  "layers.1.attention.to_out.0.row_norms": [
720
  3840
721
  ],
722
  "layers.1.feed_forward.w1.packed_weight_indices": [
723
- 9830400
724
  ],
725
  "layers.1.feed_forward.w1.row_norms": [
726
  10240
727
  ],
728
  "layers.1.feed_forward.w2.packed_weight_indices": [
729
- 9830400
730
  ],
731
  "layers.1.feed_forward.w2.row_norms": [
732
  3840
733
  ],
734
  "layers.1.feed_forward.w3.packed_weight_indices": [
735
- 9830400
736
  ],
737
  "layers.1.feed_forward.w3.row_norms": [
738
  10240
@@ -749,9 +749,6 @@
749
  "layers.1.ffn_norm2.weight": [
750
  3840
751
  ],
752
- "layers.1.adaLN_modulation.0.bias": [
753
- 15360
754
- ],
755
  "layers.1.adaLN_modulation.0.packed_weight": [
756
  1966080
757
  ],
@@ -759,6 +756,9 @@
759
  15360,
760
  4
761
  ],
 
 
 
762
  "layers.2.attention.norm_q.weight": [
763
  128
764
  ],
@@ -766,43 +766,43 @@
766
  128
767
  ],
768
  "layers.2.attention.to_q.packed_weight_indices": [
769
- 3686400
770
  ],
771
  "layers.2.attention.to_q.row_norms": [
772
  3840
773
  ],
774
  "layers.2.attention.to_k.packed_weight_indices": [
775
- 3686400
776
  ],
777
  "layers.2.attention.to_k.row_norms": [
778
  3840
779
  ],
780
  "layers.2.attention.to_v.packed_weight_indices": [
781
- 3686400
782
  ],
783
  "layers.2.attention.to_v.row_norms": [
784
  3840
785
  ],
786
  "layers.2.attention.to_out.0.packed_weight_indices": [
787
- 3686400
788
  ],
789
  "layers.2.attention.to_out.0.row_norms": [
790
  3840
791
  ],
792
  "layers.2.feed_forward.w1.packed_weight_indices": [
793
- 9830400
794
  ],
795
  "layers.2.feed_forward.w1.row_norms": [
796
  10240
797
  ],
798
  "layers.2.feed_forward.w2.packed_weight_indices": [
799
- 9830400
800
  ],
801
  "layers.2.feed_forward.w2.row_norms": [
802
  3840
803
  ],
804
  "layers.2.feed_forward.w3.packed_weight_indices": [
805
- 9830400
806
  ],
807
  "layers.2.feed_forward.w3.row_norms": [
808
  10240
@@ -819,9 +819,6 @@
819
  "layers.2.ffn_norm2.weight": [
820
  3840
821
  ],
822
- "layers.2.adaLN_modulation.0.bias": [
823
- 15360
824
- ],
825
  "layers.2.adaLN_modulation.0.packed_weight": [
826
  1966080
827
  ],
@@ -829,6 +826,9 @@
829
  15360,
830
  4
831
  ],
 
 
 
832
  "layers.3.attention.norm_q.weight": [
833
  128
834
  ],
@@ -836,43 +836,43 @@
836
  128
837
  ],
838
  "layers.3.attention.to_q.packed_weight_indices": [
839
- 3686400
840
  ],
841
  "layers.3.attention.to_q.row_norms": [
842
  3840
843
  ],
844
  "layers.3.attention.to_k.packed_weight_indices": [
845
- 3686400
846
  ],
847
  "layers.3.attention.to_k.row_norms": [
848
  3840
849
  ],
850
  "layers.3.attention.to_v.packed_weight_indices": [
851
- 3686400
852
  ],
853
  "layers.3.attention.to_v.row_norms": [
854
  3840
855
  ],
856
  "layers.3.attention.to_out.0.packed_weight_indices": [
857
- 3686400
858
  ],
859
  "layers.3.attention.to_out.0.row_norms": [
860
  3840
861
  ],
862
  "layers.3.feed_forward.w1.packed_weight_indices": [
863
- 9830400
864
  ],
865
  "layers.3.feed_forward.w1.row_norms": [
866
  10240
867
  ],
868
  "layers.3.feed_forward.w2.packed_weight_indices": [
869
- 9830400
870
  ],
871
  "layers.3.feed_forward.w2.row_norms": [
872
  3840
873
  ],
874
  "layers.3.feed_forward.w3.packed_weight_indices": [
875
- 9830400
876
  ],
877
  "layers.3.feed_forward.w3.row_norms": [
878
  10240
@@ -889,9 +889,6 @@
889
  "layers.3.ffn_norm2.weight": [
890
  3840
891
  ],
892
- "layers.3.adaLN_modulation.0.bias": [
893
- 15360
894
- ],
895
  "layers.3.adaLN_modulation.0.packed_weight": [
896
  1966080
897
  ],
@@ -899,6 +896,9 @@
899
  15360,
900
  4
901
  ],
 
 
 
902
  "layers.4.attention.norm_q.weight": [
903
  128
904
  ],
@@ -918,31 +918,31 @@
918
  3840
919
  ],
920
  "layers.4.attention.to_v.packed_weight_indices": [
921
- 3686400
922
  ],
923
  "layers.4.attention.to_v.row_norms": [
924
  3840
925
  ],
926
  "layers.4.attention.to_out.0.packed_weight_indices": [
927
- 3686400
928
  ],
929
  "layers.4.attention.to_out.0.row_norms": [
930
  3840
931
  ],
932
  "layers.4.feed_forward.w1.packed_weight_indices": [
933
- 9830400
934
  ],
935
  "layers.4.feed_forward.w1.row_norms": [
936
  10240
937
  ],
938
  "layers.4.feed_forward.w2.packed_weight_indices": [
939
- 9830400
940
  ],
941
  "layers.4.feed_forward.w2.row_norms": [
942
  3840
943
  ],
944
  "layers.4.feed_forward.w3.packed_weight_indices": [
945
- 9830400
946
  ],
947
  "layers.4.feed_forward.w3.row_norms": [
948
  10240
@@ -959,9 +959,6 @@
959
  "layers.4.ffn_norm2.weight": [
960
  3840
961
  ],
962
- "layers.4.adaLN_modulation.0.bias": [
963
- 15360
964
- ],
965
  "layers.4.adaLN_modulation.0.packed_weight": [
966
  1966080
967
  ],
@@ -969,6 +966,9 @@
969
  15360,
970
  4
971
  ],
 
 
 
972
  "layers.5.attention.norm_q.weight": [
973
  128
974
  ],
@@ -988,31 +988,31 @@
988
  3840
989
  ],
990
  "layers.5.attention.to_v.packed_weight_indices": [
991
- 3686400
992
  ],
993
  "layers.5.attention.to_v.row_norms": [
994
  3840
995
  ],
996
  "layers.5.attention.to_out.0.packed_weight_indices": [
997
- 3686400
998
  ],
999
  "layers.5.attention.to_out.0.row_norms": [
1000
  3840
1001
  ],
1002
  "layers.5.feed_forward.w1.packed_weight_indices": [
1003
- 9830400
1004
  ],
1005
  "layers.5.feed_forward.w1.row_norms": [
1006
  10240
1007
  ],
1008
  "layers.5.feed_forward.w2.packed_weight_indices": [
1009
- 9830400
1010
  ],
1011
  "layers.5.feed_forward.w2.row_norms": [
1012
  3840
1013
  ],
1014
  "layers.5.feed_forward.w3.packed_weight_indices": [
1015
- 9830400
1016
  ],
1017
  "layers.5.feed_forward.w3.row_norms": [
1018
  10240
@@ -1029,9 +1029,6 @@
1029
  "layers.5.ffn_norm2.weight": [
1030
  3840
1031
  ],
1032
- "layers.5.adaLN_modulation.0.bias": [
1033
- 15360
1034
- ],
1035
  "layers.5.adaLN_modulation.0.packed_weight": [
1036
  1966080
1037
  ],
@@ -1039,6 +1036,9 @@
1039
  15360,
1040
  4
1041
  ],
 
 
 
1042
  "layers.6.attention.norm_q.weight": [
1043
  128
1044
  ],
@@ -1058,31 +1058,31 @@
1058
  3840
1059
  ],
1060
  "layers.6.attention.to_v.packed_weight_indices": [
1061
- 3686400
1062
  ],
1063
  "layers.6.attention.to_v.row_norms": [
1064
  3840
1065
  ],
1066
  "layers.6.attention.to_out.0.packed_weight_indices": [
1067
- 3686400
1068
  ],
1069
  "layers.6.attention.to_out.0.row_norms": [
1070
  3840
1071
  ],
1072
  "layers.6.feed_forward.w1.packed_weight_indices": [
1073
- 9830400
1074
  ],
1075
  "layers.6.feed_forward.w1.row_norms": [
1076
  10240
1077
  ],
1078
  "layers.6.feed_forward.w2.packed_weight_indices": [
1079
- 9830400
1080
  ],
1081
  "layers.6.feed_forward.w2.row_norms": [
1082
  3840
1083
  ],
1084
  "layers.6.feed_forward.w3.packed_weight_indices": [
1085
- 9830400
1086
  ],
1087
  "layers.6.feed_forward.w3.row_norms": [
1088
  10240
@@ -1099,9 +1099,6 @@
1099
  "layers.6.ffn_norm2.weight": [
1100
  3840
1101
  ],
1102
- "layers.6.adaLN_modulation.0.bias": [
1103
- 15360
1104
- ],
1105
  "layers.6.adaLN_modulation.0.packed_weight": [
1106
  1966080
1107
  ],
@@ -1109,6 +1106,9 @@
1109
  15360,
1110
  4
1111
  ],
 
 
 
1112
  "layers.7.attention.norm_q.weight": [
1113
  128
1114
  ],
@@ -1128,31 +1128,31 @@
1128
  3840
1129
  ],
1130
  "layers.7.attention.to_v.packed_weight_indices": [
1131
- 3686400
1132
  ],
1133
  "layers.7.attention.to_v.row_norms": [
1134
  3840
1135
  ],
1136
  "layers.7.attention.to_out.0.packed_weight_indices": [
1137
- 3686400
1138
  ],
1139
  "layers.7.attention.to_out.0.row_norms": [
1140
  3840
1141
  ],
1142
  "layers.7.feed_forward.w1.packed_weight_indices": [
1143
- 9830400
1144
  ],
1145
  "layers.7.feed_forward.w1.row_norms": [
1146
  10240
1147
  ],
1148
  "layers.7.feed_forward.w2.packed_weight_indices": [
1149
- 9830400
1150
  ],
1151
  "layers.7.feed_forward.w2.row_norms": [
1152
  3840
1153
  ],
1154
  "layers.7.feed_forward.w3.packed_weight_indices": [
1155
- 9830400
1156
  ],
1157
  "layers.7.feed_forward.w3.row_norms": [
1158
  10240
@@ -1169,9 +1169,6 @@
1169
  "layers.7.ffn_norm2.weight": [
1170
  3840
1171
  ],
1172
- "layers.7.adaLN_modulation.0.bias": [
1173
- 15360
1174
- ],
1175
  "layers.7.adaLN_modulation.0.packed_weight": [
1176
  1966080
1177
  ],
@@ -1179,6 +1176,9 @@
1179
  15360,
1180
  4
1181
  ],
 
 
 
1182
  "layers.8.attention.norm_q.weight": [
1183
  128
1184
  ],
@@ -1198,31 +1198,31 @@
1198
  3840
1199
  ],
1200
  "layers.8.attention.to_v.packed_weight_indices": [
1201
- 3686400
1202
  ],
1203
  "layers.8.attention.to_v.row_norms": [
1204
  3840
1205
  ],
1206
  "layers.8.attention.to_out.0.packed_weight_indices": [
1207
- 3686400
1208
  ],
1209
  "layers.8.attention.to_out.0.row_norms": [
1210
  3840
1211
  ],
1212
  "layers.8.feed_forward.w1.packed_weight_indices": [
1213
- 9830400
1214
  ],
1215
  "layers.8.feed_forward.w1.row_norms": [
1216
  10240
1217
  ],
1218
  "layers.8.feed_forward.w2.packed_weight_indices": [
1219
- 9830400
1220
  ],
1221
  "layers.8.feed_forward.w2.row_norms": [
1222
  3840
1223
  ],
1224
  "layers.8.feed_forward.w3.packed_weight_indices": [
1225
- 9830400
1226
  ],
1227
  "layers.8.feed_forward.w3.row_norms": [
1228
  10240
@@ -1239,9 +1239,6 @@
1239
  "layers.8.ffn_norm2.weight": [
1240
  3840
1241
  ],
1242
- "layers.8.adaLN_modulation.0.bias": [
1243
- 15360
1244
- ],
1245
  "layers.8.adaLN_modulation.0.packed_weight": [
1246
  1966080
1247
  ],
@@ -1249,6 +1246,9 @@
1249
  15360,
1250
  4
1251
  ],
 
 
 
1252
  "layers.9.attention.norm_q.weight": [
1253
  128
1254
  ],
@@ -1268,31 +1268,31 @@
1268
  3840
1269
  ],
1270
  "layers.9.attention.to_v.packed_weight_indices": [
1271
- 3686400
1272
  ],
1273
  "layers.9.attention.to_v.row_norms": [
1274
  3840
1275
  ],
1276
  "layers.9.attention.to_out.0.packed_weight_indices": [
1277
- 3686400
1278
  ],
1279
  "layers.9.attention.to_out.0.row_norms": [
1280
  3840
1281
  ],
1282
  "layers.9.feed_forward.w1.packed_weight_indices": [
1283
- 9830400
1284
  ],
1285
  "layers.9.feed_forward.w1.row_norms": [
1286
  10240
1287
  ],
1288
  "layers.9.feed_forward.w2.packed_weight_indices": [
1289
- 9830400
1290
  ],
1291
  "layers.9.feed_forward.w2.row_norms": [
1292
  3840
1293
  ],
1294
  "layers.9.feed_forward.w3.packed_weight_indices": [
1295
- 9830400
1296
  ],
1297
  "layers.9.feed_forward.w3.row_norms": [
1298
  10240
@@ -1309,9 +1309,6 @@
1309
  "layers.9.ffn_norm2.weight": [
1310
  3840
1311
  ],
1312
- "layers.9.adaLN_modulation.0.bias": [
1313
- 15360
1314
- ],
1315
  "layers.9.adaLN_modulation.0.packed_weight": [
1316
  1966080
1317
  ],
@@ -1319,6 +1316,9 @@
1319
  15360,
1320
  4
1321
  ],
 
 
 
1322
  "layers.10.attention.norm_q.weight": [
1323
  128
1324
  ],
@@ -1338,31 +1338,31 @@
1338
  3840
1339
  ],
1340
  "layers.10.attention.to_v.packed_weight_indices": [
1341
- 3686400
1342
  ],
1343
  "layers.10.attention.to_v.row_norms": [
1344
  3840
1345
  ],
1346
  "layers.10.attention.to_out.0.packed_weight_indices": [
1347
- 3686400
1348
  ],
1349
  "layers.10.attention.to_out.0.row_norms": [
1350
  3840
1351
  ],
1352
  "layers.10.feed_forward.w1.packed_weight_indices": [
1353
- 9830400
1354
  ],
1355
  "layers.10.feed_forward.w1.row_norms": [
1356
  10240
1357
  ],
1358
  "layers.10.feed_forward.w2.packed_weight_indices": [
1359
- 9830400
1360
  ],
1361
  "layers.10.feed_forward.w2.row_norms": [
1362
  3840
1363
  ],
1364
  "layers.10.feed_forward.w3.packed_weight_indices": [
1365
- 9830400
1366
  ],
1367
  "layers.10.feed_forward.w3.row_norms": [
1368
  10240
@@ -1379,9 +1379,6 @@
1379
  "layers.10.ffn_norm2.weight": [
1380
  3840
1381
  ],
1382
- "layers.10.adaLN_modulation.0.bias": [
1383
- 15360
1384
- ],
1385
  "layers.10.adaLN_modulation.0.packed_weight": [
1386
  1966080
1387
  ],
@@ -1389,6 +1386,9 @@
1389
  15360,
1390
  4
1391
  ],
 
 
 
1392
  "layers.11.attention.norm_q.weight": [
1393
  128
1394
  ],
@@ -1408,31 +1408,31 @@
1408
  3840
1409
  ],
1410
  "layers.11.attention.to_v.packed_weight_indices": [
1411
- 3686400
1412
  ],
1413
  "layers.11.attention.to_v.row_norms": [
1414
  3840
1415
  ],
1416
  "layers.11.attention.to_out.0.packed_weight_indices": [
1417
- 3686400
1418
  ],
1419
  "layers.11.attention.to_out.0.row_norms": [
1420
  3840
1421
  ],
1422
  "layers.11.feed_forward.w1.packed_weight_indices": [
1423
- 9830400
1424
  ],
1425
  "layers.11.feed_forward.w1.row_norms": [
1426
  10240
1427
  ],
1428
  "layers.11.feed_forward.w2.packed_weight_indices": [
1429
- 9830400
1430
  ],
1431
  "layers.11.feed_forward.w2.row_norms": [
1432
  3840
1433
  ],
1434
  "layers.11.feed_forward.w3.packed_weight_indices": [
1435
- 9830400
1436
  ],
1437
  "layers.11.feed_forward.w3.row_norms": [
1438
  10240
@@ -1449,9 +1449,6 @@
1449
  "layers.11.ffn_norm2.weight": [
1450
  3840
1451
  ],
1452
- "layers.11.adaLN_modulation.0.bias": [
1453
- 15360
1454
- ],
1455
  "layers.11.adaLN_modulation.0.packed_weight": [
1456
  1966080
1457
  ],
@@ -1459,6 +1456,9 @@
1459
  15360,
1460
  4
1461
  ],
 
 
 
1462
  "layers.12.attention.norm_q.weight": [
1463
  128
1464
  ],
@@ -1478,31 +1478,31 @@
1478
  3840
1479
  ],
1480
  "layers.12.attention.to_v.packed_weight_indices": [
1481
- 3686400
1482
  ],
1483
  "layers.12.attention.to_v.row_norms": [
1484
  3840
1485
  ],
1486
  "layers.12.attention.to_out.0.packed_weight_indices": [
1487
- 3686400
1488
  ],
1489
  "layers.12.attention.to_out.0.row_norms": [
1490
  3840
1491
  ],
1492
  "layers.12.feed_forward.w1.packed_weight_indices": [
1493
- 9830400
1494
  ],
1495
  "layers.12.feed_forward.w1.row_norms": [
1496
  10240
1497
  ],
1498
  "layers.12.feed_forward.w2.packed_weight_indices": [
1499
- 9830400
1500
  ],
1501
  "layers.12.feed_forward.w2.row_norms": [
1502
  3840
1503
  ],
1504
  "layers.12.feed_forward.w3.packed_weight_indices": [
1505
- 9830400
1506
  ],
1507
  "layers.12.feed_forward.w3.row_norms": [
1508
  10240
@@ -1519,9 +1519,6 @@
1519
  "layers.12.ffn_norm2.weight": [
1520
  3840
1521
  ],
1522
- "layers.12.adaLN_modulation.0.bias": [
1523
- 15360
1524
- ],
1525
  "layers.12.adaLN_modulation.0.packed_weight": [
1526
  1966080
1527
  ],
@@ -1529,6 +1526,9 @@
1529
  15360,
1530
  4
1531
  ],
 
 
 
1532
  "layers.13.attention.norm_q.weight": [
1533
  128
1534
  ],
@@ -1548,31 +1548,31 @@
1548
  3840
1549
  ],
1550
  "layers.13.attention.to_v.packed_weight_indices": [
1551
- 3686400
1552
  ],
1553
  "layers.13.attention.to_v.row_norms": [
1554
  3840
1555
  ],
1556
  "layers.13.attention.to_out.0.packed_weight_indices": [
1557
- 3686400
1558
  ],
1559
  "layers.13.attention.to_out.0.row_norms": [
1560
  3840
1561
  ],
1562
  "layers.13.feed_forward.w1.packed_weight_indices": [
1563
- 9830400
1564
  ],
1565
  "layers.13.feed_forward.w1.row_norms": [
1566
  10240
1567
  ],
1568
  "layers.13.feed_forward.w2.packed_weight_indices": [
1569
- 9830400
1570
  ],
1571
  "layers.13.feed_forward.w2.row_norms": [
1572
  3840
1573
  ],
1574
  "layers.13.feed_forward.w3.packed_weight_indices": [
1575
- 9830400
1576
  ],
1577
  "layers.13.feed_forward.w3.row_norms": [
1578
  10240
@@ -1589,9 +1589,6 @@
1589
  "layers.13.ffn_norm2.weight": [
1590
  3840
1591
  ],
1592
- "layers.13.adaLN_modulation.0.bias": [
1593
- 15360
1594
- ],
1595
  "layers.13.adaLN_modulation.0.packed_weight": [
1596
  1966080
1597
  ],
@@ -1599,6 +1596,9 @@
1599
  15360,
1600
  4
1601
  ],
 
 
 
1602
  "layers.14.attention.norm_q.weight": [
1603
  128
1604
  ],
@@ -1618,31 +1618,31 @@
1618
  3840
1619
  ],
1620
  "layers.14.attention.to_v.packed_weight_indices": [
1621
- 3686400
1622
  ],
1623
  "layers.14.attention.to_v.row_norms": [
1624
  3840
1625
  ],
1626
  "layers.14.attention.to_out.0.packed_weight_indices": [
1627
- 3686400
1628
  ],
1629
  "layers.14.attention.to_out.0.row_norms": [
1630
  3840
1631
  ],
1632
  "layers.14.feed_forward.w1.packed_weight_indices": [
1633
- 9830400
1634
  ],
1635
  "layers.14.feed_forward.w1.row_norms": [
1636
  10240
1637
  ],
1638
  "layers.14.feed_forward.w2.packed_weight_indices": [
1639
- 9830400
1640
  ],
1641
  "layers.14.feed_forward.w2.row_norms": [
1642
  3840
1643
  ],
1644
  "layers.14.feed_forward.w3.packed_weight_indices": [
1645
- 9830400
1646
  ],
1647
  "layers.14.feed_forward.w3.row_norms": [
1648
  10240
@@ -1659,9 +1659,6 @@
1659
  "layers.14.ffn_norm2.weight": [
1660
  3840
1661
  ],
1662
- "layers.14.adaLN_modulation.0.bias": [
1663
- 15360
1664
- ],
1665
  "layers.14.adaLN_modulation.0.packed_weight": [
1666
  1966080
1667
  ],
@@ -1669,6 +1666,9 @@
1669
  15360,
1670
  4
1671
  ],
 
 
 
1672
  "layers.15.attention.norm_q.weight": [
1673
  128
1674
  ],
@@ -1688,31 +1688,31 @@
1688
  3840
1689
  ],
1690
  "layers.15.attention.to_v.packed_weight_indices": [
1691
- 3686400
1692
  ],
1693
  "layers.15.attention.to_v.row_norms": [
1694
  3840
1695
  ],
1696
  "layers.15.attention.to_out.0.packed_weight_indices": [
1697
- 3686400
1698
  ],
1699
  "layers.15.attention.to_out.0.row_norms": [
1700
  3840
1701
  ],
1702
  "layers.15.feed_forward.w1.packed_weight_indices": [
1703
- 9830400
1704
  ],
1705
  "layers.15.feed_forward.w1.row_norms": [
1706
  10240
1707
  ],
1708
  "layers.15.feed_forward.w2.packed_weight_indices": [
1709
- 9830400
1710
  ],
1711
  "layers.15.feed_forward.w2.row_norms": [
1712
  3840
1713
  ],
1714
  "layers.15.feed_forward.w3.packed_weight_indices": [
1715
- 9830400
1716
  ],
1717
  "layers.15.feed_forward.w3.row_norms": [
1718
  10240
@@ -1729,9 +1729,6 @@
1729
  "layers.15.ffn_norm2.weight": [
1730
  3840
1731
  ],
1732
- "layers.15.adaLN_modulation.0.bias": [
1733
- 15360
1734
- ],
1735
  "layers.15.adaLN_modulation.0.packed_weight": [
1736
  1966080
1737
  ],
@@ -1739,6 +1736,9 @@
1739
  15360,
1740
  4
1741
  ],
 
 
 
1742
  "layers.16.attention.norm_q.weight": [
1743
  128
1744
  ],
@@ -1758,31 +1758,31 @@
1758
  3840
1759
  ],
1760
  "layers.16.attention.to_v.packed_weight_indices": [
1761
- 3686400
1762
  ],
1763
  "layers.16.attention.to_v.row_norms": [
1764
  3840
1765
  ],
1766
  "layers.16.attention.to_out.0.packed_weight_indices": [
1767
- 3686400
1768
  ],
1769
  "layers.16.attention.to_out.0.row_norms": [
1770
  3840
1771
  ],
1772
  "layers.16.feed_forward.w1.packed_weight_indices": [
1773
- 9830400
1774
  ],
1775
  "layers.16.feed_forward.w1.row_norms": [
1776
  10240
1777
  ],
1778
  "layers.16.feed_forward.w2.packed_weight_indices": [
1779
- 9830400
1780
  ],
1781
  "layers.16.feed_forward.w2.row_norms": [
1782
  3840
1783
  ],
1784
  "layers.16.feed_forward.w3.packed_weight_indices": [
1785
- 9830400
1786
  ],
1787
  "layers.16.feed_forward.w3.row_norms": [
1788
  10240
@@ -1799,9 +1799,6 @@
1799
  "layers.16.ffn_norm2.weight": [
1800
  3840
1801
  ],
1802
- "layers.16.adaLN_modulation.0.bias": [
1803
- 15360
1804
- ],
1805
  "layers.16.adaLN_modulation.0.packed_weight": [
1806
  1966080
1807
  ],
@@ -1809,6 +1806,9 @@
1809
  15360,
1810
  4
1811
  ],
 
 
 
1812
  "layers.17.attention.norm_q.weight": [
1813
  128
1814
  ],
@@ -1828,31 +1828,31 @@
1828
  3840
1829
  ],
1830
  "layers.17.attention.to_v.packed_weight_indices": [
1831
- 3686400
1832
  ],
1833
  "layers.17.attention.to_v.row_norms": [
1834
  3840
1835
  ],
1836
  "layers.17.attention.to_out.0.packed_weight_indices": [
1837
- 3686400
1838
  ],
1839
  "layers.17.attention.to_out.0.row_norms": [
1840
  3840
1841
  ],
1842
  "layers.17.feed_forward.w1.packed_weight_indices": [
1843
- 9830400
1844
  ],
1845
  "layers.17.feed_forward.w1.row_norms": [
1846
  10240
1847
  ],
1848
  "layers.17.feed_forward.w2.packed_weight_indices": [
1849
- 9830400
1850
  ],
1851
  "layers.17.feed_forward.w2.row_norms": [
1852
  3840
1853
  ],
1854
  "layers.17.feed_forward.w3.packed_weight_indices": [
1855
- 9830400
1856
  ],
1857
  "layers.17.feed_forward.w3.row_norms": [
1858
  10240
@@ -1869,9 +1869,6 @@
1869
  "layers.17.ffn_norm2.weight": [
1870
  3840
1871
  ],
1872
- "layers.17.adaLN_modulation.0.bias": [
1873
- 15360
1874
- ],
1875
  "layers.17.adaLN_modulation.0.packed_weight": [
1876
  1966080
1877
  ],
@@ -1879,6 +1876,9 @@
1879
  15360,
1880
  4
1881
  ],
 
 
 
1882
  "layers.18.attention.norm_q.weight": [
1883
  128
1884
  ],
@@ -1898,31 +1898,31 @@
1898
  3840
1899
  ],
1900
  "layers.18.attention.to_v.packed_weight_indices": [
1901
- 3686400
1902
  ],
1903
  "layers.18.attention.to_v.row_norms": [
1904
  3840
1905
  ],
1906
  "layers.18.attention.to_out.0.packed_weight_indices": [
1907
- 3686400
1908
  ],
1909
  "layers.18.attention.to_out.0.row_norms": [
1910
  3840
1911
  ],
1912
  "layers.18.feed_forward.w1.packed_weight_indices": [
1913
- 9830400
1914
  ],
1915
  "layers.18.feed_forward.w1.row_norms": [
1916
  10240
1917
  ],
1918
  "layers.18.feed_forward.w2.packed_weight_indices": [
1919
- 9830400
1920
  ],
1921
  "layers.18.feed_forward.w2.row_norms": [
1922
  3840
1923
  ],
1924
  "layers.18.feed_forward.w3.packed_weight_indices": [
1925
- 9830400
1926
  ],
1927
  "layers.18.feed_forward.w3.row_norms": [
1928
  10240
@@ -1939,9 +1939,6 @@
1939
  "layers.18.ffn_norm2.weight": [
1940
  3840
1941
  ],
1942
- "layers.18.adaLN_modulation.0.bias": [
1943
- 15360
1944
- ],
1945
  "layers.18.adaLN_modulation.0.packed_weight": [
1946
  1966080
1947
  ],
@@ -1949,6 +1946,9 @@
1949
  15360,
1950
  4
1951
  ],
 
 
 
1952
  "layers.19.attention.norm_q.weight": [
1953
  128
1954
  ],
@@ -1968,31 +1968,31 @@
1968
  3840
1969
  ],
1970
  "layers.19.attention.to_v.packed_weight_indices": [
1971
- 3686400
1972
  ],
1973
  "layers.19.attention.to_v.row_norms": [
1974
  3840
1975
  ],
1976
  "layers.19.attention.to_out.0.packed_weight_indices": [
1977
- 3686400
1978
  ],
1979
  "layers.19.attention.to_out.0.row_norms": [
1980
  3840
1981
  ],
1982
  "layers.19.feed_forward.w1.packed_weight_indices": [
1983
- 9830400
1984
  ],
1985
  "layers.19.feed_forward.w1.row_norms": [
1986
  10240
1987
  ],
1988
  "layers.19.feed_forward.w2.packed_weight_indices": [
1989
- 9830400
1990
  ],
1991
  "layers.19.feed_forward.w2.row_norms": [
1992
  3840
1993
  ],
1994
  "layers.19.feed_forward.w3.packed_weight_indices": [
1995
- 9830400
1996
  ],
1997
  "layers.19.feed_forward.w3.row_norms": [
1998
  10240
@@ -2009,9 +2009,6 @@
2009
  "layers.19.ffn_norm2.weight": [
2010
  3840
2011
  ],
2012
- "layers.19.adaLN_modulation.0.bias": [
2013
- 15360
2014
- ],
2015
  "layers.19.adaLN_modulation.0.packed_weight": [
2016
  1966080
2017
  ],
@@ -2019,6 +2016,9 @@
2019
  15360,
2020
  4
2021
  ],
 
 
 
2022
  "layers.20.attention.norm_q.weight": [
2023
  128
2024
  ],
@@ -2038,31 +2038,31 @@
2038
  3840
2039
  ],
2040
  "layers.20.attention.to_v.packed_weight_indices": [
2041
- 3686400
2042
  ],
2043
  "layers.20.attention.to_v.row_norms": [
2044
  3840
2045
  ],
2046
  "layers.20.attention.to_out.0.packed_weight_indices": [
2047
- 3686400
2048
  ],
2049
  "layers.20.attention.to_out.0.row_norms": [
2050
  3840
2051
  ],
2052
  "layers.20.feed_forward.w1.packed_weight_indices": [
2053
- 9830400
2054
  ],
2055
  "layers.20.feed_forward.w1.row_norms": [
2056
  10240
2057
  ],
2058
  "layers.20.feed_forward.w2.packed_weight_indices": [
2059
- 9830400
2060
  ],
2061
  "layers.20.feed_forward.w2.row_norms": [
2062
  3840
2063
  ],
2064
  "layers.20.feed_forward.w3.packed_weight_indices": [
2065
- 9830400
2066
  ],
2067
  "layers.20.feed_forward.w3.row_norms": [
2068
  10240
@@ -2079,9 +2079,6 @@
2079
  "layers.20.ffn_norm2.weight": [
2080
  3840
2081
  ],
2082
- "layers.20.adaLN_modulation.0.bias": [
2083
- 15360
2084
- ],
2085
  "layers.20.adaLN_modulation.0.packed_weight": [
2086
  1966080
2087
  ],
@@ -2089,6 +2086,9 @@
2089
  15360,
2090
  4
2091
  ],
 
 
 
2092
  "layers.21.attention.norm_q.weight": [
2093
  128
2094
  ],
@@ -2108,31 +2108,31 @@
2108
  3840
2109
  ],
2110
  "layers.21.attention.to_v.packed_weight_indices": [
2111
- 3686400
2112
  ],
2113
  "layers.21.attention.to_v.row_norms": [
2114
  3840
2115
  ],
2116
  "layers.21.attention.to_out.0.packed_weight_indices": [
2117
- 3686400
2118
  ],
2119
  "layers.21.attention.to_out.0.row_norms": [
2120
  3840
2121
  ],
2122
  "layers.21.feed_forward.w1.packed_weight_indices": [
2123
- 9830400
2124
  ],
2125
  "layers.21.feed_forward.w1.row_norms": [
2126
  10240
2127
  ],
2128
  "layers.21.feed_forward.w2.packed_weight_indices": [
2129
- 9830400
2130
  ],
2131
  "layers.21.feed_forward.w2.row_norms": [
2132
  3840
2133
  ],
2134
  "layers.21.feed_forward.w3.packed_weight_indices": [
2135
- 9830400
2136
  ],
2137
  "layers.21.feed_forward.w3.row_norms": [
2138
  10240
@@ -2149,9 +2149,6 @@
2149
  "layers.21.ffn_norm2.weight": [
2150
  3840
2151
  ],
2152
- "layers.21.adaLN_modulation.0.bias": [
2153
- 15360
2154
- ],
2155
  "layers.21.adaLN_modulation.0.packed_weight": [
2156
  1966080
2157
  ],
@@ -2159,6 +2156,9 @@
2159
  15360,
2160
  4
2161
  ],
 
 
 
2162
  "layers.22.attention.norm_q.weight": [
2163
  128
2164
  ],
@@ -2178,31 +2178,31 @@
2178
  3840
2179
  ],
2180
  "layers.22.attention.to_v.packed_weight_indices": [
2181
- 3686400
2182
  ],
2183
  "layers.22.attention.to_v.row_norms": [
2184
  3840
2185
  ],
2186
  "layers.22.attention.to_out.0.packed_weight_indices": [
2187
- 3686400
2188
  ],
2189
  "layers.22.attention.to_out.0.row_norms": [
2190
  3840
2191
  ],
2192
  "layers.22.feed_forward.w1.packed_weight_indices": [
2193
- 9830400
2194
  ],
2195
  "layers.22.feed_forward.w1.row_norms": [
2196
  10240
2197
  ],
2198
  "layers.22.feed_forward.w2.packed_weight_indices": [
2199
- 9830400
2200
  ],
2201
  "layers.22.feed_forward.w2.row_norms": [
2202
  3840
2203
  ],
2204
  "layers.22.feed_forward.w3.packed_weight_indices": [
2205
- 9830400
2206
  ],
2207
  "layers.22.feed_forward.w3.row_norms": [
2208
  10240
@@ -2219,9 +2219,6 @@
2219
  "layers.22.ffn_norm2.weight": [
2220
  3840
2221
  ],
2222
- "layers.22.adaLN_modulation.0.bias": [
2223
- 15360
2224
- ],
2225
  "layers.22.adaLN_modulation.0.packed_weight": [
2226
  1966080
2227
  ],
@@ -2229,6 +2226,9 @@
2229
  15360,
2230
  4
2231
  ],
 
 
 
2232
  "layers.23.attention.norm_q.weight": [
2233
  128
2234
  ],
@@ -2248,31 +2248,31 @@
2248
  3840
2249
  ],
2250
  "layers.23.attention.to_v.packed_weight_indices": [
2251
- 3686400
2252
  ],
2253
  "layers.23.attention.to_v.row_norms": [
2254
  3840
2255
  ],
2256
  "layers.23.attention.to_out.0.packed_weight_indices": [
2257
- 3686400
2258
  ],
2259
  "layers.23.attention.to_out.0.row_norms": [
2260
  3840
2261
  ],
2262
  "layers.23.feed_forward.w1.packed_weight_indices": [
2263
- 9830400
2264
  ],
2265
  "layers.23.feed_forward.w1.row_norms": [
2266
  10240
2267
  ],
2268
  "layers.23.feed_forward.w2.packed_weight_indices": [
2269
- 9830400
2270
  ],
2271
  "layers.23.feed_forward.w2.row_norms": [
2272
  3840
2273
  ],
2274
  "layers.23.feed_forward.w3.packed_weight_indices": [
2275
- 9830400
2276
  ],
2277
  "layers.23.feed_forward.w3.row_norms": [
2278
  10240
@@ -2289,9 +2289,6 @@
2289
  "layers.23.ffn_norm2.weight": [
2290
  3840
2291
  ],
2292
- "layers.23.adaLN_modulation.0.bias": [
2293
- 15360
2294
- ],
2295
  "layers.23.adaLN_modulation.0.packed_weight": [
2296
  1966080
2297
  ],
@@ -2299,6 +2296,9 @@
2299
  15360,
2300
  4
2301
  ],
 
 
 
2302
  "layers.24.attention.norm_q.weight": [
2303
  128
2304
  ],
@@ -2318,31 +2318,31 @@
2318
  3840
2319
  ],
2320
  "layers.24.attention.to_v.packed_weight_indices": [
2321
- 3686400
2322
  ],
2323
  "layers.24.attention.to_v.row_norms": [
2324
  3840
2325
  ],
2326
  "layers.24.attention.to_out.0.packed_weight_indices": [
2327
- 3686400
2328
  ],
2329
  "layers.24.attention.to_out.0.row_norms": [
2330
  3840
2331
  ],
2332
  "layers.24.feed_forward.w1.packed_weight_indices": [
2333
- 9830400
2334
  ],
2335
  "layers.24.feed_forward.w1.row_norms": [
2336
  10240
2337
  ],
2338
  "layers.24.feed_forward.w2.packed_weight_indices": [
2339
- 9830400
2340
  ],
2341
  "layers.24.feed_forward.w2.row_norms": [
2342
  3840
2343
  ],
2344
  "layers.24.feed_forward.w3.packed_weight_indices": [
2345
- 9830400
2346
  ],
2347
  "layers.24.feed_forward.w3.row_norms": [
2348
  10240
@@ -2359,9 +2359,6 @@
2359
  "layers.24.ffn_norm2.weight": [
2360
  3840
2361
  ],
2362
- "layers.24.adaLN_modulation.0.bias": [
2363
- 15360
2364
- ],
2365
  "layers.24.adaLN_modulation.0.packed_weight": [
2366
  1966080
2367
  ],
@@ -2369,6 +2366,9 @@
2369
  15360,
2370
  4
2371
  ],
 
 
 
2372
  "layers.25.attention.norm_q.weight": [
2373
  128
2374
  ],
@@ -2388,31 +2388,31 @@
2388
  3840
2389
  ],
2390
  "layers.25.attention.to_v.packed_weight_indices": [
2391
- 3686400
2392
  ],
2393
  "layers.25.attention.to_v.row_norms": [
2394
  3840
2395
  ],
2396
  "layers.25.attention.to_out.0.packed_weight_indices": [
2397
- 3686400
2398
  ],
2399
  "layers.25.attention.to_out.0.row_norms": [
2400
  3840
2401
  ],
2402
  "layers.25.feed_forward.w1.packed_weight_indices": [
2403
- 9830400
2404
  ],
2405
  "layers.25.feed_forward.w1.row_norms": [
2406
  10240
2407
  ],
2408
  "layers.25.feed_forward.w2.packed_weight_indices": [
2409
- 9830400
2410
  ],
2411
  "layers.25.feed_forward.w2.row_norms": [
2412
  3840
2413
  ],
2414
  "layers.25.feed_forward.w3.packed_weight_indices": [
2415
- 9830400
2416
  ],
2417
  "layers.25.feed_forward.w3.row_norms": [
2418
  10240
@@ -2429,9 +2429,6 @@
2429
  "layers.25.ffn_norm2.weight": [
2430
  3840
2431
  ],
2432
- "layers.25.adaLN_modulation.0.bias": [
2433
- 15360
2434
- ],
2435
  "layers.25.adaLN_modulation.0.packed_weight": [
2436
  1966080
2437
  ],
@@ -2439,6 +2436,9 @@
2439
  15360,
2440
  4
2441
  ],
 
 
 
2442
  "layers.26.attention.norm_q.weight": [
2443
  128
2444
  ],
@@ -2446,43 +2446,43 @@
2446
  128
2447
  ],
2448
  "layers.26.attention.to_q.packed_weight_indices": [
2449
- 3686400
2450
  ],
2451
  "layers.26.attention.to_q.row_norms": [
2452
  3840
2453
  ],
2454
  "layers.26.attention.to_k.packed_weight_indices": [
2455
- 3686400
2456
  ],
2457
  "layers.26.attention.to_k.row_norms": [
2458
  3840
2459
  ],
2460
  "layers.26.attention.to_v.packed_weight_indices": [
2461
- 3686400
2462
  ],
2463
  "layers.26.attention.to_v.row_norms": [
2464
  3840
2465
  ],
2466
  "layers.26.attention.to_out.0.packed_weight_indices": [
2467
- 3686400
2468
  ],
2469
  "layers.26.attention.to_out.0.row_norms": [
2470
  3840
2471
  ],
2472
  "layers.26.feed_forward.w1.packed_weight_indices": [
2473
- 9830400
2474
  ],
2475
  "layers.26.feed_forward.w1.row_norms": [
2476
  10240
2477
  ],
2478
  "layers.26.feed_forward.w2.packed_weight_indices": [
2479
- 9830400
2480
  ],
2481
  "layers.26.feed_forward.w2.row_norms": [
2482
  3840
2483
  ],
2484
  "layers.26.feed_forward.w3.packed_weight_indices": [
2485
- 9830400
2486
  ],
2487
  "layers.26.feed_forward.w3.row_norms": [
2488
  10240
@@ -2499,9 +2499,6 @@
2499
  "layers.26.ffn_norm2.weight": [
2500
  3840
2501
  ],
2502
- "layers.26.adaLN_modulation.0.bias": [
2503
- 15360
2504
- ],
2505
  "layers.26.adaLN_modulation.0.packed_weight": [
2506
  1966080
2507
  ],
@@ -2509,6 +2506,9 @@
2509
  15360,
2510
  4
2511
  ],
 
 
 
2512
  "layers.27.attention.norm_q.weight": [
2513
  128
2514
  ],
@@ -2516,43 +2516,43 @@
2516
  128
2517
  ],
2518
  "layers.27.attention.to_q.packed_weight_indices": [
2519
- 3686400
2520
  ],
2521
  "layers.27.attention.to_q.row_norms": [
2522
  3840
2523
  ],
2524
  "layers.27.attention.to_k.packed_weight_indices": [
2525
- 3686400
2526
  ],
2527
  "layers.27.attention.to_k.row_norms": [
2528
  3840
2529
  ],
2530
  "layers.27.attention.to_v.packed_weight_indices": [
2531
- 3686400
2532
  ],
2533
  "layers.27.attention.to_v.row_norms": [
2534
  3840
2535
  ],
2536
  "layers.27.attention.to_out.0.packed_weight_indices": [
2537
- 3686400
2538
  ],
2539
  "layers.27.attention.to_out.0.row_norms": [
2540
  3840
2541
  ],
2542
  "layers.27.feed_forward.w1.packed_weight_indices": [
2543
- 9830400
2544
  ],
2545
  "layers.27.feed_forward.w1.row_norms": [
2546
  10240
2547
  ],
2548
  "layers.27.feed_forward.w2.packed_weight_indices": [
2549
- 9830400
2550
  ],
2551
  "layers.27.feed_forward.w2.row_norms": [
2552
  3840
2553
  ],
2554
  "layers.27.feed_forward.w3.packed_weight_indices": [
2555
- 9830400
2556
  ],
2557
  "layers.27.feed_forward.w3.row_norms": [
2558
  10240
@@ -2569,9 +2569,6 @@
2569
  "layers.27.ffn_norm2.weight": [
2570
  3840
2571
  ],
2572
- "layers.27.adaLN_modulation.0.bias": [
2573
- 15360
2574
- ],
2575
  "layers.27.adaLN_modulation.0.packed_weight": [
2576
  1966080
2577
  ],
@@ -2579,6 +2576,9 @@
2579
  15360,
2580
  4
2581
  ],
 
 
 
2582
  "layers.28.attention.norm_q.weight": [
2583
  128
2584
  ],
@@ -2586,43 +2586,43 @@
2586
  128
2587
  ],
2588
  "layers.28.attention.to_q.packed_weight_indices": [
2589
- 3686400
2590
  ],
2591
  "layers.28.attention.to_q.row_norms": [
2592
  3840
2593
  ],
2594
  "layers.28.attention.to_k.packed_weight_indices": [
2595
- 3686400
2596
  ],
2597
  "layers.28.attention.to_k.row_norms": [
2598
  3840
2599
  ],
2600
  "layers.28.attention.to_v.packed_weight_indices": [
2601
- 3686400
2602
  ],
2603
  "layers.28.attention.to_v.row_norms": [
2604
  3840
2605
  ],
2606
  "layers.28.attention.to_out.0.packed_weight_indices": [
2607
- 3686400
2608
  ],
2609
  "layers.28.attention.to_out.0.row_norms": [
2610
  3840
2611
  ],
2612
  "layers.28.feed_forward.w1.packed_weight_indices": [
2613
- 9830400
2614
  ],
2615
  "layers.28.feed_forward.w1.row_norms": [
2616
  10240
2617
  ],
2618
  "layers.28.feed_forward.w2.packed_weight_indices": [
2619
- 9830400
2620
  ],
2621
  "layers.28.feed_forward.w2.row_norms": [
2622
  3840
2623
  ],
2624
  "layers.28.feed_forward.w3.packed_weight_indices": [
2625
- 9830400
2626
  ],
2627
  "layers.28.feed_forward.w3.row_norms": [
2628
  10240
@@ -2639,9 +2639,6 @@
2639
  "layers.28.ffn_norm2.weight": [
2640
  3840
2641
  ],
2642
- "layers.28.adaLN_modulation.0.bias": [
2643
- 15360
2644
- ],
2645
  "layers.28.adaLN_modulation.0.packed_weight": [
2646
  1966080
2647
  ],
@@ -2649,6 +2646,9 @@
2649
  15360,
2650
  4
2651
  ],
 
 
 
2652
  "layers.29.attention.norm_q.weight": [
2653
  128
2654
  ],
@@ -2656,43 +2656,43 @@
2656
  128
2657
  ],
2658
  "layers.29.attention.to_q.packed_weight_indices": [
2659
- 3686400
2660
  ],
2661
  "layers.29.attention.to_q.row_norms": [
2662
  3840
2663
  ],
2664
  "layers.29.attention.to_k.packed_weight_indices": [
2665
- 3686400
2666
  ],
2667
  "layers.29.attention.to_k.row_norms": [
2668
  3840
2669
  ],
2670
  "layers.29.attention.to_v.packed_weight_indices": [
2671
- 3686400
2672
  ],
2673
  "layers.29.attention.to_v.row_norms": [
2674
  3840
2675
  ],
2676
  "layers.29.attention.to_out.0.packed_weight_indices": [
2677
- 3686400
2678
  ],
2679
  "layers.29.attention.to_out.0.row_norms": [
2680
  3840
2681
  ],
2682
  "layers.29.feed_forward.w1.packed_weight_indices": [
2683
- 9830400
2684
  ],
2685
  "layers.29.feed_forward.w1.row_norms": [
2686
  10240
2687
  ],
2688
  "layers.29.feed_forward.w2.packed_weight_indices": [
2689
- 9830400
2690
  ],
2691
  "layers.29.feed_forward.w2.row_norms": [
2692
  3840
2693
  ],
2694
  "layers.29.feed_forward.w3.packed_weight_indices": [
2695
- 9830400
2696
  ],
2697
  "layers.29.feed_forward.w3.row_norms": [
2698
  10240
@@ -2709,25 +2709,221 @@
2709
  "layers.29.ffn_norm2.weight": [
2710
  3840
2711
  ],
2712
- "layers.29.adaLN_modulation.0.bias": [
2713
- 15360
2714
- ],
2715
  "layers.29.adaLN_modulation.0.packed_weight": [
2716
  1966080
2717
  ],
2718
  "layers.29.adaLN_modulation.0.scales": [
2719
  15360,
2720
  4
 
 
 
2721
  ]
2722
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2723
  "checksums": {
2724
- "assets/image_generation_comparison_matrix.webp": "b92d0cacf0d8c4a7224dbde6bc3736d3dac2fbf231d465f8b992de6c7de16179",
2725
- "benchmark/summary.json": "3637c05001dfc37110277c582656a29310a0ceca8667007229c5f05a9fa0ab32",
2726
- "model.safetensors": "aa8e04810221025dc262f66632975bdbf8466b341ad19db61ba4ad4f174f402f",
2727
- "model_index.json": "33663f26f6c41269f71b4300ba1e92d57e426e52128854b079c25fec0e9080c3",
2728
- "orbitquant_codebooks.safetensors": "9ee09b3fcc9615269db665df2a76b9a94099aea67fb5c9ce1523f5c53f0e7723",
2729
  "orbitquant_rotations.safetensors": "e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825",
2730
- "prompts.json": "0aee42e64abd66ff9293ec140df94e62bf7836eb96a21528a728282a0278c6ed",
2731
- "quantization_config.json": "6cac0dabb20f4fa9cceda63aa86ba4ca5df66902aeec52b98ab133f0c381bc71"
2732
  }
2733
  }
 
20
  "adaln_group_size": 64,
21
  "quantization_device": "cuda",
22
  "weight_quantization_backend": "triton_cuda",
23
+ "quantization_staging_mode": "component",
24
  "target_policy": "z_image",
25
  "adaln_policy": "int4_rtn_group64_bf16_activation",
26
  "quantized_modules": [
 
342
  128
343
  ],
344
  "noise_refiner.0.attention.to_q.packed_weight_indices": [
345
+ 7372800
346
  ],
347
  "noise_refiner.0.attention.to_q.row_norms": [
348
  3840
349
  ],
350
  "noise_refiner.0.attention.to_k.packed_weight_indices": [
351
+ 7372800
352
  ],
353
  "noise_refiner.0.attention.to_k.row_norms": [
354
  3840
355
  ],
356
  "noise_refiner.0.attention.to_v.packed_weight_indices": [
357
+ 7372800
358
  ],
359
  "noise_refiner.0.attention.to_v.row_norms": [
360
  3840
361
  ],
362
  "noise_refiner.0.attention.to_out.0.packed_weight_indices": [
363
+ 7372800
364
  ],
365
  "noise_refiner.0.attention.to_out.0.row_norms": [
366
  3840
367
  ],
368
  "noise_refiner.0.feed_forward.w1.packed_weight_indices": [
369
+ 19660800
370
  ],
371
  "noise_refiner.0.feed_forward.w1.row_norms": [
372
  10240
373
  ],
374
  "noise_refiner.0.feed_forward.w2.packed_weight_indices": [
375
+ 19660800
376
  ],
377
  "noise_refiner.0.feed_forward.w2.row_norms": [
378
  3840
379
  ],
380
  "noise_refiner.0.feed_forward.w3.packed_weight_indices": [
381
+ 19660800
382
  ],
383
  "noise_refiner.0.feed_forward.w3.row_norms": [
384
  10240
 
395
  "noise_refiner.0.ffn_norm2.weight": [
396
  3840
397
  ],
 
 
 
398
  "noise_refiner.0.adaLN_modulation.0.packed_weight": [
399
  1966080
400
  ],
 
402
  15360,
403
  4
404
  ],
405
+ "noise_refiner.0.adaLN_modulation.0.bias": [
406
+ 15360
407
+ ],
408
  "noise_refiner.1.attention.norm_q.weight": [
409
  128
410
  ],
 
412
  128
413
  ],
414
  "noise_refiner.1.attention.to_q.packed_weight_indices": [
415
+ 7372800
416
  ],
417
  "noise_refiner.1.attention.to_q.row_norms": [
418
  3840
419
  ],
420
  "noise_refiner.1.attention.to_k.packed_weight_indices": [
421
+ 7372800
422
  ],
423
  "noise_refiner.1.attention.to_k.row_norms": [
424
  3840
425
  ],
426
  "noise_refiner.1.attention.to_v.packed_weight_indices": [
427
+ 7372800
428
  ],
429
  "noise_refiner.1.attention.to_v.row_norms": [
430
  3840
431
  ],
432
  "noise_refiner.1.attention.to_out.0.packed_weight_indices": [
433
+ 7372800
434
  ],
435
  "noise_refiner.1.attention.to_out.0.row_norms": [
436
  3840
437
  ],
438
  "noise_refiner.1.feed_forward.w1.packed_weight_indices": [
439
+ 19660800
440
  ],
441
  "noise_refiner.1.feed_forward.w1.row_norms": [
442
  10240
443
  ],
444
  "noise_refiner.1.feed_forward.w2.packed_weight_indices": [
445
+ 19660800
446
  ],
447
  "noise_refiner.1.feed_forward.w2.row_norms": [
448
  3840
449
  ],
450
  "noise_refiner.1.feed_forward.w3.packed_weight_indices": [
451
+ 19660800
452
  ],
453
  "noise_refiner.1.feed_forward.w3.row_norms": [
454
  10240
 
465
  "noise_refiner.1.ffn_norm2.weight": [
466
  3840
467
  ],
 
 
 
468
  "noise_refiner.1.adaLN_modulation.0.packed_weight": [
469
  1966080
470
  ],
 
472
  15360,
473
  4
474
  ],
475
+ "noise_refiner.1.adaLN_modulation.0.bias": [
476
+ 15360
477
+ ],
478
  "context_refiner.0.attention.norm_q.weight": [
479
  128
480
  ],
 
482
  128
483
  ],
484
  "context_refiner.0.attention.to_q.packed_weight_indices": [
485
+ 7372800
486
  ],
487
  "context_refiner.0.attention.to_q.row_norms": [
488
  3840
489
  ],
490
  "context_refiner.0.attention.to_k.packed_weight_indices": [
491
+ 7372800
492
  ],
493
  "context_refiner.0.attention.to_k.row_norms": [
494
  3840
495
  ],
496
  "context_refiner.0.attention.to_v.packed_weight_indices": [
497
+ 7372800
498
  ],
499
  "context_refiner.0.attention.to_v.row_norms": [
500
  3840
501
  ],
502
  "context_refiner.0.attention.to_out.0.packed_weight_indices": [
503
+ 7372800
504
  ],
505
  "context_refiner.0.attention.to_out.0.row_norms": [
506
  3840
507
  ],
508
  "context_refiner.0.feed_forward.w1.packed_weight_indices": [
509
+ 19660800
510
  ],
511
  "context_refiner.0.feed_forward.w1.row_norms": [
512
  10240
513
  ],
514
  "context_refiner.0.feed_forward.w2.packed_weight_indices": [
515
+ 19660800
516
  ],
517
  "context_refiner.0.feed_forward.w2.row_norms": [
518
  3840
519
  ],
520
  "context_refiner.0.feed_forward.w3.packed_weight_indices": [
521
+ 19660800
522
  ],
523
  "context_refiner.0.feed_forward.w3.row_norms": [
524
  10240
 
542
  128
543
  ],
544
  "context_refiner.1.attention.to_q.packed_weight_indices": [
545
+ 7372800
546
  ],
547
  "context_refiner.1.attention.to_q.row_norms": [
548
  3840
549
  ],
550
  "context_refiner.1.attention.to_k.packed_weight_indices": [
551
+ 7372800
552
  ],
553
  "context_refiner.1.attention.to_k.row_norms": [
554
  3840
555
  ],
556
  "context_refiner.1.attention.to_v.packed_weight_indices": [
557
+ 7372800
558
  ],
559
  "context_refiner.1.attention.to_v.row_norms": [
560
  3840
561
  ],
562
  "context_refiner.1.attention.to_out.0.packed_weight_indices": [
563
+ 7372800
564
  ],
565
  "context_refiner.1.attention.to_out.0.row_norms": [
566
  3840
567
  ],
568
  "context_refiner.1.feed_forward.w1.packed_weight_indices": [
569
+ 19660800
570
  ],
571
  "context_refiner.1.feed_forward.w1.row_norms": [
572
  10240
573
  ],
574
  "context_refiner.1.feed_forward.w2.packed_weight_indices": [
575
+ 19660800
576
  ],
577
  "context_refiner.1.feed_forward.w2.row_norms": [
578
  3840
579
  ],
580
  "context_refiner.1.feed_forward.w3.packed_weight_indices": [
581
+ 19660800
582
  ],
583
  "context_refiner.1.feed_forward.w3.row_norms": [
584
  10240
 
626
  128
627
  ],
628
  "layers.0.attention.to_q.packed_weight_indices": [
629
+ 7372800
630
  ],
631
  "layers.0.attention.to_q.row_norms": [
632
  3840
633
  ],
634
  "layers.0.attention.to_k.packed_weight_indices": [
635
+ 7372800
636
  ],
637
  "layers.0.attention.to_k.row_norms": [
638
  3840
639
  ],
640
  "layers.0.attention.to_v.packed_weight_indices": [
641
+ 7372800
642
  ],
643
  "layers.0.attention.to_v.row_norms": [
644
  3840
645
  ],
646
  "layers.0.attention.to_out.0.packed_weight_indices": [
647
+ 7372800
648
  ],
649
  "layers.0.attention.to_out.0.row_norms": [
650
  3840
651
  ],
652
  "layers.0.feed_forward.w1.packed_weight_indices": [
653
+ 19660800
654
  ],
655
  "layers.0.feed_forward.w1.row_norms": [
656
  10240
657
  ],
658
  "layers.0.feed_forward.w2.packed_weight_indices": [
659
+ 19660800
660
  ],
661
  "layers.0.feed_forward.w2.row_norms": [
662
  3840
663
  ],
664
  "layers.0.feed_forward.w3.packed_weight_indices": [
665
+ 19660800
666
  ],
667
  "layers.0.feed_forward.w3.row_norms": [
668
  10240
 
679
  "layers.0.ffn_norm2.weight": [
680
  3840
681
  ],
 
 
 
682
  "layers.0.adaLN_modulation.0.packed_weight": [
683
  1966080
684
  ],
 
686
  15360,
687
  4
688
  ],
689
+ "layers.0.adaLN_modulation.0.bias": [
690
+ 15360
691
+ ],
692
  "layers.1.attention.norm_q.weight": [
693
  128
694
  ],
 
696
  128
697
  ],
698
  "layers.1.attention.to_q.packed_weight_indices": [
699
+ 7372800
700
  ],
701
  "layers.1.attention.to_q.row_norms": [
702
  3840
703
  ],
704
  "layers.1.attention.to_k.packed_weight_indices": [
705
+ 7372800
706
  ],
707
  "layers.1.attention.to_k.row_norms": [
708
  3840
709
  ],
710
  "layers.1.attention.to_v.packed_weight_indices": [
711
+ 7372800
712
  ],
713
  "layers.1.attention.to_v.row_norms": [
714
  3840
715
  ],
716
  "layers.1.attention.to_out.0.packed_weight_indices": [
717
+ 7372800
718
  ],
719
  "layers.1.attention.to_out.0.row_norms": [
720
  3840
721
  ],
722
  "layers.1.feed_forward.w1.packed_weight_indices": [
723
+ 19660800
724
  ],
725
  "layers.1.feed_forward.w1.row_norms": [
726
  10240
727
  ],
728
  "layers.1.feed_forward.w2.packed_weight_indices": [
729
+ 19660800
730
  ],
731
  "layers.1.feed_forward.w2.row_norms": [
732
  3840
733
  ],
734
  "layers.1.feed_forward.w3.packed_weight_indices": [
735
+ 19660800
736
  ],
737
  "layers.1.feed_forward.w3.row_norms": [
738
  10240
 
749
  "layers.1.ffn_norm2.weight": [
750
  3840
751
  ],
 
 
 
752
  "layers.1.adaLN_modulation.0.packed_weight": [
753
  1966080
754
  ],
 
756
  15360,
757
  4
758
  ],
759
+ "layers.1.adaLN_modulation.0.bias": [
760
+ 15360
761
+ ],
762
  "layers.2.attention.norm_q.weight": [
763
  128
764
  ],
 
766
  128
767
  ],
768
  "layers.2.attention.to_q.packed_weight_indices": [
769
+ 7372800
770
  ],
771
  "layers.2.attention.to_q.row_norms": [
772
  3840
773
  ],
774
  "layers.2.attention.to_k.packed_weight_indices": [
775
+ 7372800
776
  ],
777
  "layers.2.attention.to_k.row_norms": [
778
  3840
779
  ],
780
  "layers.2.attention.to_v.packed_weight_indices": [
781
+ 7372800
782
  ],
783
  "layers.2.attention.to_v.row_norms": [
784
  3840
785
  ],
786
  "layers.2.attention.to_out.0.packed_weight_indices": [
787
+ 7372800
788
  ],
789
  "layers.2.attention.to_out.0.row_norms": [
790
  3840
791
  ],
792
  "layers.2.feed_forward.w1.packed_weight_indices": [
793
+ 19660800
794
  ],
795
  "layers.2.feed_forward.w1.row_norms": [
796
  10240
797
  ],
798
  "layers.2.feed_forward.w2.packed_weight_indices": [
799
+ 19660800
800
  ],
801
  "layers.2.feed_forward.w2.row_norms": [
802
  3840
803
  ],
804
  "layers.2.feed_forward.w3.packed_weight_indices": [
805
+ 19660800
806
  ],
807
  "layers.2.feed_forward.w3.row_norms": [
808
  10240
 
819
  "layers.2.ffn_norm2.weight": [
820
  3840
821
  ],
 
 
 
822
  "layers.2.adaLN_modulation.0.packed_weight": [
823
  1966080
824
  ],
 
826
  15360,
827
  4
828
  ],
829
+ "layers.2.adaLN_modulation.0.bias": [
830
+ 15360
831
+ ],
832
  "layers.3.attention.norm_q.weight": [
833
  128
834
  ],
 
836
  128
837
  ],
838
  "layers.3.attention.to_q.packed_weight_indices": [
839
+ 7372800
840
  ],
841
  "layers.3.attention.to_q.row_norms": [
842
  3840
843
  ],
844
  "layers.3.attention.to_k.packed_weight_indices": [
845
+ 7372800
846
  ],
847
  "layers.3.attention.to_k.row_norms": [
848
  3840
849
  ],
850
  "layers.3.attention.to_v.packed_weight_indices": [
851
+ 7372800
852
  ],
853
  "layers.3.attention.to_v.row_norms": [
854
  3840
855
  ],
856
  "layers.3.attention.to_out.0.packed_weight_indices": [
857
+ 7372800
858
  ],
859
  "layers.3.attention.to_out.0.row_norms": [
860
  3840
861
  ],
862
  "layers.3.feed_forward.w1.packed_weight_indices": [
863
+ 19660800
864
  ],
865
  "layers.3.feed_forward.w1.row_norms": [
866
  10240
867
  ],
868
  "layers.3.feed_forward.w2.packed_weight_indices": [
869
+ 19660800
870
  ],
871
  "layers.3.feed_forward.w2.row_norms": [
872
  3840
873
  ],
874
  "layers.3.feed_forward.w3.packed_weight_indices": [
875
+ 19660800
876
  ],
877
  "layers.3.feed_forward.w3.row_norms": [
878
  10240
 
889
  "layers.3.ffn_norm2.weight": [
890
  3840
891
  ],
 
 
 
892
  "layers.3.adaLN_modulation.0.packed_weight": [
893
  1966080
894
  ],
 
896
  15360,
897
  4
898
  ],
899
+ "layers.3.adaLN_modulation.0.bias": [
900
+ 15360
901
+ ],
902
  "layers.4.attention.norm_q.weight": [
903
  128
904
  ],
 
918
  3840
919
  ],
920
  "layers.4.attention.to_v.packed_weight_indices": [
921
+ 5529600
922
  ],
923
  "layers.4.attention.to_v.row_norms": [
924
  3840
925
  ],
926
  "layers.4.attention.to_out.0.packed_weight_indices": [
927
+ 5529600
928
  ],
929
  "layers.4.attention.to_out.0.row_norms": [
930
  3840
931
  ],
932
  "layers.4.feed_forward.w1.packed_weight_indices": [
933
+ 14745600
934
  ],
935
  "layers.4.feed_forward.w1.row_norms": [
936
  10240
937
  ],
938
  "layers.4.feed_forward.w2.packed_weight_indices": [
939
+ 14745600
940
  ],
941
  "layers.4.feed_forward.w2.row_norms": [
942
  3840
943
  ],
944
  "layers.4.feed_forward.w3.packed_weight_indices": [
945
+ 14745600
946
  ],
947
  "layers.4.feed_forward.w3.row_norms": [
948
  10240
 
959
  "layers.4.ffn_norm2.weight": [
960
  3840
961
  ],
 
 
 
962
  "layers.4.adaLN_modulation.0.packed_weight": [
963
  1966080
964
  ],
 
966
  15360,
967
  4
968
  ],
969
+ "layers.4.adaLN_modulation.0.bias": [
970
+ 15360
971
+ ],
972
  "layers.5.attention.norm_q.weight": [
973
  128
974
  ],
 
988
  3840
989
  ],
990
  "layers.5.attention.to_v.packed_weight_indices": [
991
+ 5529600
992
  ],
993
  "layers.5.attention.to_v.row_norms": [
994
  3840
995
  ],
996
  "layers.5.attention.to_out.0.packed_weight_indices": [
997
+ 5529600
998
  ],
999
  "layers.5.attention.to_out.0.row_norms": [
1000
  3840
1001
  ],
1002
  "layers.5.feed_forward.w1.packed_weight_indices": [
1003
+ 14745600
1004
  ],
1005
  "layers.5.feed_forward.w1.row_norms": [
1006
  10240
1007
  ],
1008
  "layers.5.feed_forward.w2.packed_weight_indices": [
1009
+ 14745600
1010
  ],
1011
  "layers.5.feed_forward.w2.row_norms": [
1012
  3840
1013
  ],
1014
  "layers.5.feed_forward.w3.packed_weight_indices": [
1015
+ 14745600
1016
  ],
1017
  "layers.5.feed_forward.w3.row_norms": [
1018
  10240
 
1029
  "layers.5.ffn_norm2.weight": [
1030
  3840
1031
  ],
 
 
 
1032
  "layers.5.adaLN_modulation.0.packed_weight": [
1033
  1966080
1034
  ],
 
1036
  15360,
1037
  4
1038
  ],
1039
+ "layers.5.adaLN_modulation.0.bias": [
1040
+ 15360
1041
+ ],
1042
  "layers.6.attention.norm_q.weight": [
1043
  128
1044
  ],
 
1058
  3840
1059
  ],
1060
  "layers.6.attention.to_v.packed_weight_indices": [
1061
+ 5529600
1062
  ],
1063
  "layers.6.attention.to_v.row_norms": [
1064
  3840
1065
  ],
1066
  "layers.6.attention.to_out.0.packed_weight_indices": [
1067
+ 5529600
1068
  ],
1069
  "layers.6.attention.to_out.0.row_norms": [
1070
  3840
1071
  ],
1072
  "layers.6.feed_forward.w1.packed_weight_indices": [
1073
+ 14745600
1074
  ],
1075
  "layers.6.feed_forward.w1.row_norms": [
1076
  10240
1077
  ],
1078
  "layers.6.feed_forward.w2.packed_weight_indices": [
1079
+ 14745600
1080
  ],
1081
  "layers.6.feed_forward.w2.row_norms": [
1082
  3840
1083
  ],
1084
  "layers.6.feed_forward.w3.packed_weight_indices": [
1085
+ 14745600
1086
  ],
1087
  "layers.6.feed_forward.w3.row_norms": [
1088
  10240
 
1099
  "layers.6.ffn_norm2.weight": [
1100
  3840
1101
  ],
 
 
 
1102
  "layers.6.adaLN_modulation.0.packed_weight": [
1103
  1966080
1104
  ],
 
1106
  15360,
1107
  4
1108
  ],
1109
+ "layers.6.adaLN_modulation.0.bias": [
1110
+ 15360
1111
+ ],
1112
  "layers.7.attention.norm_q.weight": [
1113
  128
1114
  ],
 
1128
  3840
1129
  ],
1130
  "layers.7.attention.to_v.packed_weight_indices": [
1131
+ 5529600
1132
  ],
1133
  "layers.7.attention.to_v.row_norms": [
1134
  3840
1135
  ],
1136
  "layers.7.attention.to_out.0.packed_weight_indices": [
1137
+ 5529600
1138
  ],
1139
  "layers.7.attention.to_out.0.row_norms": [
1140
  3840
1141
  ],
1142
  "layers.7.feed_forward.w1.packed_weight_indices": [
1143
+ 14745600
1144
  ],
1145
  "layers.7.feed_forward.w1.row_norms": [
1146
  10240
1147
  ],
1148
  "layers.7.feed_forward.w2.packed_weight_indices": [
1149
+ 14745600
1150
  ],
1151
  "layers.7.feed_forward.w2.row_norms": [
1152
  3840
1153
  ],
1154
  "layers.7.feed_forward.w3.packed_weight_indices": [
1155
+ 14745600
1156
  ],
1157
  "layers.7.feed_forward.w3.row_norms": [
1158
  10240
 
1169
  "layers.7.ffn_norm2.weight": [
1170
  3840
1171
  ],
 
 
 
1172
  "layers.7.adaLN_modulation.0.packed_weight": [
1173
  1966080
1174
  ],
 
1176
  15360,
1177
  4
1178
  ],
1179
+ "layers.7.adaLN_modulation.0.bias": [
1180
+ 15360
1181
+ ],
1182
  "layers.8.attention.norm_q.weight": [
1183
  128
1184
  ],
 
1198
  3840
1199
  ],
1200
  "layers.8.attention.to_v.packed_weight_indices": [
1201
+ 5529600
1202
  ],
1203
  "layers.8.attention.to_v.row_norms": [
1204
  3840
1205
  ],
1206
  "layers.8.attention.to_out.0.packed_weight_indices": [
1207
+ 5529600
1208
  ],
1209
  "layers.8.attention.to_out.0.row_norms": [
1210
  3840
1211
  ],
1212
  "layers.8.feed_forward.w1.packed_weight_indices": [
1213
+ 14745600
1214
  ],
1215
  "layers.8.feed_forward.w1.row_norms": [
1216
  10240
1217
  ],
1218
  "layers.8.feed_forward.w2.packed_weight_indices": [
1219
+ 14745600
1220
  ],
1221
  "layers.8.feed_forward.w2.row_norms": [
1222
  3840
1223
  ],
1224
  "layers.8.feed_forward.w3.packed_weight_indices": [
1225
+ 14745600
1226
  ],
1227
  "layers.8.feed_forward.w3.row_norms": [
1228
  10240
 
1239
  "layers.8.ffn_norm2.weight": [
1240
  3840
1241
  ],
 
 
 
1242
  "layers.8.adaLN_modulation.0.packed_weight": [
1243
  1966080
1244
  ],
 
1246
  15360,
1247
  4
1248
  ],
1249
+ "layers.8.adaLN_modulation.0.bias": [
1250
+ 15360
1251
+ ],
1252
  "layers.9.attention.norm_q.weight": [
1253
  128
1254
  ],
 
1268
  3840
1269
  ],
1270
  "layers.9.attention.to_v.packed_weight_indices": [
1271
+ 5529600
1272
  ],
1273
  "layers.9.attention.to_v.row_norms": [
1274
  3840
1275
  ],
1276
  "layers.9.attention.to_out.0.packed_weight_indices": [
1277
+ 5529600
1278
  ],
1279
  "layers.9.attention.to_out.0.row_norms": [
1280
  3840
1281
  ],
1282
  "layers.9.feed_forward.w1.packed_weight_indices": [
1283
+ 14745600
1284
  ],
1285
  "layers.9.feed_forward.w1.row_norms": [
1286
  10240
1287
  ],
1288
  "layers.9.feed_forward.w2.packed_weight_indices": [
1289
+ 14745600
1290
  ],
1291
  "layers.9.feed_forward.w2.row_norms": [
1292
  3840
1293
  ],
1294
  "layers.9.feed_forward.w3.packed_weight_indices": [
1295
+ 14745600
1296
  ],
1297
  "layers.9.feed_forward.w3.row_norms": [
1298
  10240
 
1309
  "layers.9.ffn_norm2.weight": [
1310
  3840
1311
  ],
 
 
 
1312
  "layers.9.adaLN_modulation.0.packed_weight": [
1313
  1966080
1314
  ],
 
1316
  15360,
1317
  4
1318
  ],
1319
+ "layers.9.adaLN_modulation.0.bias": [
1320
+ 15360
1321
+ ],
1322
  "layers.10.attention.norm_q.weight": [
1323
  128
1324
  ],
 
1338
  3840
1339
  ],
1340
  "layers.10.attention.to_v.packed_weight_indices": [
1341
+ 5529600
1342
  ],
1343
  "layers.10.attention.to_v.row_norms": [
1344
  3840
1345
  ],
1346
  "layers.10.attention.to_out.0.packed_weight_indices": [
1347
+ 5529600
1348
  ],
1349
  "layers.10.attention.to_out.0.row_norms": [
1350
  3840
1351
  ],
1352
  "layers.10.feed_forward.w1.packed_weight_indices": [
1353
+ 14745600
1354
  ],
1355
  "layers.10.feed_forward.w1.row_norms": [
1356
  10240
1357
  ],
1358
  "layers.10.feed_forward.w2.packed_weight_indices": [
1359
+ 14745600
1360
  ],
1361
  "layers.10.feed_forward.w2.row_norms": [
1362
  3840
1363
  ],
1364
  "layers.10.feed_forward.w3.packed_weight_indices": [
1365
+ 14745600
1366
  ],
1367
  "layers.10.feed_forward.w3.row_norms": [
1368
  10240
 
1379
  "layers.10.ffn_norm2.weight": [
1380
  3840
1381
  ],
 
 
 
1382
  "layers.10.adaLN_modulation.0.packed_weight": [
1383
  1966080
1384
  ],
 
1386
  15360,
1387
  4
1388
  ],
1389
+ "layers.10.adaLN_modulation.0.bias": [
1390
+ 15360
1391
+ ],
1392
  "layers.11.attention.norm_q.weight": [
1393
  128
1394
  ],
 
1408
  3840
1409
  ],
1410
  "layers.11.attention.to_v.packed_weight_indices": [
1411
+ 5529600
1412
  ],
1413
  "layers.11.attention.to_v.row_norms": [
1414
  3840
1415
  ],
1416
  "layers.11.attention.to_out.0.packed_weight_indices": [
1417
+ 5529600
1418
  ],
1419
  "layers.11.attention.to_out.0.row_norms": [
1420
  3840
1421
  ],
1422
  "layers.11.feed_forward.w1.packed_weight_indices": [
1423
+ 14745600
1424
  ],
1425
  "layers.11.feed_forward.w1.row_norms": [
1426
  10240
1427
  ],
1428
  "layers.11.feed_forward.w2.packed_weight_indices": [
1429
+ 14745600
1430
  ],
1431
  "layers.11.feed_forward.w2.row_norms": [
1432
  3840
1433
  ],
1434
  "layers.11.feed_forward.w3.packed_weight_indices": [
1435
+ 14745600
1436
  ],
1437
  "layers.11.feed_forward.w3.row_norms": [
1438
  10240
 
1449
  "layers.11.ffn_norm2.weight": [
1450
  3840
1451
  ],
 
 
 
1452
  "layers.11.adaLN_modulation.0.packed_weight": [
1453
  1966080
1454
  ],
 
1456
  15360,
1457
  4
1458
  ],
1459
+ "layers.11.adaLN_modulation.0.bias": [
1460
+ 15360
1461
+ ],
1462
  "layers.12.attention.norm_q.weight": [
1463
  128
1464
  ],
 
1478
  3840
1479
  ],
1480
  "layers.12.attention.to_v.packed_weight_indices": [
1481
+ 5529600
1482
  ],
1483
  "layers.12.attention.to_v.row_norms": [
1484
  3840
1485
  ],
1486
  "layers.12.attention.to_out.0.packed_weight_indices": [
1487
+ 5529600
1488
  ],
1489
  "layers.12.attention.to_out.0.row_norms": [
1490
  3840
1491
  ],
1492
  "layers.12.feed_forward.w1.packed_weight_indices": [
1493
+ 14745600
1494
  ],
1495
  "layers.12.feed_forward.w1.row_norms": [
1496
  10240
1497
  ],
1498
  "layers.12.feed_forward.w2.packed_weight_indices": [
1499
+ 14745600
1500
  ],
1501
  "layers.12.feed_forward.w2.row_norms": [
1502
  3840
1503
  ],
1504
  "layers.12.feed_forward.w3.packed_weight_indices": [
1505
+ 14745600
1506
  ],
1507
  "layers.12.feed_forward.w3.row_norms": [
1508
  10240
 
1519
  "layers.12.ffn_norm2.weight": [
1520
  3840
1521
  ],
 
 
 
1522
  "layers.12.adaLN_modulation.0.packed_weight": [
1523
  1966080
1524
  ],
 
1526
  15360,
1527
  4
1528
  ],
1529
+ "layers.12.adaLN_modulation.0.bias": [
1530
+ 15360
1531
+ ],
1532
  "layers.13.attention.norm_q.weight": [
1533
  128
1534
  ],
 
1548
  3840
1549
  ],
1550
  "layers.13.attention.to_v.packed_weight_indices": [
1551
+ 5529600
1552
  ],
1553
  "layers.13.attention.to_v.row_norms": [
1554
  3840
1555
  ],
1556
  "layers.13.attention.to_out.0.packed_weight_indices": [
1557
+ 5529600
1558
  ],
1559
  "layers.13.attention.to_out.0.row_norms": [
1560
  3840
1561
  ],
1562
  "layers.13.feed_forward.w1.packed_weight_indices": [
1563
+ 14745600
1564
  ],
1565
  "layers.13.feed_forward.w1.row_norms": [
1566
  10240
1567
  ],
1568
  "layers.13.feed_forward.w2.packed_weight_indices": [
1569
+ 14745600
1570
  ],
1571
  "layers.13.feed_forward.w2.row_norms": [
1572
  3840
1573
  ],
1574
  "layers.13.feed_forward.w3.packed_weight_indices": [
1575
+ 14745600
1576
  ],
1577
  "layers.13.feed_forward.w3.row_norms": [
1578
  10240
 
1589
  "layers.13.ffn_norm2.weight": [
1590
  3840
1591
  ],
 
 
 
1592
  "layers.13.adaLN_modulation.0.packed_weight": [
1593
  1966080
1594
  ],
 
1596
  15360,
1597
  4
1598
  ],
1599
+ "layers.13.adaLN_modulation.0.bias": [
1600
+ 15360
1601
+ ],
1602
  "layers.14.attention.norm_q.weight": [
1603
  128
1604
  ],
 
1618
  3840
1619
  ],
1620
  "layers.14.attention.to_v.packed_weight_indices": [
1621
+ 5529600
1622
  ],
1623
  "layers.14.attention.to_v.row_norms": [
1624
  3840
1625
  ],
1626
  "layers.14.attention.to_out.0.packed_weight_indices": [
1627
+ 5529600
1628
  ],
1629
  "layers.14.attention.to_out.0.row_norms": [
1630
  3840
1631
  ],
1632
  "layers.14.feed_forward.w1.packed_weight_indices": [
1633
+ 14745600
1634
  ],
1635
  "layers.14.feed_forward.w1.row_norms": [
1636
  10240
1637
  ],
1638
  "layers.14.feed_forward.w2.packed_weight_indices": [
1639
+ 14745600
1640
  ],
1641
  "layers.14.feed_forward.w2.row_norms": [
1642
  3840
1643
  ],
1644
  "layers.14.feed_forward.w3.packed_weight_indices": [
1645
+ 14745600
1646
  ],
1647
  "layers.14.feed_forward.w3.row_norms": [
1648
  10240
 
1659
  "layers.14.ffn_norm2.weight": [
1660
  3840
1661
  ],
 
 
 
1662
  "layers.14.adaLN_modulation.0.packed_weight": [
1663
  1966080
1664
  ],
 
1666
  15360,
1667
  4
1668
  ],
1669
+ "layers.14.adaLN_modulation.0.bias": [
1670
+ 15360
1671
+ ],
1672
  "layers.15.attention.norm_q.weight": [
1673
  128
1674
  ],
 
1688
  3840
1689
  ],
1690
  "layers.15.attention.to_v.packed_weight_indices": [
1691
+ 5529600
1692
  ],
1693
  "layers.15.attention.to_v.row_norms": [
1694
  3840
1695
  ],
1696
  "layers.15.attention.to_out.0.packed_weight_indices": [
1697
+ 5529600
1698
  ],
1699
  "layers.15.attention.to_out.0.row_norms": [
1700
  3840
1701
  ],
1702
  "layers.15.feed_forward.w1.packed_weight_indices": [
1703
+ 14745600
1704
  ],
1705
  "layers.15.feed_forward.w1.row_norms": [
1706
  10240
1707
  ],
1708
  "layers.15.feed_forward.w2.packed_weight_indices": [
1709
+ 14745600
1710
  ],
1711
  "layers.15.feed_forward.w2.row_norms": [
1712
  3840
1713
  ],
1714
  "layers.15.feed_forward.w3.packed_weight_indices": [
1715
+ 14745600
1716
  ],
1717
  "layers.15.feed_forward.w3.row_norms": [
1718
  10240
 
1729
  "layers.15.ffn_norm2.weight": [
1730
  3840
1731
  ],
 
 
 
1732
  "layers.15.adaLN_modulation.0.packed_weight": [
1733
  1966080
1734
  ],
 
1736
  15360,
1737
  4
1738
  ],
1739
+ "layers.15.adaLN_modulation.0.bias": [
1740
+ 15360
1741
+ ],
1742
  "layers.16.attention.norm_q.weight": [
1743
  128
1744
  ],
 
1758
  3840
1759
  ],
1760
  "layers.16.attention.to_v.packed_weight_indices": [
1761
+ 5529600
1762
  ],
1763
  "layers.16.attention.to_v.row_norms": [
1764
  3840
1765
  ],
1766
  "layers.16.attention.to_out.0.packed_weight_indices": [
1767
+ 5529600
1768
  ],
1769
  "layers.16.attention.to_out.0.row_norms": [
1770
  3840
1771
  ],
1772
  "layers.16.feed_forward.w1.packed_weight_indices": [
1773
+ 14745600
1774
  ],
1775
  "layers.16.feed_forward.w1.row_norms": [
1776
  10240
1777
  ],
1778
  "layers.16.feed_forward.w2.packed_weight_indices": [
1779
+ 14745600
1780
  ],
1781
  "layers.16.feed_forward.w2.row_norms": [
1782
  3840
1783
  ],
1784
  "layers.16.feed_forward.w3.packed_weight_indices": [
1785
+ 14745600
1786
  ],
1787
  "layers.16.feed_forward.w3.row_norms": [
1788
  10240
 
1799
  "layers.16.ffn_norm2.weight": [
1800
  3840
1801
  ],
 
 
 
1802
  "layers.16.adaLN_modulation.0.packed_weight": [
1803
  1966080
1804
  ],
 
1806
  15360,
1807
  4
1808
  ],
1809
+ "layers.16.adaLN_modulation.0.bias": [
1810
+ 15360
1811
+ ],
1812
  "layers.17.attention.norm_q.weight": [
1813
  128
1814
  ],
 
1828
  3840
1829
  ],
1830
  "layers.17.attention.to_v.packed_weight_indices": [
1831
+ 5529600
1832
  ],
1833
  "layers.17.attention.to_v.row_norms": [
1834
  3840
1835
  ],
1836
  "layers.17.attention.to_out.0.packed_weight_indices": [
1837
+ 5529600
1838
  ],
1839
  "layers.17.attention.to_out.0.row_norms": [
1840
  3840
1841
  ],
1842
  "layers.17.feed_forward.w1.packed_weight_indices": [
1843
+ 14745600
1844
  ],
1845
  "layers.17.feed_forward.w1.row_norms": [
1846
  10240
1847
  ],
1848
  "layers.17.feed_forward.w2.packed_weight_indices": [
1849
+ 14745600
1850
  ],
1851
  "layers.17.feed_forward.w2.row_norms": [
1852
  3840
1853
  ],
1854
  "layers.17.feed_forward.w3.packed_weight_indices": [
1855
+ 14745600
1856
  ],
1857
  "layers.17.feed_forward.w3.row_norms": [
1858
  10240
 
1869
  "layers.17.ffn_norm2.weight": [
1870
  3840
1871
  ],
 
 
 
1872
  "layers.17.adaLN_modulation.0.packed_weight": [
1873
  1966080
1874
  ],
 
1876
  15360,
1877
  4
1878
  ],
1879
+ "layers.17.adaLN_modulation.0.bias": [
1880
+ 15360
1881
+ ],
1882
  "layers.18.attention.norm_q.weight": [
1883
  128
1884
  ],
 
1898
  3840
1899
  ],
1900
  "layers.18.attention.to_v.packed_weight_indices": [
1901
+ 5529600
1902
  ],
1903
  "layers.18.attention.to_v.row_norms": [
1904
  3840
1905
  ],
1906
  "layers.18.attention.to_out.0.packed_weight_indices": [
1907
+ 5529600
1908
  ],
1909
  "layers.18.attention.to_out.0.row_norms": [
1910
  3840
1911
  ],
1912
  "layers.18.feed_forward.w1.packed_weight_indices": [
1913
+ 14745600
1914
  ],
1915
  "layers.18.feed_forward.w1.row_norms": [
1916
  10240
1917
  ],
1918
  "layers.18.feed_forward.w2.packed_weight_indices": [
1919
+ 14745600
1920
  ],
1921
  "layers.18.feed_forward.w2.row_norms": [
1922
  3840
1923
  ],
1924
  "layers.18.feed_forward.w3.packed_weight_indices": [
1925
+ 14745600
1926
  ],
1927
  "layers.18.feed_forward.w3.row_norms": [
1928
  10240
 
1939
  "layers.18.ffn_norm2.weight": [
1940
  3840
1941
  ],
 
 
 
1942
  "layers.18.adaLN_modulation.0.packed_weight": [
1943
  1966080
1944
  ],
 
1946
  15360,
1947
  4
1948
  ],
1949
+ "layers.18.adaLN_modulation.0.bias": [
1950
+ 15360
1951
+ ],
1952
  "layers.19.attention.norm_q.weight": [
1953
  128
1954
  ],
 
1968
  3840
1969
  ],
1970
  "layers.19.attention.to_v.packed_weight_indices": [
1971
+ 5529600
1972
  ],
1973
  "layers.19.attention.to_v.row_norms": [
1974
  3840
1975
  ],
1976
  "layers.19.attention.to_out.0.packed_weight_indices": [
1977
+ 5529600
1978
  ],
1979
  "layers.19.attention.to_out.0.row_norms": [
1980
  3840
1981
  ],
1982
  "layers.19.feed_forward.w1.packed_weight_indices": [
1983
+ 14745600
1984
  ],
1985
  "layers.19.feed_forward.w1.row_norms": [
1986
  10240
1987
  ],
1988
  "layers.19.feed_forward.w2.packed_weight_indices": [
1989
+ 14745600
1990
  ],
1991
  "layers.19.feed_forward.w2.row_norms": [
1992
  3840
1993
  ],
1994
  "layers.19.feed_forward.w3.packed_weight_indices": [
1995
+ 14745600
1996
  ],
1997
  "layers.19.feed_forward.w3.row_norms": [
1998
  10240
 
2009
  "layers.19.ffn_norm2.weight": [
2010
  3840
2011
  ],
 
 
 
2012
  "layers.19.adaLN_modulation.0.packed_weight": [
2013
  1966080
2014
  ],
 
2016
  15360,
2017
  4
2018
  ],
2019
+ "layers.19.adaLN_modulation.0.bias": [
2020
+ 15360
2021
+ ],
2022
  "layers.20.attention.norm_q.weight": [
2023
  128
2024
  ],
 
2038
  3840
2039
  ],
2040
  "layers.20.attention.to_v.packed_weight_indices": [
2041
+ 5529600
2042
  ],
2043
  "layers.20.attention.to_v.row_norms": [
2044
  3840
2045
  ],
2046
  "layers.20.attention.to_out.0.packed_weight_indices": [
2047
+ 5529600
2048
  ],
2049
  "layers.20.attention.to_out.0.row_norms": [
2050
  3840
2051
  ],
2052
  "layers.20.feed_forward.w1.packed_weight_indices": [
2053
+ 14745600
2054
  ],
2055
  "layers.20.feed_forward.w1.row_norms": [
2056
  10240
2057
  ],
2058
  "layers.20.feed_forward.w2.packed_weight_indices": [
2059
+ 14745600
2060
  ],
2061
  "layers.20.feed_forward.w2.row_norms": [
2062
  3840
2063
  ],
2064
  "layers.20.feed_forward.w3.packed_weight_indices": [
2065
+ 14745600
2066
  ],
2067
  "layers.20.feed_forward.w3.row_norms": [
2068
  10240
 
2079
  "layers.20.ffn_norm2.weight": [
2080
  3840
2081
  ],
 
 
 
2082
  "layers.20.adaLN_modulation.0.packed_weight": [
2083
  1966080
2084
  ],
 
2086
  15360,
2087
  4
2088
  ],
2089
+ "layers.20.adaLN_modulation.0.bias": [
2090
+ 15360
2091
+ ],
2092
  "layers.21.attention.norm_q.weight": [
2093
  128
2094
  ],
 
2108
  3840
2109
  ],
2110
  "layers.21.attention.to_v.packed_weight_indices": [
2111
+ 5529600
2112
  ],
2113
  "layers.21.attention.to_v.row_norms": [
2114
  3840
2115
  ],
2116
  "layers.21.attention.to_out.0.packed_weight_indices": [
2117
+ 5529600
2118
  ],
2119
  "layers.21.attention.to_out.0.row_norms": [
2120
  3840
2121
  ],
2122
  "layers.21.feed_forward.w1.packed_weight_indices": [
2123
+ 14745600
2124
  ],
2125
  "layers.21.feed_forward.w1.row_norms": [
2126
  10240
2127
  ],
2128
  "layers.21.feed_forward.w2.packed_weight_indices": [
2129
+ 14745600
2130
  ],
2131
  "layers.21.feed_forward.w2.row_norms": [
2132
  3840
2133
  ],
2134
  "layers.21.feed_forward.w3.packed_weight_indices": [
2135
+ 14745600
2136
  ],
2137
  "layers.21.feed_forward.w3.row_norms": [
2138
  10240
 
2149
  "layers.21.ffn_norm2.weight": [
2150
  3840
2151
  ],
 
 
 
2152
  "layers.21.adaLN_modulation.0.packed_weight": [
2153
  1966080
2154
  ],
 
2156
  15360,
2157
  4
2158
  ],
2159
+ "layers.21.adaLN_modulation.0.bias": [
2160
+ 15360
2161
+ ],
2162
  "layers.22.attention.norm_q.weight": [
2163
  128
2164
  ],
 
2178
  3840
2179
  ],
2180
  "layers.22.attention.to_v.packed_weight_indices": [
2181
+ 5529600
2182
  ],
2183
  "layers.22.attention.to_v.row_norms": [
2184
  3840
2185
  ],
2186
  "layers.22.attention.to_out.0.packed_weight_indices": [
2187
+ 5529600
2188
  ],
2189
  "layers.22.attention.to_out.0.row_norms": [
2190
  3840
2191
  ],
2192
  "layers.22.feed_forward.w1.packed_weight_indices": [
2193
+ 14745600
2194
  ],
2195
  "layers.22.feed_forward.w1.row_norms": [
2196
  10240
2197
  ],
2198
  "layers.22.feed_forward.w2.packed_weight_indices": [
2199
+ 14745600
2200
  ],
2201
  "layers.22.feed_forward.w2.row_norms": [
2202
  3840
2203
  ],
2204
  "layers.22.feed_forward.w3.packed_weight_indices": [
2205
+ 14745600
2206
  ],
2207
  "layers.22.feed_forward.w3.row_norms": [
2208
  10240
 
2219
  "layers.22.ffn_norm2.weight": [
2220
  3840
2221
  ],
 
 
 
2222
  "layers.22.adaLN_modulation.0.packed_weight": [
2223
  1966080
2224
  ],
 
2226
  15360,
2227
  4
2228
  ],
2229
+ "layers.22.adaLN_modulation.0.bias": [
2230
+ 15360
2231
+ ],
2232
  "layers.23.attention.norm_q.weight": [
2233
  128
2234
  ],
 
2248
  3840
2249
  ],
2250
  "layers.23.attention.to_v.packed_weight_indices": [
2251
+ 5529600
2252
  ],
2253
  "layers.23.attention.to_v.row_norms": [
2254
  3840
2255
  ],
2256
  "layers.23.attention.to_out.0.packed_weight_indices": [
2257
+ 5529600
2258
  ],
2259
  "layers.23.attention.to_out.0.row_norms": [
2260
  3840
2261
  ],
2262
  "layers.23.feed_forward.w1.packed_weight_indices": [
2263
+ 14745600
2264
  ],
2265
  "layers.23.feed_forward.w1.row_norms": [
2266
  10240
2267
  ],
2268
  "layers.23.feed_forward.w2.packed_weight_indices": [
2269
+ 14745600
2270
  ],
2271
  "layers.23.feed_forward.w2.row_norms": [
2272
  3840
2273
  ],
2274
  "layers.23.feed_forward.w3.packed_weight_indices": [
2275
+ 14745600
2276
  ],
2277
  "layers.23.feed_forward.w3.row_norms": [
2278
  10240
 
2289
  "layers.23.ffn_norm2.weight": [
2290
  3840
2291
  ],
 
 
 
2292
  "layers.23.adaLN_modulation.0.packed_weight": [
2293
  1966080
2294
  ],
 
2296
  15360,
2297
  4
2298
  ],
2299
+ "layers.23.adaLN_modulation.0.bias": [
2300
+ 15360
2301
+ ],
2302
  "layers.24.attention.norm_q.weight": [
2303
  128
2304
  ],
 
2318
  3840
2319
  ],
2320
  "layers.24.attention.to_v.packed_weight_indices": [
2321
+ 5529600
2322
  ],
2323
  "layers.24.attention.to_v.row_norms": [
2324
  3840
2325
  ],
2326
  "layers.24.attention.to_out.0.packed_weight_indices": [
2327
+ 5529600
2328
  ],
2329
  "layers.24.attention.to_out.0.row_norms": [
2330
  3840
2331
  ],
2332
  "layers.24.feed_forward.w1.packed_weight_indices": [
2333
+ 14745600
2334
  ],
2335
  "layers.24.feed_forward.w1.row_norms": [
2336
  10240
2337
  ],
2338
  "layers.24.feed_forward.w2.packed_weight_indices": [
2339
+ 14745600
2340
  ],
2341
  "layers.24.feed_forward.w2.row_norms": [
2342
  3840
2343
  ],
2344
  "layers.24.feed_forward.w3.packed_weight_indices": [
2345
+ 14745600
2346
  ],
2347
  "layers.24.feed_forward.w3.row_norms": [
2348
  10240
 
2359
  "layers.24.ffn_norm2.weight": [
2360
  3840
2361
  ],
 
 
 
2362
  "layers.24.adaLN_modulation.0.packed_weight": [
2363
  1966080
2364
  ],
 
2366
  15360,
2367
  4
2368
  ],
2369
+ "layers.24.adaLN_modulation.0.bias": [
2370
+ 15360
2371
+ ],
2372
  "layers.25.attention.norm_q.weight": [
2373
  128
2374
  ],
 
2388
  3840
2389
  ],
2390
  "layers.25.attention.to_v.packed_weight_indices": [
2391
+ 5529600
2392
  ],
2393
  "layers.25.attention.to_v.row_norms": [
2394
  3840
2395
  ],
2396
  "layers.25.attention.to_out.0.packed_weight_indices": [
2397
+ 5529600
2398
  ],
2399
  "layers.25.attention.to_out.0.row_norms": [
2400
  3840
2401
  ],
2402
  "layers.25.feed_forward.w1.packed_weight_indices": [
2403
+ 14745600
2404
  ],
2405
  "layers.25.feed_forward.w1.row_norms": [
2406
  10240
2407
  ],
2408
  "layers.25.feed_forward.w2.packed_weight_indices": [
2409
+ 14745600
2410
  ],
2411
  "layers.25.feed_forward.w2.row_norms": [
2412
  3840
2413
  ],
2414
  "layers.25.feed_forward.w3.packed_weight_indices": [
2415
+ 14745600
2416
  ],
2417
  "layers.25.feed_forward.w3.row_norms": [
2418
  10240
 
2429
  "layers.25.ffn_norm2.weight": [
2430
  3840
2431
  ],
 
 
 
2432
  "layers.25.adaLN_modulation.0.packed_weight": [
2433
  1966080
2434
  ],
 
2436
  15360,
2437
  4
2438
  ],
2439
+ "layers.25.adaLN_modulation.0.bias": [
2440
+ 15360
2441
+ ],
2442
  "layers.26.attention.norm_q.weight": [
2443
  128
2444
  ],
 
2446
  128
2447
  ],
2448
  "layers.26.attention.to_q.packed_weight_indices": [
2449
+ 7372800
2450
  ],
2451
  "layers.26.attention.to_q.row_norms": [
2452
  3840
2453
  ],
2454
  "layers.26.attention.to_k.packed_weight_indices": [
2455
+ 7372800
2456
  ],
2457
  "layers.26.attention.to_k.row_norms": [
2458
  3840
2459
  ],
2460
  "layers.26.attention.to_v.packed_weight_indices": [
2461
+ 7372800
2462
  ],
2463
  "layers.26.attention.to_v.row_norms": [
2464
  3840
2465
  ],
2466
  "layers.26.attention.to_out.0.packed_weight_indices": [
2467
+ 7372800
2468
  ],
2469
  "layers.26.attention.to_out.0.row_norms": [
2470
  3840
2471
  ],
2472
  "layers.26.feed_forward.w1.packed_weight_indices": [
2473
+ 19660800
2474
  ],
2475
  "layers.26.feed_forward.w1.row_norms": [
2476
  10240
2477
  ],
2478
  "layers.26.feed_forward.w2.packed_weight_indices": [
2479
+ 19660800
2480
  ],
2481
  "layers.26.feed_forward.w2.row_norms": [
2482
  3840
2483
  ],
2484
  "layers.26.feed_forward.w3.packed_weight_indices": [
2485
+ 19660800
2486
  ],
2487
  "layers.26.feed_forward.w3.row_norms": [
2488
  10240
 
2499
  "layers.26.ffn_norm2.weight": [
2500
  3840
2501
  ],
 
 
 
2502
  "layers.26.adaLN_modulation.0.packed_weight": [
2503
  1966080
2504
  ],
 
2506
  15360,
2507
  4
2508
  ],
2509
+ "layers.26.adaLN_modulation.0.bias": [
2510
+ 15360
2511
+ ],
2512
  "layers.27.attention.norm_q.weight": [
2513
  128
2514
  ],
 
2516
  128
2517
  ],
2518
  "layers.27.attention.to_q.packed_weight_indices": [
2519
+ 7372800
2520
  ],
2521
  "layers.27.attention.to_q.row_norms": [
2522
  3840
2523
  ],
2524
  "layers.27.attention.to_k.packed_weight_indices": [
2525
+ 7372800
2526
  ],
2527
  "layers.27.attention.to_k.row_norms": [
2528
  3840
2529
  ],
2530
  "layers.27.attention.to_v.packed_weight_indices": [
2531
+ 7372800
2532
  ],
2533
  "layers.27.attention.to_v.row_norms": [
2534
  3840
2535
  ],
2536
  "layers.27.attention.to_out.0.packed_weight_indices": [
2537
+ 7372800
2538
  ],
2539
  "layers.27.attention.to_out.0.row_norms": [
2540
  3840
2541
  ],
2542
  "layers.27.feed_forward.w1.packed_weight_indices": [
2543
+ 19660800
2544
  ],
2545
  "layers.27.feed_forward.w1.row_norms": [
2546
  10240
2547
  ],
2548
  "layers.27.feed_forward.w2.packed_weight_indices": [
2549
+ 19660800
2550
  ],
2551
  "layers.27.feed_forward.w2.row_norms": [
2552
  3840
2553
  ],
2554
  "layers.27.feed_forward.w3.packed_weight_indices": [
2555
+ 19660800
2556
  ],
2557
  "layers.27.feed_forward.w3.row_norms": [
2558
  10240
 
2569
  "layers.27.ffn_norm2.weight": [
2570
  3840
2571
  ],
 
 
 
2572
  "layers.27.adaLN_modulation.0.packed_weight": [
2573
  1966080
2574
  ],
 
2576
  15360,
2577
  4
2578
  ],
2579
+ "layers.27.adaLN_modulation.0.bias": [
2580
+ 15360
2581
+ ],
2582
  "layers.28.attention.norm_q.weight": [
2583
  128
2584
  ],
 
2586
  128
2587
  ],
2588
  "layers.28.attention.to_q.packed_weight_indices": [
2589
+ 7372800
2590
  ],
2591
  "layers.28.attention.to_q.row_norms": [
2592
  3840
2593
  ],
2594
  "layers.28.attention.to_k.packed_weight_indices": [
2595
+ 7372800
2596
  ],
2597
  "layers.28.attention.to_k.row_norms": [
2598
  3840
2599
  ],
2600
  "layers.28.attention.to_v.packed_weight_indices": [
2601
+ 7372800
2602
  ],
2603
  "layers.28.attention.to_v.row_norms": [
2604
  3840
2605
  ],
2606
  "layers.28.attention.to_out.0.packed_weight_indices": [
2607
+ 7372800
2608
  ],
2609
  "layers.28.attention.to_out.0.row_norms": [
2610
  3840
2611
  ],
2612
  "layers.28.feed_forward.w1.packed_weight_indices": [
2613
+ 19660800
2614
  ],
2615
  "layers.28.feed_forward.w1.row_norms": [
2616
  10240
2617
  ],
2618
  "layers.28.feed_forward.w2.packed_weight_indices": [
2619
+ 19660800
2620
  ],
2621
  "layers.28.feed_forward.w2.row_norms": [
2622
  3840
2623
  ],
2624
  "layers.28.feed_forward.w3.packed_weight_indices": [
2625
+ 19660800
2626
  ],
2627
  "layers.28.feed_forward.w3.row_norms": [
2628
  10240
 
2639
  "layers.28.ffn_norm2.weight": [
2640
  3840
2641
  ],
 
 
 
2642
  "layers.28.adaLN_modulation.0.packed_weight": [
2643
  1966080
2644
  ],
 
2646
  15360,
2647
  4
2648
  ],
2649
+ "layers.28.adaLN_modulation.0.bias": [
2650
+ 15360
2651
+ ],
2652
  "layers.29.attention.norm_q.weight": [
2653
  128
2654
  ],
 
2656
  128
2657
  ],
2658
  "layers.29.attention.to_q.packed_weight_indices": [
2659
+ 7372800
2660
  ],
2661
  "layers.29.attention.to_q.row_norms": [
2662
  3840
2663
  ],
2664
  "layers.29.attention.to_k.packed_weight_indices": [
2665
+ 7372800
2666
  ],
2667
  "layers.29.attention.to_k.row_norms": [
2668
  3840
2669
  ],
2670
  "layers.29.attention.to_v.packed_weight_indices": [
2671
+ 7372800
2672
  ],
2673
  "layers.29.attention.to_v.row_norms": [
2674
  3840
2675
  ],
2676
  "layers.29.attention.to_out.0.packed_weight_indices": [
2677
+ 7372800
2678
  ],
2679
  "layers.29.attention.to_out.0.row_norms": [
2680
  3840
2681
  ],
2682
  "layers.29.feed_forward.w1.packed_weight_indices": [
2683
+ 19660800
2684
  ],
2685
  "layers.29.feed_forward.w1.row_norms": [
2686
  10240
2687
  ],
2688
  "layers.29.feed_forward.w2.packed_weight_indices": [
2689
+ 19660800
2690
  ],
2691
  "layers.29.feed_forward.w2.row_norms": [
2692
  3840
2693
  ],
2694
  "layers.29.feed_forward.w3.packed_weight_indices": [
2695
+ 19660800
2696
  ],
2697
  "layers.29.feed_forward.w3.row_norms": [
2698
  10240
 
2709
  "layers.29.ffn_norm2.weight": [
2710
  3840
2711
  ],
 
 
 
2712
  "layers.29.adaLN_modulation.0.packed_weight": [
2713
  1966080
2714
  ],
2715
  "layers.29.adaLN_modulation.0.scales": [
2716
  15360,
2717
  4
2718
+ ],
2719
+ "layers.29.adaLN_modulation.0.bias": [
2720
+ 15360
2721
  ]
2722
  },
2723
+ "module_bits": {
2724
+ "noise_refiner.0.attention.to_q": 4,
2725
+ "noise_refiner.0.attention.to_k": 4,
2726
+ "noise_refiner.0.attention.to_v": 4,
2727
+ "noise_refiner.0.attention.to_out.0": 4,
2728
+ "noise_refiner.0.feed_forward.w1": 4,
2729
+ "noise_refiner.0.feed_forward.w2": 4,
2730
+ "noise_refiner.0.feed_forward.w3": 4,
2731
+ "noise_refiner.1.attention.to_q": 4,
2732
+ "noise_refiner.1.attention.to_k": 4,
2733
+ "noise_refiner.1.attention.to_v": 4,
2734
+ "noise_refiner.1.attention.to_out.0": 4,
2735
+ "noise_refiner.1.feed_forward.w1": 4,
2736
+ "noise_refiner.1.feed_forward.w2": 4,
2737
+ "noise_refiner.1.feed_forward.w3": 4,
2738
+ "context_refiner.0.attention.to_q": 4,
2739
+ "context_refiner.0.attention.to_k": 4,
2740
+ "context_refiner.0.attention.to_v": 4,
2741
+ "context_refiner.0.attention.to_out.0": 4,
2742
+ "context_refiner.0.feed_forward.w1": 4,
2743
+ "context_refiner.0.feed_forward.w2": 4,
2744
+ "context_refiner.0.feed_forward.w3": 4,
2745
+ "context_refiner.1.attention.to_q": 4,
2746
+ "context_refiner.1.attention.to_k": 4,
2747
+ "context_refiner.1.attention.to_v": 4,
2748
+ "context_refiner.1.attention.to_out.0": 4,
2749
+ "context_refiner.1.feed_forward.w1": 4,
2750
+ "context_refiner.1.feed_forward.w2": 4,
2751
+ "context_refiner.1.feed_forward.w3": 4,
2752
+ "layers.0.attention.to_q": 4,
2753
+ "layers.0.attention.to_k": 4,
2754
+ "layers.0.attention.to_v": 4,
2755
+ "layers.0.attention.to_out.0": 4,
2756
+ "layers.0.feed_forward.w1": 4,
2757
+ "layers.0.feed_forward.w2": 4,
2758
+ "layers.0.feed_forward.w3": 4,
2759
+ "layers.1.attention.to_q": 4,
2760
+ "layers.1.attention.to_k": 4,
2761
+ "layers.1.attention.to_v": 4,
2762
+ "layers.1.attention.to_out.0": 4,
2763
+ "layers.1.feed_forward.w1": 4,
2764
+ "layers.1.feed_forward.w2": 4,
2765
+ "layers.1.feed_forward.w3": 4,
2766
+ "layers.2.attention.to_q": 4,
2767
+ "layers.2.attention.to_k": 4,
2768
+ "layers.2.attention.to_v": 4,
2769
+ "layers.2.attention.to_out.0": 4,
2770
+ "layers.2.feed_forward.w1": 4,
2771
+ "layers.2.feed_forward.w2": 4,
2772
+ "layers.2.feed_forward.w3": 4,
2773
+ "layers.3.attention.to_q": 4,
2774
+ "layers.3.attention.to_k": 4,
2775
+ "layers.3.attention.to_v": 4,
2776
+ "layers.3.attention.to_out.0": 4,
2777
+ "layers.3.feed_forward.w1": 4,
2778
+ "layers.3.feed_forward.w2": 4,
2779
+ "layers.3.feed_forward.w3": 4,
2780
+ "layers.4.attention.to_v": 3,
2781
+ "layers.4.attention.to_out.0": 3,
2782
+ "layers.4.feed_forward.w1": 3,
2783
+ "layers.4.feed_forward.w2": 3,
2784
+ "layers.4.feed_forward.w3": 3,
2785
+ "layers.5.attention.to_v": 3,
2786
+ "layers.5.attention.to_out.0": 3,
2787
+ "layers.5.feed_forward.w1": 3,
2788
+ "layers.5.feed_forward.w2": 3,
2789
+ "layers.5.feed_forward.w3": 3,
2790
+ "layers.6.attention.to_v": 3,
2791
+ "layers.6.attention.to_out.0": 3,
2792
+ "layers.6.feed_forward.w1": 3,
2793
+ "layers.6.feed_forward.w2": 3,
2794
+ "layers.6.feed_forward.w3": 3,
2795
+ "layers.7.attention.to_v": 3,
2796
+ "layers.7.attention.to_out.0": 3,
2797
+ "layers.7.feed_forward.w1": 3,
2798
+ "layers.7.feed_forward.w2": 3,
2799
+ "layers.7.feed_forward.w3": 3,
2800
+ "layers.8.attention.to_v": 3,
2801
+ "layers.8.attention.to_out.0": 3,
2802
+ "layers.8.feed_forward.w1": 3,
2803
+ "layers.8.feed_forward.w2": 3,
2804
+ "layers.8.feed_forward.w3": 3,
2805
+ "layers.9.attention.to_v": 3,
2806
+ "layers.9.attention.to_out.0": 3,
2807
+ "layers.9.feed_forward.w1": 3,
2808
+ "layers.9.feed_forward.w2": 3,
2809
+ "layers.9.feed_forward.w3": 3,
2810
+ "layers.10.attention.to_v": 3,
2811
+ "layers.10.attention.to_out.0": 3,
2812
+ "layers.10.feed_forward.w1": 3,
2813
+ "layers.10.feed_forward.w2": 3,
2814
+ "layers.10.feed_forward.w3": 3,
2815
+ "layers.11.attention.to_v": 3,
2816
+ "layers.11.attention.to_out.0": 3,
2817
+ "layers.11.feed_forward.w1": 3,
2818
+ "layers.11.feed_forward.w2": 3,
2819
+ "layers.11.feed_forward.w3": 3,
2820
+ "layers.12.attention.to_v": 3,
2821
+ "layers.12.attention.to_out.0": 3,
2822
+ "layers.12.feed_forward.w1": 3,
2823
+ "layers.12.feed_forward.w2": 3,
2824
+ "layers.12.feed_forward.w3": 3,
2825
+ "layers.13.attention.to_v": 3,
2826
+ "layers.13.attention.to_out.0": 3,
2827
+ "layers.13.feed_forward.w1": 3,
2828
+ "layers.13.feed_forward.w2": 3,
2829
+ "layers.13.feed_forward.w3": 3,
2830
+ "layers.14.attention.to_v": 3,
2831
+ "layers.14.attention.to_out.0": 3,
2832
+ "layers.14.feed_forward.w1": 3,
2833
+ "layers.14.feed_forward.w2": 3,
2834
+ "layers.14.feed_forward.w3": 3,
2835
+ "layers.15.attention.to_v": 3,
2836
+ "layers.15.attention.to_out.0": 3,
2837
+ "layers.15.feed_forward.w1": 3,
2838
+ "layers.15.feed_forward.w2": 3,
2839
+ "layers.15.feed_forward.w3": 3,
2840
+ "layers.16.attention.to_v": 3,
2841
+ "layers.16.attention.to_out.0": 3,
2842
+ "layers.16.feed_forward.w1": 3,
2843
+ "layers.16.feed_forward.w2": 3,
2844
+ "layers.16.feed_forward.w3": 3,
2845
+ "layers.17.attention.to_v": 3,
2846
+ "layers.17.attention.to_out.0": 3,
2847
+ "layers.17.feed_forward.w1": 3,
2848
+ "layers.17.feed_forward.w2": 3,
2849
+ "layers.17.feed_forward.w3": 3,
2850
+ "layers.18.attention.to_v": 3,
2851
+ "layers.18.attention.to_out.0": 3,
2852
+ "layers.18.feed_forward.w1": 3,
2853
+ "layers.18.feed_forward.w2": 3,
2854
+ "layers.18.feed_forward.w3": 3,
2855
+ "layers.19.attention.to_v": 3,
2856
+ "layers.19.attention.to_out.0": 3,
2857
+ "layers.19.feed_forward.w1": 3,
2858
+ "layers.19.feed_forward.w2": 3,
2859
+ "layers.19.feed_forward.w3": 3,
2860
+ "layers.20.attention.to_v": 3,
2861
+ "layers.20.attention.to_out.0": 3,
2862
+ "layers.20.feed_forward.w1": 3,
2863
+ "layers.20.feed_forward.w2": 3,
2864
+ "layers.20.feed_forward.w3": 3,
2865
+ "layers.21.attention.to_v": 3,
2866
+ "layers.21.attention.to_out.0": 3,
2867
+ "layers.21.feed_forward.w1": 3,
2868
+ "layers.21.feed_forward.w2": 3,
2869
+ "layers.21.feed_forward.w3": 3,
2870
+ "layers.22.attention.to_v": 3,
2871
+ "layers.22.attention.to_out.0": 3,
2872
+ "layers.22.feed_forward.w1": 3,
2873
+ "layers.22.feed_forward.w2": 3,
2874
+ "layers.22.feed_forward.w3": 3,
2875
+ "layers.23.attention.to_v": 3,
2876
+ "layers.23.attention.to_out.0": 3,
2877
+ "layers.23.feed_forward.w1": 3,
2878
+ "layers.23.feed_forward.w2": 3,
2879
+ "layers.23.feed_forward.w3": 3,
2880
+ "layers.24.attention.to_v": 3,
2881
+ "layers.24.attention.to_out.0": 3,
2882
+ "layers.24.feed_forward.w1": 3,
2883
+ "layers.24.feed_forward.w2": 3,
2884
+ "layers.24.feed_forward.w3": 3,
2885
+ "layers.25.attention.to_v": 3,
2886
+ "layers.25.attention.to_out.0": 3,
2887
+ "layers.25.feed_forward.w1": 3,
2888
+ "layers.25.feed_forward.w2": 3,
2889
+ "layers.25.feed_forward.w3": 3,
2890
+ "layers.26.attention.to_q": 4,
2891
+ "layers.26.attention.to_k": 4,
2892
+ "layers.26.attention.to_v": 4,
2893
+ "layers.26.attention.to_out.0": 4,
2894
+ "layers.26.feed_forward.w1": 4,
2895
+ "layers.26.feed_forward.w2": 4,
2896
+ "layers.26.feed_forward.w3": 4,
2897
+ "layers.27.attention.to_q": 4,
2898
+ "layers.27.attention.to_k": 4,
2899
+ "layers.27.attention.to_v": 4,
2900
+ "layers.27.attention.to_out.0": 4,
2901
+ "layers.27.feed_forward.w1": 4,
2902
+ "layers.27.feed_forward.w2": 4,
2903
+ "layers.27.feed_forward.w3": 4,
2904
+ "layers.28.attention.to_q": 4,
2905
+ "layers.28.attention.to_k": 4,
2906
+ "layers.28.attention.to_v": 4,
2907
+ "layers.28.attention.to_out.0": 4,
2908
+ "layers.28.feed_forward.w1": 4,
2909
+ "layers.28.feed_forward.w2": 4,
2910
+ "layers.28.feed_forward.w3": 4,
2911
+ "layers.29.attention.to_q": 4,
2912
+ "layers.29.attention.to_k": 4,
2913
+ "layers.29.attention.to_v": 4,
2914
+ "layers.29.attention.to_out.0": 4,
2915
+ "layers.29.feed_forward.w1": 4,
2916
+ "layers.29.feed_forward.w2": 4,
2917
+ "layers.29.feed_forward.w3": 4
2918
+ },
2919
  "checksums": {
2920
+ "assets/image_generation_comparison_matrix.webp": "2ed2129a41b99b6c0f7f1222c590c97a34e5bccac1a2736b360c21dae22fe45e",
2921
+ "benchmark/summary.json": "2268eba92998556016340e097a7113ecb2b249f3690cf4621f67963cae7c4380",
2922
+ "model.safetensors": "667324a633b4e4d92931ca523943c093da46c24472d3f7dc315782ca57847585",
2923
+ "model_index.json": "a954b766fd9af50a2bcb5326e28b48f1f449b21eb2c9bafe21e9ce287192d15e",
2924
+ "orbitquant_codebooks.safetensors": "e7737934089114db3107ba8c16e8c6a882f259fc97d02352cf5f4fb1b0656fe9",
2925
  "orbitquant_rotations.safetensors": "e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825",
2926
+ "prompts.json": "f48aca6d4d681d1ed10dcce8ad43fedb965b4a6a4b11003eb5e37ea53e4eb526",
2927
+ "quantization_config.json": "b80a7d0bbb5779a442d0ac5d6a318caf1a4dc76f76063c2acbd05b1d5bb06cc0"
2928
  }
2929
  }
prompts.json CHANGED
@@ -49,19 +49,19 @@
49
  "id": "cyrillic-text-rendering",
50
  "title": "08 Russian Constructivist typography",
51
  "category": "russian_typography",
52
- "prompt": "A richly detailed Russian Constructivist science-fiction poster with the exact Cyrillic headline \"КВАНТОВАЯ ОРБИТА\", the exact subtitle \"МОСКВА 2049\", and a small exact stamp \"КВАНТОВАНИЕ\"; diagonal red and black geometry, cream paper, cosmonaut portrait, orbital diagrams, halftone grain, folded corners, layered ink, museum archival photograph, all letters crisp and correctly ordered"
53
  },
54
  {
55
  "id": "style-heavy",
56
  "title": "09 Japanese typography and mixed style",
57
  "category": "japanese_typography",
58
- "prompt": "An elaborate Japanese art magazine cover combining Edo woodblock printing with a futuristic Tokyo skyline, with the exact vertical title \"量子の軌道\" and the exact subtitle \"東京の未来\"; giant indigo waves curl around glass towers, red-crowned cranes cross a gold moon, tiny pedestrians and trains fill the lower streets, visible washi fibers, layered spot colors, precise Japanese glyphs, balanced editorial layout"
59
  },
60
  {
61
  "id": "occlusion-reflection",
62
  "title": "10 Chinese typography, reflection, occlusion",
63
  "category": "chinese_typography_reflection_occlusion",
64
- "prompt": "A luxurious Chinese retro-futurist department-store window at night with the exact gold title \"量子轨道\" and the exact red subtitle \"未来之城\"; a curved chrome robot is partly occluded by peonies and blue-and-white porcelain, the calligraphy and neon street must appear coherently reflected across its body, multiple glass layers, silk textures, passing bicycles, cinematic rain, fine product-photography detail"
65
  }
66
  ]
67
  }
 
49
  "id": "cyrillic-text-rendering",
50
  "title": "08 Russian Constructivist typography",
51
  "category": "russian_typography",
52
+ "prompt": "A richly detailed Russian Constructivist science-fiction poster with the exact Cyrillic headline \"\u041a\u0412\u0410\u041d\u0422\u041e\u0412\u0410\u042f \u041e\u0420\u0411\u0418\u0422\u0410\", the exact subtitle \"\u041c\u041e\u0421\u041a\u0412\u0410 2049\", and a small exact stamp \"\u041a\u0412\u0410\u041d\u0422\u041e\u0412\u0410\u041d\u0418\u0415\"; diagonal red and black geometry, cream paper, cosmonaut portrait, orbital diagrams, halftone grain, folded corners, layered ink, museum archival photograph, all letters crisp and correctly ordered"
53
  },
54
  {
55
  "id": "style-heavy",
56
  "title": "09 Japanese typography and mixed style",
57
  "category": "japanese_typography",
58
+ "prompt": "An elaborate Japanese art magazine cover combining Edo woodblock printing with a futuristic Tokyo skyline, with the exact vertical title \"\u91cf\u5b50\u306e\u8ecc\u9053\" and the exact subtitle \"\u6771\u4eac\u306e\u672a\u6765\"; giant indigo waves curl around glass towers, red-crowned cranes cross a gold moon, tiny pedestrians and trains fill the lower streets, visible washi fibers, layered spot colors, precise Japanese glyphs, balanced editorial layout"
59
  },
60
  {
61
  "id": "occlusion-reflection",
62
  "title": "10 Chinese typography, reflection, occlusion",
63
  "category": "chinese_typography_reflection_occlusion",
64
+ "prompt": "A luxurious Chinese retro-futurist department-store window at night with the exact gold title \"\u91cf\u5b50\u8f68\u9053\" and the exact red subtitle \"\u672a\u6765\u4e4b\u57ce\"; a curved chrome robot is partly occluded by peonies and blue-and-white porcelain, the calligraphy and neon street must appear coherently reflected across its body, multiple glass layers, silk textures, passing bicycles, cinematic rain, fine product-photography detail"
65
  }
66
  ]
67
  }
quantization_config.json CHANGED
@@ -15,6 +15,8 @@
15
  "target_policy": "z_image",
16
  "adaln_policy": "int4_rtn",
17
  "adaln_group_size": 64,
 
 
18
  "modules_to_not_convert": [],
19
  "modules_dtype_dict": {},
20
  "artifact_format_version": 1,
@@ -23,5 +25,11 @@
23
  "packed_matmul_block_m": 64,
24
  "packed_matmul_block_n": 64,
25
  "packed_matmul_block_k": 128,
26
- "packed_matmul_num_warps": 4
 
 
 
 
 
 
27
  }
 
15
  "target_policy": "z_image",
16
  "adaln_policy": "int4_rtn",
17
  "adaln_group_size": 64,
18
+ "modules_to_convert": [],
19
+ "modules_to_use_adaln": [],
20
  "modules_to_not_convert": [],
21
  "modules_dtype_dict": {},
22
  "artifact_format_version": 1,
 
25
  "packed_matmul_block_m": 64,
26
  "packed_matmul_block_n": 64,
27
  "packed_matmul_block_k": 128,
28
+ "packed_matmul_num_warps": 4,
29
+ "weight_row_tile_size": 256,
30
+ "w4a4_int8_weight_cache": false,
31
+ "lowbit_boundary_protection": "auto",
32
+ "lowbit_protected_blocks": 4,
33
+ "lowbit_protected_bits": 4,
34
+ "lowbit_interior_protection": true
35
  }