Hugging Face's logo

isek-ai
/
SDPrompt-RetNet-300M

Text Generation
Transformers
PyTorch
Safetensors
English
retnet
Generated from Trainer
custom_code
Model card Files
xet
Community
p1atdev commited on
Commit
2339d5d
·
1 Parent(s): af3e66a

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +243 -0
README.md ADDED
@@ -0,0 +1,243 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - generated_from_trainer
4
+ - retnet
5
+ model-index:
6
+ - name: sdprompt-retnet-300m
7
+ results: []
8
+ license: mit
9
+ datasets:
10
+ - Gustavosta/Stable-Diffusion-Prompts
11
+ - FredZhang7/anime-prompts-180K
12
+ language:
13
+ - en
14
+ library_name: transformers
15
+ pipeline_tag: text-generation
16
+ ---
17
+
18
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
19
+ should probably proofread and complete it, then remove this comment. -->
20
+
21
+ # SDPrompt-RetNet-300M
22
+
23
+ This model is a RetNet model trained from scratch using https://github.com/syncdoth/RetNet.
24
+ It achieves the following results on the evaluation set:
25
+ - Loss: 0.3616
26
+
27
+ ## Usage
28
+
29
+ ```
30
+ pip install transformers safetensors timm
31
+ ```
32
+
33
+ ```py
34
+ from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
35
+
36
+ MODEL_NAME = ""
37
+
38
+ DEVICE = "cuda"
39
+
40
+ tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
41
+ model = AutoModelForCausalLM.from_pretrained(
42
+ MODEL_NAME,
43
+ trust_remote_code=True,
44
+ ).to(DEVICE)
45
+
46
+ streamer = TextStreamer(tokenizer)
47
+
48
+ prompt = "<s>1girl"
49
+
50
+ inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
51
+
52
+ _ = model.generate(
53
+ inputs["input_ids"],
54
+ max_new_tokens=256,
55
+ do_sample=True,
56
+ top_p=0.9,
57
+ top_k=20,
58
+ temperature=0.9,
59
+ streamer=streamer,
60
+ )
61
+ # <s> 1girl, absurdres, animal ear fluff, animal ears, bangs, bare shoulders, black hair, blue archive, blunt bangs, blush, closed mouth, collarbone, commentary request, eyes visible through hair, green eyes, hair between eyes, halo, hand on own face, hand up, highres, jacket, kisaki blue archive, long hair, long sleeves, looking at viewer, open clothes, open jacket, shinonome asu, simple background, solo, track jacket, upper body, white background, white jacket</s>
62
+ ```
63
+
64
+ ## Model description
65
+
66
+ This model is trained with Stable Diffusion prompts and Danbooru tags to generate prompts for image generation models.
67
+
68
+ ## Training data
69
+
70
+ - [Gustavosta/Stable-Diffusion-Prompts](https://huggingface.co/datasets/Gustavosta/Stable-Diffusion-Prompts)
71
+ - [FredZhang7/anime-prompts-180K](https://huggingface.co/datasets/FredZhang7/anime-prompts-180K)
72
+
73
+
74
+ ## Training procedure
75
+
76
+ ### Training hyperparameters
77
+
78
+ The following hyperparameters were used during training:
79
+ - learning_rate: 0.0006
80
+ - train_batch_size: 8
81
+ - eval_batch_size: 8
82
+ - seed: 42
83
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
84
+ - lr_scheduler_type: cosine
85
+ - lr_scheduler_warmup_steps: 500
86
+ - num_epochs: 5
87
+
88
+ ### Training results
89
+
90
+ | Training Loss | Epoch | Step | Validation Loss |
91
+ |:-------------:|:-----:|:------:|:---------------:|
92
+ | 2.6714 | 0.03 | 1000 | 2.5787 |
93
+ | 2.1551 | 0.07 | 2000 | 2.3981 |
94
+ | 2.1439 | 0.1 | 3000 | 2.1160 |
95
+ | 1.8406 | 0.14 | 4000 | 1.9138 |
96
+ | 1.7485 | 0.17 | 5000 | 1.7847 |
97
+ | 1.6417 | 0.21 | 6000 | 1.7120 |
98
+ | 1.6084 | 0.24 | 7000 | 1.6055 |
99
+ | 1.4805 | 0.28 | 8000 | 1.5946 |
100
+ | 1.5524 | 0.31 | 9000 | 1.5027 |
101
+ | 1.4425 | 0.35 | 10000 | 1.4876 |
102
+ | 1.4007 | 0.38 | 11000 | 1.4364 |
103
+ | 1.4637 | 0.42 | 12000 | 1.3896 |
104
+ | 1.3211 | 0.45 | 13000 | 1.3968 |
105
+ | 1.3246 | 0.49 | 14000 | 1.3403 |
106
+ | 1.3461 | 0.52 | 15000 | 1.3156 |
107
+ | 1.2897 | 0.56 | 16000 | 1.2977 |
108
+ | 1.2748 | 0.59 | 17000 | 1.2823 |
109
+ | 1.2424 | 0.62 | 18000 | 1.2649 |
110
+ | 1.348 | 0.66 | 19000 | 1.2134 |
111
+ | 1.1797 | 0.69 | 20000 | 1.2030 |
112
+ | 1.2116 | 0.73 | 21000 | 1.2033 |
113
+ | 1.1702 | 0.76 | 22000 | 1.1453 |
114
+ | 1.1027 | 0.8 | 23000 | 1.1597 |
115
+ | 1.1932 | 0.83 | 24000 | 1.1506 |
116
+ | 1.3669 | 0.87 | 25000 | 1.1428 |
117
+ | 1.0705 | 0.9 | 26000 | 1.1239 |
118
+ | 1.1474 | 0.94 | 27000 | 1.1239 |
119
+ | 1.0879 | 0.97 | 28000 | 1.1168 |
120
+ | 0.9879 | 1.01 | 29000 | 1.0848 |
121
+ | 0.9928 | 1.04 | 30000 | 1.0953 |
122
+ | 0.9095 | 1.08 | 31000 | 1.1043 |
123
+ | 1.0423 | 1.11 | 32000 | 1.0823 |
124
+ | 0.9478 | 1.15 | 33000 | 1.0840 |
125
+ | 0.9979 | 1.18 | 34000 | 1.0387 |
126
+ | 1.0316 | 1.22 | 35000 | 1.0282 |
127
+ | 1.0531 | 1.25 | 36000 | 1.0369 |
128
+ | 0.919 | 1.28 | 37000 | 1.0398 |
129
+ | 1.0596 | 1.32 | 38000 | 1.0410 |
130
+ | 0.9076 | 1.35 | 39000 | 0.9889 |
131
+ | 0.9698 | 1.39 | 40000 | 1.0004 |
132
+ | 0.9633 | 1.42 | 41000 | 1.0038 |
133
+ | 0.9622 | 1.46 | 42000 | 0.9933 |
134
+ | 0.9809 | 1.49 | 43000 | 0.9805 |
135
+ | 0.9496 | 1.53 | 44000 | 0.9755 |
136
+ | 0.9435 | 1.56 | 45000 | 0.9759 |
137
+ | 0.9337 | 1.6 | 46000 | 0.9615 |
138
+ | 0.8844 | 1.63 | 47000 | 0.9524 |
139
+ | 0.9039 | 1.67 | 48000 | 0.9567 |
140
+ | 0.905 | 1.7 | 49000 | 0.9430 |
141
+ | 0.9491 | 1.74 | 50000 | 0.9205 |
142
+ | 0.8464 | 1.77 | 51000 | 0.9109 |
143
+ | 0.9384 | 1.81 | 52000 | 0.9056 |
144
+ | 0.8121 | 1.84 | 53000 | 0.8969 |
145
+ | 0.8381 | 1.88 | 54000 | 0.8869 |
146
+ | 0.8171 | 1.91 | 55000 | 0.8946 |
147
+ | 0.9024 | 1.94 | 56000 | 0.8993 |
148
+ | 0.84 | 1.98 | 57000 | 0.9011 |
149
+ | 0.6702 | 2.01 | 58000 | 0.8876 |
150
+ | 0.6278 | 2.05 | 59000 | 0.8716 |
151
+ | 0.6876 | 2.08 | 60000 | 0.8546 |
152
+ | 0.6754 | 2.12 | 61000 | 0.8639 |
153
+ | 0.6479 | 2.15 | 62000 | 0.8425 |
154
+ | 0.698 | 2.19 | 63000 | 0.8533 |
155
+ | 0.708 | 2.22 | 64000 | 0.8407 |
156
+ | 0.7021 | 2.26 | 65000 | 0.8160 |
157
+ | 0.5881 | 2.29 | 66000 | 0.8251 |
158
+ | 0.6181 | 2.33 | 67000 | 0.8205 |
159
+ | 0.6789 | 2.36 | 68000 | 0.8066 |
160
+ | 0.6452 | 2.4 | 69000 | 0.8037 |
161
+ | 0.6483 | 2.43 | 70000 | 0.7915 |
162
+ | 0.5868 | 2.47 | 71000 | 0.7864 |
163
+ | 0.6257 | 2.5 | 72000 | 0.7895 |
164
+ | 0.6593 | 2.53 | 73000 | 0.7718 |
165
+ | 0.5957 | 2.57 | 74000 | 0.7490 |
166
+ | 0.6351 | 2.6 | 75000 | 0.7481 |
167
+ | 0.699 | 2.64 | 76000 | 0.7628 |
168
+ | 0.566 | 2.67 | 77000 | 0.7590 |
169
+ | 0.5892 | 2.71 | 78000 | 0.7628 |
170
+ | 0.6052 | 2.74 | 79000 | 0.7633 |
171
+ | 0.6494 | 2.78 | 80000 | 0.7588 |
172
+ | 0.5917 | 2.81 | 81000 | 0.7118 |
173
+ | 0.508 | 2.85 | 82000 | 0.6857 |
174
+ | 0.523 | 2.88 | 83000 | 0.6738 |
175
+ | 0.4894 | 2.92 | 84000 | 0.6713 |
176
+ | 0.5096 | 2.95 | 85000 | 0.6625 |
177
+ | 0.352 | 2.99 | 86000 | 0.6802 |
178
+ | 0.3927 | 3.02 | 87000 | 0.6606 |
179
+ | 0.3468 | 3.06 | 88000 | 0.6546 |
180
+ | 0.3368 | 3.09 | 89000 | 0.6520 |
181
+ | 0.352 | 3.12 | 90000 | 0.6495 |
182
+ | 0.3613 | 3.16 | 91000 | 0.6324 |
183
+ | 0.3501 | 3.19 | 92000 | 0.6227 |
184
+ | 0.3269 | 3.23 | 93000 | 0.6091 |
185
+ | 0.3583 | 3.26 | 94000 | 0.6153 |
186
+ | 0.3278 | 3.3 | 95000 | 0.6178 |
187
+ | 0.3216 | 3.33 | 96000 | 0.6208 |
188
+ | 0.3383 | 3.37 | 97000 | 0.6195 |
189
+ | 0.3326 | 3.4 | 98000 | 0.6088 |
190
+ | 0.3081 | 3.44 | 99000 | 0.5956 |
191
+ | 0.3459 | 3.47 | 100000 | 0.5840 |
192
+ | 0.3139 | 3.51 | 101000 | 0.5712 |
193
+ | 0.3087 | 3.54 | 102000 | 0.5677 |
194
+ | 0.2798 | 3.58 | 103000 | 0.5566 |
195
+ | 0.3166 | 3.61 | 104000 | 0.5332 |
196
+ | 0.2981 | 3.65 | 105000 | 0.5333 |
197
+ | 0.3027 | 3.68 | 106000 | 0.5276 |
198
+ | 0.2815 | 3.72 | 107000 | 0.5024 |
199
+ | 0.2294 | 3.75 | 108000 | 0.5081 |
200
+ | 0.2452 | 3.78 | 109000 | 0.4824 |
201
+ | 0.2733 | 3.82 | 110000 | 0.4695 |
202
+ | 0.3001 | 3.85 | 111000 | 0.4627 |
203
+ | 0.2322 | 3.89 | 112000 | 0.4580 |
204
+ | 0.2362 | 3.92 | 113000 | 0.4402 |
205
+ | 0.2488 | 3.96 | 114000 | 0.4263 |
206
+ | 0.2449 | 3.99 | 115000 | 0.3999 |
207
+ | 0.1798 | 4.03 | 116000 | 0.4038 |
208
+ | 0.1956 | 4.06 | 117000 | 0.4037 |
209
+ | 0.1831 | 4.1 | 118000 | 0.4040 |
210
+ | 0.1802 | 4.13 | 119000 | 0.4039 |
211
+ | 0.1641 | 4.17 | 120000 | 0.4029 |
212
+ | 0.1769 | 4.2 | 121000 | 0.4016 |
213
+ | 0.1564 | 4.24 | 122000 | 0.4026 |
214
+ | 0.1552 | 4.27 | 123000 | 0.3988 |
215
+ | 0.1806 | 4.31 | 124000 | 0.3995 |
216
+ | 0.1783 | 4.34 | 125000 | 0.3995 |
217
+ | 0.1736 | 4.38 | 126000 | 0.3940 |
218
+ | 0.1657 | 4.41 | 127000 | 0.3913 |
219
+ | 0.1598 | 4.44 | 128000 | 0.3871 |
220
+ | 0.1599 | 4.48 | 129000 | 0.3831 |
221
+ | 0.1606 | 4.51 | 130000 | 0.3776 |
222
+ | 0.1639 | 4.55 | 131000 | 0.3754 |
223
+ | 0.1736 | 4.58 | 132000 | 0.3742 |
224
+ | 0.1653 | 4.62 | 133000 | 0.3703 |
225
+ | 0.1708 | 4.65 | 134000 | 0.3681 |
226
+ | 0.1729 | 4.69 | 135000 | 0.3674 |
227
+ | 0.1564 | 4.72 | 136000 | 0.3660 |
228
+ | 0.1734 | 4.76 | 137000 | 0.3641 |
229
+ | 0.163 | 4.79 | 138000 | 0.3632 |
230
+ | 0.1585 | 4.83 | 139000 | 0.3626 |
231
+ | 0.1603 | 4.86 | 140000 | 0.3619 |
232
+ | 0.1751 | 4.9 | 141000 | 0.3617 |
233
+ | 0.1622 | 4.93 | 142000 | 0.3617 |
234
+ | 0.161 | 4.97 | 143000 | 0.3617 |
235
+ | 0.1541 | 5.0 | 144000 | 0.3616 |
236
+
237
+
238
+ ### Framework versions
239
+
240
+ - Transformers 4.34.1
241
+ - Pytorch 2.0.0+cu118
242
+ - Datasets 2.14.5
243
+ - Tokenizers 0.14.0