Hugging Face's logo

circulus
/
Sana-Sprint-0.6B-ov-int4

Model card Files Community
rippertnt commited on
Commit
479b447
·
verified ·
1 Parent(s): bbe78be

Upload 22 files

Browse files
Files changed (15) hide show
  1. model_index.json +3 -3
  2. scheduler/scheduler_config.json +26 -3
  3. text_encoder/config.json +1 -1
  4. text_encoder/openvino_model.xml +0 -0
  5. tokenizer/openvino_detokenizer.xml +17 -17
  6. tokenizer/openvino_tokenizer.xml +55 -55
  7. transformer/config.json +3 -3
  8. transformer/openvino_model.bin +2 -2
  9. transformer/openvino_model.xml +0 -0
  10. vae_decoder/config.json +1 -1
  11. vae_decoder/openvino_model.bin +1 -1
  12. vae_decoder/openvino_model.xml +0 -0
  13. vae_encoder/config.json +1 -1
  14. vae_encoder/openvino_model.bin +1 -1
  15. vae_encoder/openvino_model.xml +0 -0
model_index.json CHANGED
@@ -1,10 +1,10 @@
1
  {
2
- "_class_name": "SanaSprintPipeline",
3
  "_diffusers_version": "0.33.1",
4
- "_name_or_path": "./sana-sprint-fp16",
5
  "scheduler": [
6
  "diffusers",
7
- "SCMScheduler"
8
  ],
9
  "text_encoder": [
10
  "transformers",
 
1
  {
2
+ "_class_name": "SanaPipeline",
3
  "_diffusers_version": "0.33.1",
4
+ "_name_or_path": "Efficient-Large-Model/Sana_600M_1024px_diffusers",
5
  "scheduler": [
6
  "diffusers",
7
+ "DPMSolverMultistepScheduler"
8
  ],
9
  "text_encoder": [
10
  "transformers",
scheduler/scheduler_config.json CHANGED
@@ -1,7 +1,30 @@
1
  {
2
- "_class_name": "SCMScheduler",
3
  "_diffusers_version": "0.33.1",
 
 
 
 
 
 
 
 
 
 
4
  "num_train_timesteps": 1000,
5
- "prediction_type": "trigflow",
6
- "sigma_data": 0.5
 
 
 
 
 
 
 
 
 
 
 
 
 
7
  }
 
1
  {
2
+ "_class_name": "DPMSolverMultistepScheduler",
3
  "_diffusers_version": "0.33.1",
4
+ "algorithm_type": "dpmsolver++",
5
+ "beta_end": 0.02,
6
+ "beta_schedule": "linear",
7
+ "beta_start": 0.0001,
8
+ "dynamic_thresholding_ratio": 0.995,
9
+ "euler_at_final": false,
10
+ "final_sigmas_type": "zero",
11
+ "flow_shift": 3.0,
12
+ "lambda_min_clipped": -Infinity,
13
+ "lower_order_final": true,
14
  "num_train_timesteps": 1000,
15
+ "prediction_type": "flow_prediction",
16
+ "rescale_betas_zero_snr": false,
17
+ "sample_max_value": 1.0,
18
+ "solver_order": 2,
19
+ "solver_type": "midpoint",
20
+ "steps_offset": 0,
21
+ "thresholding": false,
22
+ "timestep_spacing": "linspace",
23
+ "trained_betas": null,
24
+ "use_beta_sigmas": false,
25
+ "use_exponential_sigmas": false,
26
+ "use_flow_sigmas": true,
27
+ "use_karras_sigmas": false,
28
+ "use_lu_lambdas": false,
29
+ "variance_type": null
30
  }
text_encoder/config.json CHANGED
@@ -29,7 +29,7 @@
29
  "rms_norm_eps": 1e-06,
30
  "rope_theta": 10000.0,
31
  "sliding_window": 4096,
32
- "torch_dtype": "float16",
33
  "transformers_version": "4.51.3",
34
  "use_cache": true,
35
  "vocab_size": 256000
 
29
  "rms_norm_eps": 1e-06,
30
  "rope_theta": 10000.0,
31
  "sliding_window": 4096,
32
+ "torch_dtype": "float32",
33
  "transformers_version": "4.51.3",
34
  "use_cache": true,
35
  "vocab_size": 256000
text_encoder/openvino_model.xml CHANGED
The diff for this file is too large to render. See raw diff
 
tokenizer/openvino_detokenizer.xml CHANGED
@@ -1,16 +1,16 @@
1
  <?xml version="1.0"?>
2
  <net name="detokenizer" version="11">
3
  <layers>
4
- <layer id="0" name="Parameter_396547" type="Parameter" version="opset1">
5
  <data shape="?,?" element_type="i64" />
6
  <output>
7
- <port id="0" precision="I64" names="Parameter_396547">
8
  <dim>-1</dim>
9
  <dim>-1</dim>
10
  </port>
11
  </output>
12
  </layer>
13
- <layer id="1" name="Convert_396717" type="Convert" version="opset1">
14
  <data destination_type="i32" />
15
  <input>
16
  <port id="0" precision="I64">
@@ -25,7 +25,7 @@
25
  </port>
26
  </output>
27
  </layer>
28
- <layer id="2" name="Constant_396502" type="Const" version="opset1">
29
  <data element_type="i32" shape="256000" offset="0" size="1024000" />
30
  <output>
31
  <port id="0" precision="I32">
@@ -33,7 +33,7 @@
33
  </port>
34
  </output>
35
  </layer>
36
- <layer id="3" name="Constant_396504" type="Const" version="opset1">
37
  <data element_type="i32" shape="256000" offset="1024000" size="1024000" />
38
  <output>
39
  <port id="0" precision="I32">
@@ -41,7 +41,7 @@
41
  </port>
42
  </output>
43
  </layer>
44
- <layer id="4" name="Constant_396506" type="Const" version="opset1">
45
  <data element_type="u8" shape="1931902" offset="2048000" size="1931902" />
46
  <output>
47
  <port id="0" precision="U8">
@@ -49,7 +49,7 @@
49
  </port>
50
  </output>
51
  </layer>
52
- <layer id="5" name="Slice_396552" type="Const" version="opset1">
53
  <data element_type="i32" shape="6" offset="3979902" size="24" />
54
  <output>
55
  <port id="0" precision="I32">
@@ -57,7 +57,7 @@
57
  </port>
58
  </output>
59
  </layer>
60
- <layer id="6" name="VocabDecoder_396554" type="VocabDecoder" version="extension">
61
  <data skip_tokens="" />
62
  <input>
63
  <port id="0" precision="I32">
@@ -95,7 +95,7 @@
95
  </port>
96
  </output>
97
  </layer>
98
- <layer id="7" name="Constant_396556" type="Const" version="opset1">
99
  <data element_type="u8" shape="3" offset="3979926" size="3" />
100
  <output>
101
  <port id="0" precision="U8">
@@ -103,7 +103,7 @@
103
  </port>
104
  </output>
105
  </layer>
106
- <layer id="8" name="Constant_396558" type="Const" version="opset1">
107
  <data element_type="u8" shape="1" offset="3979929" size="1" />
108
  <output>
109
  <port id="0" precision="U8">
@@ -111,7 +111,7 @@
111
  </port>
112
  </output>
113
  </layer>
114
- <layer id="9" name="RegexNormalization_396559" type="RegexNormalization" version="extension">
115
  <data global_replace="true" />
116
  <input>
117
  <port id="0" precision="I32">
@@ -142,7 +142,7 @@
142
  </port>
143
  </output>
144
  </layer>
145
- <layer id="10" name="ByteFallback_396560" type="ByteFallback" version="extension">
146
  <input>
147
  <port id="0" precision="I32">
148
  <dim>-1</dim>
@@ -166,7 +166,7 @@
166
  </port>
167
  </output>
168
  </layer>
169
- <layer id="11" name="FuzeRagged_396561" type="FuzeRagged" version="extension">
170
  <input>
171
  <port id="0" precision="I32">
172
  <dim>-1</dim>
@@ -190,7 +190,7 @@
190
  </port>
191
  </output>
192
  </layer>
193
- <layer id="12" name="UTF8Validate_396562" type="UTF8Validate" version="extension">
194
  <data replace_mode="true" />
195
  <input>
196
  <port id="0" precision="I32">
@@ -215,7 +215,7 @@
215
  </port>
216
  </output>
217
  </layer>
218
- <layer id="13" name="StringTensorPack_396563" type="StringTensorPack" version="opset15">
219
  <input>
220
  <port id="0" precision="I32">
221
  <dim>-1</dim>
@@ -228,12 +228,12 @@
228
  </port>
229
  </input>
230
  <output>
231
- <port id="3" precision="STRING" names="Result_396564,string_output">
232
  <dim>-1</dim>
233
  </port>
234
  </output>
235
  </layer>
236
- <layer id="14" name="Result_396564" type="Result" version="opset1" output_names="Result_396564,string_output">
237
  <input>
238
  <port id="0" precision="STRING">
239
  <dim>-1</dim>
 
1
  <?xml version="1.0"?>
2
  <net name="detokenizer" version="11">
3
  <layers>
4
+ <layer id="0" name="Parameter_377777" type="Parameter" version="opset1">
5
  <data shape="?,?" element_type="i64" />
6
  <output>
7
+ <port id="0" precision="I64" names="Parameter_377777">
8
  <dim>-1</dim>
9
  <dim>-1</dim>
10
  </port>
11
  </output>
12
  </layer>
13
+ <layer id="1" name="Convert_377947" type="Convert" version="opset1">
14
  <data destination_type="i32" />
15
  <input>
16
  <port id="0" precision="I64">
 
25
  </port>
26
  </output>
27
  </layer>
28
+ <layer id="2" name="Constant_377732" type="Const" version="opset1">
29
  <data element_type="i32" shape="256000" offset="0" size="1024000" />
30
  <output>
31
  <port id="0" precision="I32">
 
33
  </port>
34
  </output>
35
  </layer>
36
+ <layer id="3" name="Constant_377734" type="Const" version="opset1">
37
  <data element_type="i32" shape="256000" offset="1024000" size="1024000" />
38
  <output>
39
  <port id="0" precision="I32">
 
41
  </port>
42
  </output>
43
  </layer>
44
+ <layer id="4" name="Constant_377736" type="Const" version="opset1">
45
  <data element_type="u8" shape="1931902" offset="2048000" size="1931902" />
46
  <output>
47
  <port id="0" precision="U8">
 
49
  </port>
50
  </output>
51
  </layer>
52
+ <layer id="5" name="Slice_377782" type="Const" version="opset1">
53
  <data element_type="i32" shape="6" offset="3979902" size="24" />
54
  <output>
55
  <port id="0" precision="I32">
 
57
  </port>
58
  </output>
59
  </layer>
60
+ <layer id="6" name="VocabDecoder_377784" type="VocabDecoder" version="extension">
61
  <data skip_tokens="" />
62
  <input>
63
  <port id="0" precision="I32">
 
95
  </port>
96
  </output>
97
  </layer>
98
+ <layer id="7" name="Constant_377786" type="Const" version="opset1">
99
  <data element_type="u8" shape="3" offset="3979926" size="3" />
100
  <output>
101
  <port id="0" precision="U8">
 
103
  </port>
104
  </output>
105
  </layer>
106
+ <layer id="8" name="Constant_377788" type="Const" version="opset1">
107
  <data element_type="u8" shape="1" offset="3979929" size="1" />
108
  <output>
109
  <port id="0" precision="U8">
 
111
  </port>
112
  </output>
113
  </layer>
114
+ <layer id="9" name="RegexNormalization_377789" type="RegexNormalization" version="extension">
115
  <data global_replace="true" />
116
  <input>
117
  <port id="0" precision="I32">
 
142
  </port>
143
  </output>
144
  </layer>
145
+ <layer id="10" name="ByteFallback_377790" type="ByteFallback" version="extension">
146
  <input>
147
  <port id="0" precision="I32">
148
  <dim>-1</dim>
 
166
  </port>
167
  </output>
168
  </layer>
169
+ <layer id="11" name="FuzeRagged_377791" type="FuzeRagged" version="extension">
170
  <input>
171
  <port id="0" precision="I32">
172
  <dim>-1</dim>
 
190
  </port>
191
  </output>
192
  </layer>
193
+ <layer id="12" name="UTF8Validate_377792" type="UTF8Validate" version="extension">
194
  <data replace_mode="true" />
195
  <input>
196
  <port id="0" precision="I32">
 
215
  </port>
216
  </output>
217
  </layer>
218
+ <layer id="13" name="StringTensorPack_377793" type="StringTensorPack" version="opset15">
219
  <input>
220
  <port id="0" precision="I32">
221
  <dim>-1</dim>
 
228
  </port>
229
  </input>
230
  <output>
231
+ <port id="3" precision="STRING" names="Result_377794,string_output">
232
  <dim>-1</dim>
233
  </port>
234
  </output>
235
  </layer>
236
+ <layer id="14" name="Result_377794" type="Result" version="opset1" output_names="Result_377794,string_output">
237
  <input>
238
  <port id="0" precision="STRING">
239
  <dim>-1</dim>
tokenizer/openvino_tokenizer.xml CHANGED
@@ -1,27 +1,27 @@
1
  <?xml version="1.0"?>
2
  <net name="tokenizer" version="11">
3
  <layers>
4
- <layer id="0" name="Parameter_396416" type="Parameter" version="opset1">
5
  <data shape="?" element_type="string" />
6
  <output>
7
- <port id="0" precision="STRING" names="Parameter_396416">
8
  <dim>-1</dim>
9
  </port>
10
  </output>
11
  </layer>
12
- <layer id="1" name="Constant_396531" type="Const" version="opset1">
13
  <data element_type="i32" shape="" offset="0" size="4" />
14
  <output>
15
  <port id="0" precision="I32" />
16
  </output>
17
  </layer>
18
- <layer id="2" name="Constant_396532" type="Const" version="opset1">
19
  <data element_type="i32" shape="" offset="4" size="4" />
20
  <output>
21
  <port id="0" precision="I32" />
22
  </output>
23
  </layer>
24
- <layer id="3" name="Constant_396533" type="Const" version="opset1">
25
  <data element_type="i32" shape="1" offset="8" size="4" />
26
  <output>
27
  <port id="0" precision="I32">
@@ -29,13 +29,13 @@
29
  </port>
30
  </output>
31
  </layer>
32
- <layer id="4" name="Constant_396422" type="Const" version="opset1">
33
  <data element_type="i64" shape="" offset="12" size="8" />
34
  <output>
35
  <port id="0" precision="I64" />
36
  </output>
37
  </layer>
38
- <layer id="5" name="StringTensorUnpack_396417" type="StringTensorUnpack" version="opset15">
39
  <input>
40
  <port id="0" precision="STRING">
41
  <dim>-1</dim>
@@ -53,7 +53,7 @@
53
  </port>
54
  </output>
55
  </layer>
56
- <layer id="6" name="ShapeOf_396418" type="ShapeOf" version="opset3">
57
  <data output_type="i64" />
58
  <input>
59
  <port id="0" precision="I32">
@@ -66,19 +66,19 @@
66
  </port>
67
  </output>
68
  </layer>
69
- <layer id="7" name="Constant_396419" type="Const" version="opset1">
70
  <data element_type="i64" shape="" offset="12" size="8" />
71
  <output>
72
  <port id="0" precision="I64" />
73
  </output>
74
  </layer>
75
- <layer id="8" name="Constant_396420" type="Const" version="opset1">
76
  <data element_type="i64" shape="" offset="12" size="8" />
77
  <output>
78
  <port id="0" precision="I64" />
79
  </output>
80
  </layer>
81
- <layer id="9" name="Gather_396421" type="Gather" version="opset8">
82
  <data batch_dims="0" />
83
  <input>
84
  <port id="0" precision="I64">
@@ -91,13 +91,13 @@
91
  <port id="3" precision="I64" />
92
  </output>
93
  </layer>
94
- <layer id="10" name="Constant_396423" type="Const" version="opset1">
95
  <data element_type="i64" shape="" offset="20" size="8" />
96
  <output>
97
  <port id="0" precision="I64" />
98
  </output>
99
  </layer>
100
- <layer id="11" name="Range_396424" type="Range" version="opset4">
101
  <data output_type="i32" />
102
  <input>
103
  <port id="0" precision="I64" />
@@ -110,19 +110,19 @@
110
  </port>
111
  </output>
112
  </layer>
113
- <layer id="12" name="Constant_396425" type="Const" version="opset1">
114
  <data element_type="i64" shape="" offset="20" size="8" />
115
  <output>
116
  <port id="0" precision="I64" />
117
  </output>
118
  </layer>
119
- <layer id="13" name="Constant_396426" type="Const" version="opset1">
120
  <data element_type="i64" shape="" offset="20" size="8" />
121
  <output>
122
  <port id="0" precision="I64" />
123
  </output>
124
  </layer>
125
- <layer id="14" name="Add_396427" type="Add" version="opset1">
126
  <data auto_broadcast="numpy" />
127
  <input>
128
  <port id="0" precision="I64" />
@@ -132,13 +132,13 @@
132
  <port id="2" precision="I64" />
133
  </output>
134
  </layer>
135
- <layer id="15" name="Constant_396428" type="Const" version="opset1">
136
  <data element_type="i64" shape="" offset="20" size="8" />
137
  <output>
138
  <port id="0" precision="I64" />
139
  </output>
140
  </layer>
141
- <layer id="16" name="Range_396429" type="Range" version="opset4">
142
  <data output_type="i32" />
143
  <input>
144
  <port id="0" precision="I64" />
@@ -151,7 +151,7 @@
151
  </port>
152
  </output>
153
  </layer>
154
- <layer id="17" name="Constant_396491" type="Const" version="opset1">
155
  <data element_type="u8" shape="5900" offset="28" size="5900" />
156
  <output>
157
  <port id="0" precision="U8">
@@ -159,7 +159,7 @@
159
  </port>
160
  </output>
161
  </layer>
162
- <layer id="18" name="SpecialTokensSplit_396492" type="SpecialTokensSplit" version="extension">
163
  <input>
164
  <port id="0" precision="I32">
165
  <dim>-1</dim>
@@ -201,7 +201,7 @@
201
  </port>
202
  </output>
203
  </layer>
204
- <layer id="19" name="Constant_396494" type="Const" version="opset1">
205
  <data element_type="u8" shape="1" offset="5928" size="1" />
206
  <output>
207
  <port id="0" precision="U8">
@@ -209,7 +209,7 @@
209
  </port>
210
  </output>
211
  </layer>
212
- <layer id="20" name="Constant_396496" type="Const" version="opset1">
213
  <data element_type="u8" shape="3" offset="5929" size="3" />
214
  <output>
215
  <port id="0" precision="U8">
@@ -217,7 +217,7 @@
217
  </port>
218
  </output>
219
  </layer>
220
- <layer id="21" name="RegexNormalization_396497" type="RegexNormalization" version="extension">
221
  <data global_replace="true" />
222
  <input>
223
  <port id="0" precision="I32">
@@ -254,7 +254,7 @@
254
  </port>
255
  </output>
256
  </layer>
257
- <layer id="22" name="Constant_396499" type="Const" version="opset1">
258
  <data element_type="u8" shape="1" offset="5928" size="1" />
259
  <output>
260
  <port id="0" precision="U8">
@@ -262,7 +262,7 @@
262
  </port>
263
  </output>
264
  </layer>
265
- <layer id="23" name="RegexSplit_396500" type="RegexSplit" version="extension">
266
  <data behaviour="mergedwithprevious" invert="false" max_splits="-1" />
267
  <input>
268
  <port id="0" precision="I32">
@@ -308,7 +308,7 @@
308
  </port>
309
  </output>
310
  </layer>
311
- <layer id="24" name="Constant_396502" type="Const" version="opset1">
312
  <data element_type="i32" shape="256000" offset="5932" size="1024000" />
313
  <output>
314
  <port id="0" precision="I32">
@@ -316,7 +316,7 @@
316
  </port>
317
  </output>
318
  </layer>
319
- <layer id="25" name="Constant_396504" type="Const" version="opset1">
320
  <data element_type="i32" shape="256000" offset="1029932" size="1024000" />
321
  <output>
322
  <port id="0" precision="I32">
@@ -324,7 +324,7 @@
324
  </port>
325
  </output>
326
  </layer>
327
- <layer id="26" name="Constant_396506" type="Const" version="opset1">
328
  <data element_type="u8" shape="1931902" offset="2053932" size="1931902" />
329
  <output>
330
  <port id="0" precision="U8">
@@ -332,7 +332,7 @@
332
  </port>
333
  </output>
334
  </layer>
335
- <layer id="27" name="Constant_396514" type="Const" version="opset1">
336
  <data element_type="i32" shape="580604" offset="3985834" size="2322416" />
337
  <output>
338
  <port id="0" precision="I32">
@@ -340,7 +340,7 @@
340
  </port>
341
  </output>
342
  </layer>
343
- <layer id="28" name="Constant_396516" type="Const" version="opset1">
344
  <data element_type="i32" shape="580604" offset="6308250" size="2322416" />
345
  <output>
346
  <port id="0" precision="I32">
@@ -348,7 +348,7 @@
348
  </port>
349
  </output>
350
  </layer>
351
- <layer id="29" name="Constant_396518" type="Const" version="opset1">
352
  <data element_type="u8" shape="2709312" offset="8630666" size="2709312" />
353
  <output>
354
  <port id="0" precision="U8">
@@ -356,7 +356,7 @@
356
  </port>
357
  </output>
358
  </layer>
359
- <layer id="30" name="Constant_396520" type="Const" version="opset1">
360
  <data element_type="i32" shape="580604" offset="11339978" size="2322416" />
361
  <output>
362
  <port id="0" precision="I32">
@@ -364,7 +364,7 @@
364
  </port>
365
  </output>
366
  </layer>
367
- <layer id="31" name="Constant_396522" type="Const" version="opset1">
368
  <data element_type="i32" shape="580604" offset="13662394" size="2322416" />
369
  <output>
370
  <port id="0" precision="I32">
@@ -372,7 +372,7 @@
372
  </port>
373
  </output>
374
  </layer>
375
- <layer id="32" name="Constant_396524" type="Const" version="opset1">
376
  <data element_type="u8" shape="1923319" offset="15984810" size="1923319" />
377
  <output>
378
  <port id="0" precision="U8">
@@ -380,7 +380,7 @@
380
  </port>
381
  </output>
382
  </layer>
383
- <layer id="33" name="Constant_396508" type="Const" version="opset1">
384
  <data element_type="i32" shape="248" offset="17908129" size="992" />
385
  <output>
386
  <port id="0" precision="I32">
@@ -388,7 +388,7 @@
388
  </port>
389
  </output>
390
  </layer>
391
- <layer id="34" name="Constant_396510" type="Const" version="opset1">
392
  <data element_type="i32" shape="248" offset="17909121" size="992" />
393
  <output>
394
  <port id="0" precision="I32">
@@ -396,7 +396,7 @@
396
  </port>
397
  </output>
398
  </layer>
399
- <layer id="35" name="Constant_396512" type="Const" version="opset1">
400
  <data element_type="u8" shape="3815" offset="17910113" size="3815" />
401
  <output>
402
  <port id="0" precision="U8">
@@ -404,7 +404,7 @@
404
  </port>
405
  </output>
406
  </layer>
407
- <layer id="36" name="Constant_396525" type="Const" version="opset1">
408
  <data element_type="i32" shape="248" offset="17913928" size="992" />
409
  <output>
410
  <port id="0" precision="I32">
@@ -412,7 +412,7 @@
412
  </port>
413
  </output>
414
  </layer>
415
- <layer id="37" name="BPETokenizer_396526" type="BPETokenizer" version="extension">
416
  <data unk_token="&lt;unk>" fuse_unk="true" suffix_indicator="" end_suffix="" byte_fallback="true" cache_capacity="51200" />
417
  <input>
418
  <port id="0" precision="I32">
@@ -482,7 +482,7 @@
482
  </port>
483
  </output>
484
  </layer>
485
- <layer id="38" name="Subtract_396527" type="Subtract" version="opset1">
486
  <data auto_broadcast="numpy" />
487
  <input>
488
  <port id="0" precision="I32">
@@ -498,13 +498,13 @@
498
  </port>
499
  </output>
500
  </layer>
501
- <layer id="39" name="Constant_396528" type="Const" version="opset1">
502
  <data element_type="i32" shape="" offset="17914920" size="4" />
503
  <output>
504
  <port id="0" precision="I32" />
505
  </output>
506
  </layer>
507
- <layer id="40" name="Minimum_396529" type="Minimum" version="opset1">
508
  <data auto_broadcast="numpy" />
509
  <input>
510
  <port id="0" precision="I32">
@@ -518,7 +518,7 @@
518
  </port>
519
  </output>
520
  </layer>
521
- <layer id="41" name="Add_396530" type="Add" version="opset1">
522
  <data auto_broadcast="numpy" />
523
  <input>
524
  <port id="0" precision="I32">
@@ -534,7 +534,7 @@
534
  </port>
535
  </output>
536
  </layer>
537
- <layer id="42" name="Constant_396534" type="Const" version="opset1">
538
  <data element_type="i32" shape="2" offset="12" size="8" />
539
  <output>
540
  <port id="0" precision="I32">
@@ -542,7 +542,7 @@
542
  </port>
543
  </output>
544
  </layer>
545
- <layer id="43" name="CombineSegments_396535" type="CombineSegments" version="extension">
546
  <input>
547
  <port id="0" precision="I32" />
548
  <port id="1" precision="I32" />
@@ -583,7 +583,7 @@
583
  </port>
584
  </output>
585
  </layer>
586
- <layer id="44" name="Subtract_396536" type="Subtract" version="opset1">
587
  <data auto_broadcast="numpy" />
588
  <input>
589
  <port id="0" precision="I32">
@@ -599,13 +599,13 @@
599
  </port>
600
  </output>
601
  </layer>
602
- <layer id="45" name="Constant_396537" type="Const" version="opset1">
603
  <data element_type="i32" shape="" offset="0" size="4" />
604
  <output>
605
  <port id="0" precision="I32" />
606
  </output>
607
  </layer>
608
- <layer id="46" name="ReduceMax_396538" type="ReduceMax" version="opset1">
609
  <data keep_dims="false" />
610
  <input>
611
  <port id="0" precision="I32">
@@ -617,13 +617,13 @@
617
  <port id="2" precision="I32" />
618
  </output>
619
  </layer>
620
- <layer id="47" name="Constant_396539" type="Const" version="opset1">
621
  <data element_type="i32" shape="" offset="0" size="4" />
622
  <output>
623
  <port id="0" precision="I32" />
624
  </output>
625
  </layer>
626
- <layer id="48" name="RaggedToDense_396540" type="RaggedToDense" version="extension">
627
  <data pad_right="false" m_pad_max_length="false" />
628
  <input>
629
  <port id="0" precision="I32">
@@ -649,7 +649,7 @@
649
  </port>
650
  </output>
651
  </layer>
652
- <layer id="49" name="Convert_396541" type="Convert" version="opset1">
653
  <data destination_type="i32" />
654
  <input>
655
  <port id="0" precision="BOOL">
@@ -664,7 +664,7 @@
664
  </port>
665
  </output>
666
  </layer>
667
- <layer id="50" name="Convert_396541.0" type="Convert" version="opset1">
668
  <data destination_type="i64" />
669
  <input>
670
  <port id="0" precision="I32">
@@ -679,7 +679,7 @@
679
  </port>
680
  </output>
681
  </layer>
682
- <layer id="52" name="RaggedToDense_396540.0" type="Convert" version="opset1">
683
  <data destination_type="i64" />
684
  <input>
685
  <port id="0" precision="I32">
@@ -694,7 +694,7 @@
694
  </port>
695
  </output>
696
  </layer>
697
- <layer id="53" name="Result_396544" type="Result" version="opset1" output_names="input_ids">
698
  <input>
699
  <port id="0" precision="I64">
700
  <dim>-1</dim>
@@ -702,7 +702,7 @@
702
  </port>
703
  </input>
704
  </layer>
705
- <layer id="51" name="Result_396546" type="Result" version="opset1" output_names="attention_mask">
706
  <input>
707
  <port id="0" precision="I64">
708
  <dim>-1</dim>
 
1
  <?xml version="1.0"?>
2
  <net name="tokenizer" version="11">
3
  <layers>
4
+ <layer id="0" name="Parameter_377646" type="Parameter" version="opset1">
5
  <data shape="?" element_type="string" />
6
  <output>
7
+ <port id="0" precision="STRING" names="Parameter_377646">
8
  <dim>-1</dim>
9
  </port>
10
  </output>
11
  </layer>
12
+ <layer id="1" name="Constant_377761" type="Const" version="opset1">
13
  <data element_type="i32" shape="" offset="0" size="4" />
14
  <output>
15
  <port id="0" precision="I32" />
16
  </output>
17
  </layer>
18
+ <layer id="2" name="Constant_377762" type="Const" version="opset1">
19
  <data element_type="i32" shape="" offset="4" size="4" />
20
  <output>
21
  <port id="0" precision="I32" />
22
  </output>
23
  </layer>
24
+ <layer id="3" name="Constant_377763" type="Const" version="opset1">
25
  <data element_type="i32" shape="1" offset="8" size="4" />
26
  <output>
27
  <port id="0" precision="I32">
 
29
  </port>
30
  </output>
31
  </layer>
32
+ <layer id="4" name="Constant_377652" type="Const" version="opset1">
33
  <data element_type="i64" shape="" offset="12" size="8" />
34
  <output>
35
  <port id="0" precision="I64" />
36
  </output>
37
  </layer>
38
+ <layer id="5" name="StringTensorUnpack_377647" type="StringTensorUnpack" version="opset15">
39
  <input>
40
  <port id="0" precision="STRING">
41
  <dim>-1</dim>
 
53
  </port>
54
  </output>
55
  </layer>
56
+ <layer id="6" name="ShapeOf_377648" type="ShapeOf" version="opset3">
57
  <data output_type="i64" />
58
  <input>
59
  <port id="0" precision="I32">
 
66
  </port>
67
  </output>
68
  </layer>
69
+ <layer id="7" name="Constant_377649" type="Const" version="opset1">
70
  <data element_type="i64" shape="" offset="12" size="8" />
71
  <output>
72
  <port id="0" precision="I64" />
73
  </output>
74
  </layer>
75
+ <layer id="8" name="Constant_377650" type="Const" version="opset1">
76
  <data element_type="i64" shape="" offset="12" size="8" />
77
  <output>
78
  <port id="0" precision="I64" />
79
  </output>
80
  </layer>
81
+ <layer id="9" name="Gather_377651" type="Gather" version="opset8">
82
  <data batch_dims="0" />
83
  <input>
84
  <port id="0" precision="I64">
 
91
  <port id="3" precision="I64" />
92
  </output>
93
  </layer>
94
+ <layer id="10" name="Constant_377653" type="Const" version="opset1">
95
  <data element_type="i64" shape="" offset="20" size="8" />
96
  <output>
97
  <port id="0" precision="I64" />
98
  </output>
99
  </layer>
100
+ <layer id="11" name="Range_377654" type="Range" version="opset4">
101
  <data output_type="i32" />
102
  <input>
103
  <port id="0" precision="I64" />
 
110
  </port>
111
  </output>
112
  </layer>
113
+ <layer id="12" name="Constant_377655" type="Const" version="opset1">
114
  <data element_type="i64" shape="" offset="20" size="8" />
115
  <output>
116
  <port id="0" precision="I64" />
117
  </output>
118
  </layer>
119
+ <layer id="13" name="Constant_377656" type="Const" version="opset1">
120
  <data element_type="i64" shape="" offset="20" size="8" />
121
  <output>
122
  <port id="0" precision="I64" />
123
  </output>
124
  </layer>
125
+ <layer id="14" name="Add_377657" type="Add" version="opset1">
126
  <data auto_broadcast="numpy" />
127
  <input>
128
  <port id="0" precision="I64" />
 
132
  <port id="2" precision="I64" />
133
  </output>
134
  </layer>
135
+ <layer id="15" name="Constant_377658" type="Const" version="opset1">
136
  <data element_type="i64" shape="" offset="20" size="8" />
137
  <output>
138
  <port id="0" precision="I64" />
139
  </output>
140
  </layer>
141
+ <layer id="16" name="Range_377659" type="Range" version="opset4">
142
  <data output_type="i32" />
143
  <input>
144
  <port id="0" precision="I64" />
 
151
  </port>
152
  </output>
153
  </layer>
154
+ <layer id="17" name="Constant_377721" type="Const" version="opset1">
155
  <data element_type="u8" shape="5900" offset="28" size="5900" />
156
  <output>
157
  <port id="0" precision="U8">
 
159
  </port>
160
  </output>
161
  </layer>
162
+ <layer id="18" name="SpecialTokensSplit_377722" type="SpecialTokensSplit" version="extension">
163
  <input>
164
  <port id="0" precision="I32">
165
  <dim>-1</dim>
 
201
  </port>
202
  </output>
203
  </layer>
204
+ <layer id="19" name="Constant_377724" type="Const" version="opset1">
205
  <data element_type="u8" shape="1" offset="5928" size="1" />
206
  <output>
207
  <port id="0" precision="U8">
 
209
  </port>
210
  </output>
211
  </layer>
212
+ <layer id="20" name="Constant_377726" type="Const" version="opset1">
213
  <data element_type="u8" shape="3" offset="5929" size="3" />
214
  <output>
215
  <port id="0" precision="U8">
 
217
  </port>
218
  </output>
219
  </layer>
220
+ <layer id="21" name="RegexNormalization_377727" type="RegexNormalization" version="extension">
221
  <data global_replace="true" />
222
  <input>
223
  <port id="0" precision="I32">
 
254
  </port>
255
  </output>
256
  </layer>
257
+ <layer id="22" name="Constant_377729" type="Const" version="opset1">
258
  <data element_type="u8" shape="1" offset="5928" size="1" />
259
  <output>
260
  <port id="0" precision="U8">
 
262
  </port>
263
  </output>
264
  </layer>
265
+ <layer id="23" name="RegexSplit_377730" type="RegexSplit" version="extension">
266
  <data behaviour="mergedwithprevious" invert="false" max_splits="-1" />
267
  <input>
268
  <port id="0" precision="I32">
 
308
  </port>
309
  </output>
310
  </layer>
311
+ <layer id="24" name="Constant_377732" type="Const" version="opset1">
312
  <data element_type="i32" shape="256000" offset="5932" size="1024000" />
313
  <output>
314
  <port id="0" precision="I32">
 
316
  </port>
317
  </output>
318
  </layer>
319
+ <layer id="25" name="Constant_377734" type="Const" version="opset1">
320
  <data element_type="i32" shape="256000" offset="1029932" size="1024000" />
321
  <output>
322
  <port id="0" precision="I32">
 
324
  </port>
325
  </output>
326
  </layer>
327
+ <layer id="26" name="Constant_377736" type="Const" version="opset1">
328
  <data element_type="u8" shape="1931902" offset="2053932" size="1931902" />
329
  <output>
330
  <port id="0" precision="U8">
 
332
  </port>
333
  </output>
334
  </layer>
335
+ <layer id="27" name="Constant_377744" type="Const" version="opset1">
336
  <data element_type="i32" shape="580604" offset="3985834" size="2322416" />
337
  <output>
338
  <port id="0" precision="I32">
 
340
  </port>
341
  </output>
342
  </layer>
343
+ <layer id="28" name="Constant_377746" type="Const" version="opset1">
344
  <data element_type="i32" shape="580604" offset="6308250" size="2322416" />
345
  <output>
346
  <port id="0" precision="I32">
 
348
  </port>
349
  </output>
350
  </layer>
351
+ <layer id="29" name="Constant_377748" type="Const" version="opset1">
352
  <data element_type="u8" shape="2709312" offset="8630666" size="2709312" />
353
  <output>
354
  <port id="0" precision="U8">
 
356
  </port>
357
  </output>
358
  </layer>
359
+ <layer id="30" name="Constant_377750" type="Const" version="opset1">
360
  <data element_type="i32" shape="580604" offset="11339978" size="2322416" />
361
  <output>
362
  <port id="0" precision="I32">
 
364
  </port>
365
  </output>
366
  </layer>
367
+ <layer id="31" name="Constant_377752" type="Const" version="opset1">
368
  <data element_type="i32" shape="580604" offset="13662394" size="2322416" />
369
  <output>
370
  <port id="0" precision="I32">
 
372
  </port>
373
  </output>
374
  </layer>
375
+ <layer id="32" name="Constant_377754" type="Const" version="opset1">
376
  <data element_type="u8" shape="1923319" offset="15984810" size="1923319" />
377
  <output>
378
  <port id="0" precision="U8">
 
380
  </port>
381
  </output>
382
  </layer>
383
+ <layer id="33" name="Constant_377738" type="Const" version="opset1">
384
  <data element_type="i32" shape="248" offset="17908129" size="992" />
385
  <output>
386
  <port id="0" precision="I32">
 
388
  </port>
389
  </output>
390
  </layer>
391
+ <layer id="34" name="Constant_377740" type="Const" version="opset1">
392
  <data element_type="i32" shape="248" offset="17909121" size="992" />
393
  <output>
394
  <port id="0" precision="I32">
 
396
  </port>
397
  </output>
398
  </layer>
399
+ <layer id="35" name="Constant_377742" type="Const" version="opset1">
400
  <data element_type="u8" shape="3815" offset="17910113" size="3815" />
401
  <output>
402
  <port id="0" precision="U8">
 
404
  </port>
405
  </output>
406
  </layer>
407
+ <layer id="36" name="Constant_377755" type="Const" version="opset1">
408
  <data element_type="i32" shape="248" offset="17913928" size="992" />
409
  <output>
410
  <port id="0" precision="I32">
 
412
  </port>
413
  </output>
414
  </layer>
415
+ <layer id="37" name="BPETokenizer_377756" type="BPETokenizer" version="extension">
416
  <data unk_token="&lt;unk>" fuse_unk="true" suffix_indicator="" end_suffix="" byte_fallback="true" cache_capacity="51200" />
417
  <input>
418
  <port id="0" precision="I32">
 
482
  </port>
483
  </output>
484
  </layer>
485
+ <layer id="38" name="Subtract_377757" type="Subtract" version="opset1">
486
  <data auto_broadcast="numpy" />
487
  <input>
488
  <port id="0" precision="I32">
 
498
  </port>
499
  </output>
500
  </layer>
501
+ <layer id="39" name="Constant_377758" type="Const" version="opset1">
502
  <data element_type="i32" shape="" offset="17914920" size="4" />
503
  <output>
504
  <port id="0" precision="I32" />
505
  </output>
506
  </layer>
507
+ <layer id="40" name="Minimum_377759" type="Minimum" version="opset1">
508
  <data auto_broadcast="numpy" />
509
  <input>
510
  <port id="0" precision="I32">
 
518
  </port>
519
  </output>
520
  </layer>
521
+ <layer id="41" name="Add_377760" type="Add" version="opset1">
522
  <data auto_broadcast="numpy" />
523
  <input>
524
  <port id="0" precision="I32">
 
534
  </port>
535
  </output>
536
  </layer>
537
+ <layer id="42" name="Constant_377764" type="Const" version="opset1">
538
  <data element_type="i32" shape="2" offset="12" size="8" />
539
  <output>
540
  <port id="0" precision="I32">
 
542
  </port>
543
  </output>
544
  </layer>
545
+ <layer id="43" name="CombineSegments_377765" type="CombineSegments" version="extension">
546
  <input>
547
  <port id="0" precision="I32" />
548
  <port id="1" precision="I32" />
 
583
  </port>
584
  </output>
585
  </layer>
586
+ <layer id="44" name="Subtract_377766" type="Subtract" version="opset1">
587
  <data auto_broadcast="numpy" />
588
  <input>
589
  <port id="0" precision="I32">
 
599
  </port>
600
  </output>
601
  </layer>
602
+ <layer id="45" name="Constant_377767" type="Const" version="opset1">
603
  <data element_type="i32" shape="" offset="0" size="4" />
604
  <output>
605
  <port id="0" precision="I32" />
606
  </output>
607
  </layer>
608
+ <layer id="46" name="ReduceMax_377768" type="ReduceMax" version="opset1">
609
  <data keep_dims="false" />
610
  <input>
611
  <port id="0" precision="I32">
 
617
  <port id="2" precision="I32" />
618
  </output>
619
  </layer>
620
+ <layer id="47" name="Constant_377769" type="Const" version="opset1">
621
  <data element_type="i32" shape="" offset="0" size="4" />
622
  <output>
623
  <port id="0" precision="I32" />
624
  </output>
625
  </layer>
626
+ <layer id="48" name="RaggedToDense_377770" type="RaggedToDense" version="extension">
627
  <data pad_right="false" m_pad_max_length="false" />
628
  <input>
629
  <port id="0" precision="I32">
 
649
  </port>
650
  </output>
651
  </layer>
652
+ <layer id="49" name="Convert_377771" type="Convert" version="opset1">
653
  <data destination_type="i32" />
654
  <input>
655
  <port id="0" precision="BOOL">
 
664
  </port>
665
  </output>
666
  </layer>
667
+ <layer id="50" name="Convert_377771.0" type="Convert" version="opset1">
668
  <data destination_type="i64" />
669
  <input>
670
  <port id="0" precision="I32">
 
679
  </port>
680
  </output>
681
  </layer>
682
+ <layer id="52" name="RaggedToDense_377770.0" type="Convert" version="opset1">
683
  <data destination_type="i64" />
684
  <input>
685
  <port id="0" precision="I32">
 
694
  </port>
695
  </output>
696
  </layer>
697
+ <layer id="53" name="Result_377774" type="Result" version="opset1" output_names="input_ids">
698
  <input>
699
  <port id="0" precision="I64">
700
  <dim>-1</dim>
 
702
  </port>
703
  </input>
704
  </layer>
705
+ <layer id="51" name="Result_377776" type="Result" version="opset1" output_names="attention_mask">
706
  <input>
707
  <port id="0" precision="I64">
708
  <dim>-1</dim>
transformer/config.json CHANGED
@@ -1,14 +1,14 @@
1
  {
2
  "_class_name": "SanaTransformer2DModel",
3
  "_diffusers_version": "0.33.1",
4
- "_name_or_path": "./sana-sprint-fp16/transformer",
5
  "attention_bias": false,
6
  "attention_head_dim": 32,
7
  "caption_channels": 2304,
8
  "cross_attention_dim": 1152,
9
  "cross_attention_head_dim": 72,
10
  "dropout": 0.0,
11
- "guidance_embeds": true,
12
  "guidance_embeds_scale": 0.1,
13
  "in_channels": 32,
14
  "interpolation_scale": null,
@@ -20,7 +20,7 @@
20
  "num_layers": 28,
21
  "out_channels": 32,
22
  "patch_size": 1,
23
- "qk_norm": "rms_norm_across_heads",
24
  "sample_size": 32,
25
  "timestep_scale": 1.0
26
  }
 
1
  {
2
  "_class_name": "SanaTransformer2DModel",
3
  "_diffusers_version": "0.33.1",
4
+ "_name_or_path": "/home/circulus/.cache/huggingface/hub/models--Efficient-Large-Model--Sana_600M_1024px_diffusers/snapshots/b3ddc862873249650ed9f6ba90e92ede3083bb0d/transformer",
5
  "attention_bias": false,
6
  "attention_head_dim": 32,
7
  "caption_channels": 2304,
8
  "cross_attention_dim": 1152,
9
  "cross_attention_head_dim": 72,
10
  "dropout": 0.0,
11
+ "guidance_embeds": false,
12
  "guidance_embeds_scale": 0.1,
13
  "in_channels": 32,
14
  "interpolation_scale": null,
 
20
  "num_layers": 28,
21
  "out_channels": 32,
22
  "patch_size": 1,
23
+ "qk_norm": null,
24
  "sample_size": 32,
25
  "timestep_scale": 1.0
26
  }
transformer/openvino_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f0014249d7c6fbd4c2dc23e4344c452f7f96624d9581e75654332393aa893284
3
- size 441494820
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f3d99303acaaf663759e0361f9008b0ab592f6186110178baa381b754488fa4
3
+ size 440152740
transformer/openvino_model.xml CHANGED
The diff for this file is too large to render. See raw diff
 
vae_decoder/config.json CHANGED
@@ -1,7 +1,7 @@
1
  {
2
  "_class_name": "AutoencoderDC",
3
  "_diffusers_version": "0.33.1",
4
- "_name_or_path": "./sana-sprint-fp16/vae",
5
  "attention_head_dim": 32,
6
  "decoder_act_fns": "silu",
7
  "decoder_block_out_channels": [
 
1
  {
2
  "_class_name": "AutoencoderDC",
3
  "_diffusers_version": "0.33.1",
4
+ "_name_or_path": "/home/circulus/.cache/huggingface/hub/models--Efficient-Large-Model--Sana_600M_1024px_diffusers/snapshots/b3ddc862873249650ed9f6ba90e92ede3083bb0d/vae",
5
  "attention_head_dim": 32,
6
  "decoder_act_fns": "silu",
7
  "decoder_block_out_channels": [
vae_decoder/openvino_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1e97183b7503e0e1433b69449af4f007190b65e9bfaedd1a029b31c1c02429fc
3
  size 142920449
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab2f7b0e61a05bb37f83c0552b307cde1bb07d682167917c640f36a1ca18c290
3
  size 142920449
vae_decoder/openvino_model.xml CHANGED
The diff for this file is too large to render. See raw diff
 
vae_encoder/config.json CHANGED
@@ -1,7 +1,7 @@
1
  {
2
  "_class_name": "AutoencoderDC",
3
  "_diffusers_version": "0.33.1",
4
- "_name_or_path": "./sana-sprint-fp16/vae",
5
  "attention_head_dim": 32,
6
  "decoder_act_fns": "silu",
7
  "decoder_block_out_channels": [
 
1
  {
2
  "_class_name": "AutoencoderDC",
3
  "_diffusers_version": "0.33.1",
4
+ "_name_or_path": "/home/circulus/.cache/huggingface/hub/models--Efficient-Large-Model--Sana_600M_1024px_diffusers/snapshots/b3ddc862873249650ed9f6ba90e92ede3083bb0d/vae",
5
  "attention_head_dim": 32,
6
  "decoder_act_fns": "silu",
7
  "decoder_block_out_channels": [
vae_encoder/openvino_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:254009d2fe532b1d66a4d41af0a16fad438f8251f6f1578edaa3ccacc99a40bb
3
  size 137497040
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67068f378d671194e9831d6755865567c5f44729942fca0523c622da57fab168
3
  size 137497040
vae_encoder/openvino_model.xml CHANGED
The diff for this file is too large to render. See raw diff