Demosthene-OR
commited on
Commit
·
9a30cab
1
Parent(s):
c1653c1
End of training
Browse files- README.md +306 -205
- config.json +2 -1
- generation_config.json +1 -1
- pytorch_model.bin +1 -1
- tokenizer.json +0 -0
- tokenizer_config.json +0 -1
- training_args.bin +2 -2
README.md
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
3 |
tags:
|
4 |
- generated_from_trainer
|
5 |
metrics:
|
@@ -16,9 +17,9 @@ should probably proofread and complete it, then remove this comment. -->
|
|
16 |
|
17 |
This model is a fine-tuned version of [Demosthene-OR/t5-small-finetuned-en-to-fr](https://huggingface.co/Demosthene-OR/t5-small-finetuned-en-to-fr) on the None dataset.
|
18 |
It achieves the following results on the evaluation set:
|
19 |
-
- Loss: 0.
|
20 |
-
- Bleu:
|
21 |
-
- Gen Len:
|
22 |
|
23 |
## Model description
|
24 |
|
@@ -43,217 +44,317 @@ The following hyperparameters were used during training:
|
|
43 |
- seed: 42
|
44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
45 |
- lr_scheduler_type: linear
|
46 |
-
- num_epochs:
|
47 |
|
48 |
### Training results
|
49 |
|
50 |
| Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
|
51 |
|:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|
|
52 |
-
| No log | 1.0 | 1 |
|
53 |
-
| No log | 2.0 | 2 |
|
54 |
-
| No log | 3.0 | 3 |
|
55 |
-
| No log | 4.0 | 4 |
|
56 |
-
| No log | 5.0 | 5 |
|
57 |
-
| No log | 6.0 | 6 |
|
58 |
-
| No log | 7.0 | 7 |
|
59 |
-
| No log | 8.0 | 8 |
|
60 |
-
| No log | 9.0 | 9 |
|
61 |
-
| No log | 10.0 | 10 |
|
62 |
-
| No log | 11.0 | 11 |
|
63 |
-
| No log | 12.0 | 12 |
|
64 |
-
| No log | 13.0 | 13 |
|
65 |
-
| No log | 14.0 | 14 |
|
66 |
-
| No log | 15.0 | 15 |
|
67 |
-
| No log | 16.0 | 16 |
|
68 |
-
| No log | 17.0 | 17 |
|
69 |
-
| No log | 18.0 | 18 |
|
70 |
-
| No log | 19.0 | 19 |
|
71 |
-
| No log | 20.0 | 20 |
|
72 |
-
| No log | 21.0 | 21 |
|
73 |
-
| No log | 22.0 | 22 |
|
74 |
-
| No log | 23.0 | 23 |
|
75 |
-
| No log | 24.0 | 24 |
|
76 |
-
| No log | 25.0 | 25 |
|
77 |
-
| No log | 26.0 | 26 |
|
78 |
-
| No log | 27.0 | 27 | 0.
|
79 |
-
| No log | 28.0 | 28 | 0.
|
80 |
-
| No log | 29.0 | 29 | 0.
|
81 |
-
| No log | 30.0 | 30 | 0.
|
82 |
-
| No log | 31.0 | 31 | 0.
|
83 |
-
| No log | 32.0 | 32 | 0.
|
84 |
-
| No log | 33.0 | 33 | 0.
|
85 |
-
| No log | 34.0 | 34 | 0.
|
86 |
-
| No log | 35.0 | 35 | 0.
|
87 |
-
| No log | 36.0 | 36 | 0.
|
88 |
-
| No log | 37.0 | 37 | 0.
|
89 |
-
| No log | 38.0 | 38 | 0.
|
90 |
-
| No log | 39.0 | 39 | 0.
|
91 |
-
| No log | 40.0 | 40 | 0.
|
92 |
-
| No log | 41.0 | 41 | 0.
|
93 |
-
| No log | 42.0 | 42 | 0.
|
94 |
-
| No log | 43.0 | 43 | 0.
|
95 |
-
| No log | 44.0 | 44 | 0.
|
96 |
-
| No log | 45.0 | 45 | 0.
|
97 |
-
| No log | 46.0 | 46 | 0.
|
98 |
-
| No log | 47.0 | 47 | 0.
|
99 |
-
| No log | 48.0 | 48 | 0.
|
100 |
-
| No log | 49.0 | 49 | 0.
|
101 |
-
| No log | 50.0 | 50 | 0.
|
102 |
-
| No log | 51.0 | 51 | 0.
|
103 |
-
| No log | 52.0 | 52 | 0.
|
104 |
-
| No log | 53.0 | 53 | 0.
|
105 |
-
| No log | 54.0 | 54 | 0.
|
106 |
-
| No log | 55.0 | 55 | 0.
|
107 |
-
| No log | 56.0 | 56 | 0.
|
108 |
-
| No log | 57.0 | 57 | 0.
|
109 |
-
| No log | 58.0 | 58 | 0.
|
110 |
-
| No log | 59.0 | 59 | 0.
|
111 |
-
| No log | 60.0 | 60 | 0.
|
112 |
-
| No log | 61.0 | 61 | 0.
|
113 |
-
| No log | 62.0 | 62 | 0.
|
114 |
-
| No log | 63.0 | 63 | 0.
|
115 |
-
| No log | 64.0 | 64 | 0.
|
116 |
-
| No log | 65.0 | 65 | 0.
|
117 |
-
| No log | 66.0 | 66 | 0.
|
118 |
-
| No log | 67.0 | 67 | 0.
|
119 |
-
| No log | 68.0 | 68 | 0.
|
120 |
-
| No log | 69.0 | 69 | 0.
|
121 |
-
| No log | 70.0 | 70 | 0.
|
122 |
-
| No log | 71.0 | 71 | 0.
|
123 |
-
| No log | 72.0 | 72 | 0.
|
124 |
-
| No log | 73.0 | 73 | 0.
|
125 |
-
| No log | 74.0 | 74 | 0.
|
126 |
-
| No log | 75.0 | 75 | 0.
|
127 |
-
| No log | 76.0 | 76 | 0.
|
128 |
-
| No log | 77.0 | 77 | 0.
|
129 |
-
| No log | 78.0 | 78 | 0.
|
130 |
-
| No log | 79.0 | 79 | 0.
|
131 |
-
| No log | 80.0 | 80 | 0.
|
132 |
-
| No log | 81.0 | 81 | 0.
|
133 |
-
| No log | 82.0 | 82 | 0.
|
134 |
-
| No log | 83.0 | 83 | 0.
|
135 |
-
| No log | 84.0 | 84 | 0.
|
136 |
-
| No log | 85.0 | 85 | 0.
|
137 |
-
| No log | 86.0 | 86 | 0.
|
138 |
-
| No log | 87.0 | 87 | 0.
|
139 |
-
| No log | 88.0 | 88 | 0.
|
140 |
-
| No log | 89.0 | 89 | 0.
|
141 |
-
| No log | 90.0 | 90 | 0.
|
142 |
-
| No log | 91.0 | 91 | 0.
|
143 |
-
| No log | 92.0 | 92 | 0.
|
144 |
-
| No log | 93.0 | 93 | 0.
|
145 |
-
| No log | 94.0 | 94 | 0.
|
146 |
-
| No log | 95.0 | 95 | 0.
|
147 |
-
| No log | 96.0 | 96 | 0.
|
148 |
-
| No log | 97.0 | 97 | 0.
|
149 |
-
| No log | 98.0 | 98 | 0.
|
150 |
-
| No log | 99.0 | 99 | 0.
|
151 |
-
| No log | 100.0 | 100 | 0.
|
152 |
-
| No log | 101.0 | 101 | 0.
|
153 |
-
| No log | 102.0 | 102 | 0.
|
154 |
-
| No log | 103.0 | 103 | 0.
|
155 |
-
| No log | 104.0 | 104 | 0.
|
156 |
-
| No log | 105.0 | 105 | 0.
|
157 |
-
| No log | 106.0 | 106 | 0.
|
158 |
-
| No log | 107.0 | 107 | 0.
|
159 |
-
| No log | 108.0 | 108 | 0.
|
160 |
-
| No log | 109.0 | 109 | 0.
|
161 |
-
| No log | 110.0 | 110 | 0.
|
162 |
-
| No log | 111.0 | 111 | 0.
|
163 |
-
| No log | 112.0 | 112 | 0.
|
164 |
-
| No log | 113.0 | 113 | 0.
|
165 |
-
| No log | 114.0 | 114 | 0.
|
166 |
-
| No log | 115.0 | 115 | 0.
|
167 |
-
| No log | 116.0 | 116 | 0.
|
168 |
-
| No log | 117.0 | 117 | 0.
|
169 |
-
| No log | 118.0 | 118 | 0.
|
170 |
-
| No log | 119.0 | 119 | 0.
|
171 |
-
| No log | 120.0 | 120 | 0.
|
172 |
-
| No log | 121.0 | 121 | 0.
|
173 |
-
| No log | 122.0 | 122 | 0.
|
174 |
-
| No log | 123.0 | 123 | 0.
|
175 |
-
| No log | 124.0 | 124 | 0.
|
176 |
-
| No log | 125.0 | 125 | 0.
|
177 |
-
| No log | 126.0 | 126 | 0.
|
178 |
-
| No log | 127.0 | 127 | 0.
|
179 |
-
| No log | 128.0 | 128 | 0.
|
180 |
-
| No log | 129.0 | 129 | 0.
|
181 |
-
| No log | 130.0 | 130 | 0.
|
182 |
-
| No log | 131.0 | 131 | 0.
|
183 |
-
| No log | 132.0 | 132 | 0.
|
184 |
-
| No log | 133.0 | 133 | 0.
|
185 |
-
| No log | 134.0 | 134 | 0.
|
186 |
-
| No log | 135.0 | 135 | 0.
|
187 |
-
| No log | 136.0 | 136 | 0.
|
188 |
-
| No log | 137.0 | 137 | 0.
|
189 |
-
| No log | 138.0 | 138 | 0.
|
190 |
-
| No log | 139.0 | 139 | 0.
|
191 |
-
| No log | 140.0 | 140 | 0.
|
192 |
-
| No log | 141.0 | 141 | 0.
|
193 |
-
| No log | 142.0 | 142 | 0.
|
194 |
-
| No log | 143.0 | 143 | 0.
|
195 |
-
| No log | 144.0 | 144 | 0.
|
196 |
-
| No log | 145.0 | 145 | 0.
|
197 |
-
| No log | 146.0 | 146 | 0.
|
198 |
-
| No log | 147.0 | 147 | 0.
|
199 |
-
| No log | 148.0 | 148 | 0.
|
200 |
-
| No log | 149.0 | 149 | 0.
|
201 |
-
| No log | 150.0 | 150 | 0.
|
202 |
-
| No log | 151.0 | 151 | 0.
|
203 |
-
| No log | 152.0 | 152 | 0.
|
204 |
-
| No log | 153.0 | 153 | 0.
|
205 |
-
| No log | 154.0 | 154 | 0.
|
206 |
-
| No log | 155.0 | 155 | 0.
|
207 |
-
| No log | 156.0 | 156 | 0.
|
208 |
-
| No log | 157.0 | 157 | 0.
|
209 |
-
| No log | 158.0 | 158 | 0.
|
210 |
-
| No log | 159.0 | 159 | 0.
|
211 |
-
| No log | 160.0 | 160 | 0.
|
212 |
-
| No log | 161.0 | 161 | 0.
|
213 |
-
| No log | 162.0 | 162 | 0.
|
214 |
-
| No log | 163.0 | 163 | 0.
|
215 |
-
| No log | 164.0 | 164 | 0.
|
216 |
-
| No log | 165.0 | 165 | 0.
|
217 |
-
| No log | 166.0 | 166 | 0.
|
218 |
-
| No log | 167.0 | 167 | 0.
|
219 |
-
| No log | 168.0 | 168 | 0.
|
220 |
-
| No log | 169.0 | 169 | 0.
|
221 |
-
| No log | 170.0 | 170 | 0.
|
222 |
-
| No log | 171.0 | 171 | 0.
|
223 |
-
| No log | 172.0 | 172 | 0.
|
224 |
-
| No log | 173.0 | 173 | 0.
|
225 |
-
| No log | 174.0 | 174 | 0.
|
226 |
-
| No log | 175.0 | 175 | 0.
|
227 |
-
| No log | 176.0 | 176 | 0.
|
228 |
-
| No log | 177.0 | 177 | 0.
|
229 |
-
| No log | 178.0 | 178 | 0.
|
230 |
-
| No log | 179.0 | 179 | 0.
|
231 |
-
| No log | 180.0 | 180 | 0.
|
232 |
-
| No log | 181.0 | 181 | 0.
|
233 |
-
| No log | 182.0 | 182 | 0.
|
234 |
-
| No log | 183.0 | 183 | 0.
|
235 |
-
| No log | 184.0 | 184 | 0.
|
236 |
-
| No log | 185.0 | 185 | 0.
|
237 |
-
| No log | 186.0 | 186 | 0.
|
238 |
-
| No log | 187.0 | 187 | 0.
|
239 |
-
| No log | 188.0 | 188 | 0.
|
240 |
-
| No log | 189.0 | 189 | 0.
|
241 |
-
| No log | 190.0 | 190 | 0.
|
242 |
-
| No log | 191.0 | 191 | 0.
|
243 |
-
| No log | 192.0 | 192 | 0.
|
244 |
-
| No log | 193.0 | 193 | 0.
|
245 |
-
| No log | 194.0 | 194 | 0.
|
246 |
-
| No log | 195.0 | 195 | 0.
|
247 |
-
| No log | 196.0 | 196 | 0.
|
248 |
-
| No log | 197.0 | 197 | 0.
|
249 |
-
| No log | 198.0 | 198 | 0.
|
250 |
-
| No log | 199.0 | 199 | 0.
|
251 |
-
| No log | 200.0 | 200 | 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
252 |
|
253 |
|
254 |
### Framework versions
|
255 |
|
256 |
-
- Transformers 4.
|
257 |
- Pytorch 2.0.1
|
258 |
- Datasets 2.13.0
|
259 |
- Tokenizers 0.13.2
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
base_model: Demosthene-OR/t5-small-finetuned-en-to-fr
|
4 |
tags:
|
5 |
- generated_from_trainer
|
6 |
metrics:
|
|
|
17 |
|
18 |
This model is a fine-tuned version of [Demosthene-OR/t5-small-finetuned-en-to-fr](https://huggingface.co/Demosthene-OR/t5-small-finetuned-en-to-fr) on the None dataset.
|
19 |
It achieves the following results on the evaluation set:
|
20 |
+
- Loss: 0.0770
|
21 |
+
- Bleu: 98.5812
|
22 |
+
- Gen Len: 6.5
|
23 |
|
24 |
## Model description
|
25 |
|
|
|
44 |
- seed: 42
|
45 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
46 |
- lr_scheduler_type: linear
|
47 |
+
- num_epochs: 300
|
48 |
|
49 |
### Training results
|
50 |
|
51 |
| Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
|
52 |
|:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|
|
53 |
+
| No log | 1.0 | 1 | 1.6452 | 69.6781 | 6.3333 |
|
54 |
+
| No log | 2.0 | 2 | 1.6049 | 69.6781 | 6.3333 |
|
55 |
+
| No log | 3.0 | 3 | 1.5656 | 69.6781 | 6.3333 |
|
56 |
+
| No log | 4.0 | 4 | 1.5272 | 69.6781 | 6.3333 |
|
57 |
+
| No log | 5.0 | 5 | 1.4915 | 69.6781 | 6.3333 |
|
58 |
+
| No log | 6.0 | 6 | 1.4568 | 69.6781 | 6.3333 |
|
59 |
+
| No log | 7.0 | 7 | 1.4246 | 69.6781 | 6.3333 |
|
60 |
+
| No log | 8.0 | 8 | 1.3939 | 69.6781 | 6.3333 |
|
61 |
+
| No log | 9.0 | 9 | 1.3640 | 69.6781 | 6.3333 |
|
62 |
+
| No log | 10.0 | 10 | 1.3364 | 82.2802 | 6.1667 |
|
63 |
+
| No log | 11.0 | 11 | 1.3098 | 82.2802 | 6.1667 |
|
64 |
+
| No log | 12.0 | 12 | 1.2831 | 82.2802 | 6.1667 |
|
65 |
+
| No log | 13.0 | 13 | 1.2575 | 82.2802 | 6.1667 |
|
66 |
+
| No log | 14.0 | 14 | 1.2338 | 82.2802 | 6.1667 |
|
67 |
+
| No log | 15.0 | 15 | 1.2106 | 82.2802 | 6.1667 |
|
68 |
+
| No log | 16.0 | 16 | 1.1883 | 82.2802 | 6.1667 |
|
69 |
+
| No log | 17.0 | 17 | 1.1669 | 82.2802 | 6.1667 |
|
70 |
+
| No log | 18.0 | 18 | 1.1467 | 82.2802 | 6.1667 |
|
71 |
+
| No log | 19.0 | 19 | 1.1265 | 82.2802 | 6.1667 |
|
72 |
+
| No log | 20.0 | 20 | 1.1068 | 82.2802 | 6.1667 |
|
73 |
+
| No log | 21.0 | 21 | 1.0883 | 82.1956 | 6.3333 |
|
74 |
+
| No log | 22.0 | 22 | 1.0699 | 82.1956 | 6.3333 |
|
75 |
+
| No log | 23.0 | 23 | 1.0525 | 82.1956 | 6.3333 |
|
76 |
+
| No log | 24.0 | 24 | 1.0361 | 82.1956 | 6.3333 |
|
77 |
+
| No log | 25.0 | 25 | 1.0201 | 82.1956 | 6.3333 |
|
78 |
+
| No log | 26.0 | 26 | 1.0046 | 82.1956 | 6.3333 |
|
79 |
+
| No log | 27.0 | 27 | 0.9891 | 82.1956 | 5.8333 |
|
80 |
+
| No log | 28.0 | 28 | 0.9740 | 82.1956 | 5.8333 |
|
81 |
+
| No log | 29.0 | 29 | 0.9591 | 82.1956 | 5.8333 |
|
82 |
+
| No log | 30.0 | 30 | 0.9448 | 82.1956 | 5.8333 |
|
83 |
+
| No log | 31.0 | 31 | 0.9312 | 82.1956 | 5.8333 |
|
84 |
+
| No log | 32.0 | 32 | 0.9180 | 82.1956 | 5.8333 |
|
85 |
+
| No log | 33.0 | 33 | 0.9056 | 82.1956 | 5.8333 |
|
86 |
+
| No log | 34.0 | 34 | 0.8928 | 82.1956 | 5.8333 |
|
87 |
+
| No log | 35.0 | 35 | 0.8804 | 82.1956 | 5.8333 |
|
88 |
+
| No log | 36.0 | 36 | 0.8684 | 82.1956 | 5.8333 |
|
89 |
+
| No log | 37.0 | 37 | 0.8569 | 82.1956 | 5.8333 |
|
90 |
+
| No log | 38.0 | 38 | 0.8453 | 82.1956 | 5.8333 |
|
91 |
+
| No log | 39.0 | 39 | 0.8339 | 82.1956 | 5.8333 |
|
92 |
+
| No log | 40.0 | 40 | 0.8231 | 82.1956 | 5.8333 |
|
93 |
+
| No log | 41.0 | 41 | 0.8125 | 82.1956 | 5.8333 |
|
94 |
+
| No log | 42.0 | 42 | 0.8018 | 82.1956 | 5.8333 |
|
95 |
+
| No log | 43.0 | 43 | 0.7918 | 82.1956 | 5.8333 |
|
96 |
+
| No log | 44.0 | 44 | 0.7823 | 82.1956 | 5.8333 |
|
97 |
+
| No log | 45.0 | 45 | 0.7731 | 82.1956 | 5.8333 |
|
98 |
+
| No log | 46.0 | 46 | 0.7645 | 82.1956 | 5.8333 |
|
99 |
+
| No log | 47.0 | 47 | 0.7559 | 82.1956 | 5.8333 |
|
100 |
+
| No log | 48.0 | 48 | 0.7475 | 82.1956 | 5.8333 |
|
101 |
+
| No log | 49.0 | 49 | 0.7391 | 87.7051 | 5.8333 |
|
102 |
+
| No log | 50.0 | 50 | 0.7308 | 87.7051 | 5.8333 |
|
103 |
+
| No log | 51.0 | 51 | 0.7227 | 87.7051 | 5.8333 |
|
104 |
+
| No log | 52.0 | 52 | 0.7148 | 87.7051 | 5.8333 |
|
105 |
+
| No log | 53.0 | 53 | 0.7068 | 87.7051 | 5.8333 |
|
106 |
+
| No log | 54.0 | 54 | 0.6992 | 87.7051 | 5.8333 |
|
107 |
+
| No log | 55.0 | 55 | 0.6916 | 87.7051 | 5.8333 |
|
108 |
+
| No log | 56.0 | 56 | 0.6839 | 87.7051 | 5.8333 |
|
109 |
+
| No log | 57.0 | 57 | 0.6765 | 87.7051 | 5.8333 |
|
110 |
+
| No log | 58.0 | 58 | 0.6696 | 87.7051 | 5.8333 |
|
111 |
+
| No log | 59.0 | 59 | 0.6628 | 87.7051 | 5.8333 |
|
112 |
+
| No log | 60.0 | 60 | 0.6559 | 87.7051 | 5.8333 |
|
113 |
+
| No log | 61.0 | 61 | 0.6487 | 87.7051 | 5.8333 |
|
114 |
+
| No log | 62.0 | 62 | 0.6414 | 87.7051 | 5.8333 |
|
115 |
+
| No log | 63.0 | 63 | 0.6342 | 87.7051 | 5.8333 |
|
116 |
+
| No log | 64.0 | 64 | 0.6271 | 87.7051 | 5.8333 |
|
117 |
+
| No log | 65.0 | 65 | 0.6201 | 87.7051 | 5.8333 |
|
118 |
+
| No log | 66.0 | 66 | 0.6132 | 87.7051 | 5.8333 |
|
119 |
+
| No log | 67.0 | 67 | 0.6067 | 87.7051 | 5.8333 |
|
120 |
+
| No log | 68.0 | 68 | 0.6001 | 87.7051 | 5.8333 |
|
121 |
+
| No log | 69.0 | 69 | 0.5935 | 87.7051 | 5.8333 |
|
122 |
+
| No log | 70.0 | 70 | 0.5867 | 87.7051 | 5.8333 |
|
123 |
+
| No log | 71.0 | 71 | 0.5799 | 87.7051 | 5.8333 |
|
124 |
+
| No log | 72.0 | 72 | 0.5733 | 87.7051 | 5.8333 |
|
125 |
+
| No log | 73.0 | 73 | 0.5668 | 87.7051 | 5.8333 |
|
126 |
+
| No log | 74.0 | 74 | 0.5606 | 87.7051 | 5.8333 |
|
127 |
+
| No log | 75.0 | 75 | 0.5546 | 87.7051 | 5.8333 |
|
128 |
+
| No log | 76.0 | 76 | 0.5486 | 87.7051 | 5.8333 |
|
129 |
+
| No log | 77.0 | 77 | 0.5427 | 87.7051 | 5.8333 |
|
130 |
+
| No log | 78.0 | 78 | 0.5368 | 87.7051 | 5.8333 |
|
131 |
+
| No log | 79.0 | 79 | 0.5311 | 87.7051 | 5.8333 |
|
132 |
+
| No log | 80.0 | 80 | 0.5252 | 87.7051 | 5.8333 |
|
133 |
+
| No log | 81.0 | 81 | 0.5192 | 87.7051 | 5.8333 |
|
134 |
+
| No log | 82.0 | 82 | 0.5132 | 87.7051 | 5.8333 |
|
135 |
+
| No log | 83.0 | 83 | 0.5072 | 87.7051 | 5.8333 |
|
136 |
+
| No log | 84.0 | 84 | 0.5011 | 87.7051 | 5.8333 |
|
137 |
+
| No log | 85.0 | 85 | 0.4951 | 87.7051 | 5.8333 |
|
138 |
+
| No log | 86.0 | 86 | 0.4891 | 90.6824 | 5.8333 |
|
139 |
+
| No log | 87.0 | 87 | 0.4834 | 90.6824 | 5.8333 |
|
140 |
+
| No log | 88.0 | 88 | 0.4778 | 90.6824 | 5.8333 |
|
141 |
+
| No log | 89.0 | 89 | 0.4723 | 90.6824 | 5.8333 |
|
142 |
+
| No log | 90.0 | 90 | 0.4670 | 90.6824 | 5.8333 |
|
143 |
+
| No log | 91.0 | 91 | 0.4617 | 90.6824 | 5.8333 |
|
144 |
+
| No log | 92.0 | 92 | 0.4564 | 90.6824 | 5.8333 |
|
145 |
+
| No log | 93.0 | 93 | 0.4510 | 90.6824 | 5.8333 |
|
146 |
+
| No log | 94.0 | 94 | 0.4458 | 90.6824 | 5.8333 |
|
147 |
+
| No log | 95.0 | 95 | 0.4406 | 90.6824 | 5.8333 |
|
148 |
+
| No log | 96.0 | 96 | 0.4356 | 90.6824 | 5.8333 |
|
149 |
+
| No log | 97.0 | 97 | 0.4308 | 90.6824 | 5.8333 |
|
150 |
+
| No log | 98.0 | 98 | 0.4260 | 90.6824 | 5.8333 |
|
151 |
+
| No log | 99.0 | 99 | 0.4214 | 90.6824 | 5.8333 |
|
152 |
+
| No log | 100.0 | 100 | 0.4167 | 90.6824 | 5.8333 |
|
153 |
+
| No log | 101.0 | 101 | 0.4120 | 90.6824 | 5.8333 |
|
154 |
+
| No log | 102.0 | 102 | 0.4073 | 90.6824 | 5.8333 |
|
155 |
+
| No log | 103.0 | 103 | 0.4027 | 90.6824 | 5.8333 |
|
156 |
+
| No log | 104.0 | 104 | 0.3983 | 90.6824 | 5.8333 |
|
157 |
+
| No log | 105.0 | 105 | 0.3941 | 90.6824 | 5.8333 |
|
158 |
+
| No log | 106.0 | 106 | 0.3899 | 90.6824 | 5.8333 |
|
159 |
+
| No log | 107.0 | 107 | 0.3856 | 90.6824 | 5.8333 |
|
160 |
+
| No log | 108.0 | 108 | 0.3814 | 90.6824 | 5.8333 |
|
161 |
+
| No log | 109.0 | 109 | 0.3773 | 90.6824 | 5.8333 |
|
162 |
+
| No log | 110.0 | 110 | 0.3733 | 90.6824 | 5.8333 |
|
163 |
+
| No log | 111.0 | 111 | 0.3694 | 90.6824 | 5.8333 |
|
164 |
+
| No log | 112.0 | 112 | 0.3653 | 90.6824 | 5.8333 |
|
165 |
+
| No log | 113.0 | 113 | 0.3615 | 90.6824 | 5.8333 |
|
166 |
+
| No log | 114.0 | 114 | 0.3577 | 90.6824 | 5.8333 |
|
167 |
+
| No log | 115.0 | 115 | 0.3539 | 90.6824 | 5.8333 |
|
168 |
+
| No log | 116.0 | 116 | 0.3502 | 90.6824 | 5.8333 |
|
169 |
+
| No log | 117.0 | 117 | 0.3468 | 90.6824 | 5.8333 |
|
170 |
+
| No log | 118.0 | 118 | 0.3433 | 90.6824 | 5.8333 |
|
171 |
+
| No log | 119.0 | 119 | 0.3399 | 90.6824 | 5.8333 |
|
172 |
+
| No log | 120.0 | 120 | 0.3365 | 90.6824 | 5.8333 |
|
173 |
+
| No log | 121.0 | 121 | 0.3332 | 90.6824 | 5.8333 |
|
174 |
+
| No log | 122.0 | 122 | 0.3300 | 90.6824 | 5.8333 |
|
175 |
+
| No log | 123.0 | 123 | 0.3269 | 90.6824 | 5.8333 |
|
176 |
+
| No log | 124.0 | 124 | 0.3237 | 90.6824 | 5.8333 |
|
177 |
+
| No log | 125.0 | 125 | 0.3205 | 90.6824 | 5.8333 |
|
178 |
+
| No log | 126.0 | 126 | 0.3172 | 90.6824 | 5.8333 |
|
179 |
+
| No log | 127.0 | 127 | 0.3139 | 90.6824 | 5.8333 |
|
180 |
+
| No log | 128.0 | 128 | 0.3104 | 90.6824 | 5.8333 |
|
181 |
+
| No log | 129.0 | 129 | 0.3070 | 90.6824 | 5.8333 |
|
182 |
+
| No log | 130.0 | 130 | 0.3035 | 90.6824 | 5.8333 |
|
183 |
+
| No log | 131.0 | 131 | 0.2999 | 90.6824 | 5.8333 |
|
184 |
+
| No log | 132.0 | 132 | 0.2963 | 90.6824 | 5.8333 |
|
185 |
+
| No log | 133.0 | 133 | 0.2927 | 90.6824 | 5.8333 |
|
186 |
+
| No log | 134.0 | 134 | 0.2890 | 90.6824 | 5.8333 |
|
187 |
+
| No log | 135.0 | 135 | 0.2855 | 90.6824 | 5.8333 |
|
188 |
+
| No log | 136.0 | 136 | 0.2821 | 90.6824 | 5.8333 |
|
189 |
+
| No log | 137.0 | 137 | 0.2786 | 90.6824 | 5.8333 |
|
190 |
+
| No log | 138.0 | 138 | 0.2752 | 90.6824 | 5.8333 |
|
191 |
+
| No log | 139.0 | 139 | 0.2718 | 90.6824 | 5.8333 |
|
192 |
+
| No log | 140.0 | 140 | 0.2682 | 90.6824 | 5.8333 |
|
193 |
+
| No log | 141.0 | 141 | 0.2648 | 90.6824 | 5.8333 |
|
194 |
+
| No log | 142.0 | 142 | 0.2613 | 90.6824 | 5.8333 |
|
195 |
+
| No log | 143.0 | 143 | 0.2580 | 90.6824 | 5.8333 |
|
196 |
+
| No log | 144.0 | 144 | 0.2548 | 90.6824 | 5.8333 |
|
197 |
+
| No log | 145.0 | 145 | 0.2516 | 90.6824 | 5.8333 |
|
198 |
+
| No log | 146.0 | 146 | 0.2484 | 90.6824 | 5.8333 |
|
199 |
+
| No log | 147.0 | 147 | 0.2452 | 90.6824 | 5.8333 |
|
200 |
+
| No log | 148.0 | 148 | 0.2421 | 90.6824 | 5.8333 |
|
201 |
+
| No log | 149.0 | 149 | 0.2391 | 90.6824 | 5.8333 |
|
202 |
+
| No log | 150.0 | 150 | 0.2362 | 90.6824 | 5.8333 |
|
203 |
+
| No log | 151.0 | 151 | 0.2333 | 90.6824 | 5.8333 |
|
204 |
+
| No log | 152.0 | 152 | 0.2305 | 90.6824 | 5.8333 |
|
205 |
+
| No log | 153.0 | 153 | 0.2278 | 90.6824 | 5.8333 |
|
206 |
+
| No log | 154.0 | 154 | 0.2250 | 90.6824 | 5.8333 |
|
207 |
+
| No log | 155.0 | 155 | 0.2222 | 90.6824 | 5.8333 |
|
208 |
+
| No log | 156.0 | 156 | 0.2193 | 90.6824 | 5.8333 |
|
209 |
+
| No log | 157.0 | 157 | 0.2164 | 90.6824 | 5.8333 |
|
210 |
+
| No log | 158.0 | 158 | 0.2136 | 90.6824 | 5.8333 |
|
211 |
+
| No log | 159.0 | 159 | 0.2108 | 90.6824 | 5.8333 |
|
212 |
+
| No log | 160.0 | 160 | 0.2079 | 90.6824 | 5.8333 |
|
213 |
+
| No log | 161.0 | 161 | 0.2051 | 90.6824 | 5.8333 |
|
214 |
+
| No log | 162.0 | 162 | 0.2024 | 90.6824 | 5.8333 |
|
215 |
+
| No log | 163.0 | 163 | 0.1996 | 90.6824 | 5.8333 |
|
216 |
+
| No log | 164.0 | 164 | 0.1969 | 90.6824 | 5.8333 |
|
217 |
+
| No log | 165.0 | 165 | 0.1942 | 90.6824 | 5.8333 |
|
218 |
+
| No log | 166.0 | 166 | 0.1915 | 90.6824 | 5.8333 |
|
219 |
+
| No log | 167.0 | 167 | 0.1888 | 90.6824 | 5.8333 |
|
220 |
+
| No log | 168.0 | 168 | 0.1862 | 90.6824 | 5.8333 |
|
221 |
+
| No log | 169.0 | 169 | 0.1836 | 98.5812 | 6.5 |
|
222 |
+
| No log | 170.0 | 170 | 0.1811 | 98.5812 | 6.5 |
|
223 |
+
| No log | 171.0 | 171 | 0.1787 | 98.5812 | 6.5 |
|
224 |
+
| No log | 172.0 | 172 | 0.1764 | 98.5812 | 6.5 |
|
225 |
+
| No log | 173.0 | 173 | 0.1742 | 98.5812 | 6.5 |
|
226 |
+
| No log | 174.0 | 174 | 0.1721 | 98.5812 | 6.5 |
|
227 |
+
| No log | 175.0 | 175 | 0.1700 | 98.5812 | 6.5 |
|
228 |
+
| No log | 176.0 | 176 | 0.1680 | 98.5812 | 6.5 |
|
229 |
+
| No log | 177.0 | 177 | 0.1660 | 98.5812 | 6.5 |
|
230 |
+
| No log | 178.0 | 178 | 0.1640 | 98.5812 | 6.5 |
|
231 |
+
| No log | 179.0 | 179 | 0.1620 | 98.5812 | 6.5 |
|
232 |
+
| No log | 180.0 | 180 | 0.1600 | 98.5812 | 6.5 |
|
233 |
+
| No log | 181.0 | 181 | 0.1580 | 98.5812 | 6.5 |
|
234 |
+
| No log | 182.0 | 182 | 0.1560 | 98.5812 | 6.5 |
|
235 |
+
| No log | 183.0 | 183 | 0.1540 | 98.5812 | 6.5 |
|
236 |
+
| No log | 184.0 | 184 | 0.1521 | 98.5812 | 6.5 |
|
237 |
+
| No log | 185.0 | 185 | 0.1502 | 98.5812 | 6.5 |
|
238 |
+
| No log | 186.0 | 186 | 0.1483 | 98.5812 | 6.5 |
|
239 |
+
| No log | 187.0 | 187 | 0.1466 | 98.5812 | 6.5 |
|
240 |
+
| No log | 188.0 | 188 | 0.1448 | 98.5812 | 6.5 |
|
241 |
+
| No log | 189.0 | 189 | 0.1430 | 98.5812 | 6.5 |
|
242 |
+
| No log | 190.0 | 190 | 0.1414 | 98.5812 | 6.5 |
|
243 |
+
| No log | 191.0 | 191 | 0.1398 | 98.5812 | 6.5 |
|
244 |
+
| No log | 192.0 | 192 | 0.1381 | 98.5812 | 6.5 |
|
245 |
+
| No log | 193.0 | 193 | 0.1365 | 98.5812 | 6.5 |
|
246 |
+
| No log | 194.0 | 194 | 0.1349 | 98.5812 | 6.5 |
|
247 |
+
| No log | 195.0 | 195 | 0.1334 | 98.5812 | 6.5 |
|
248 |
+
| No log | 196.0 | 196 | 0.1319 | 98.5812 | 6.5 |
|
249 |
+
| No log | 197.0 | 197 | 0.1304 | 98.5812 | 6.5 |
|
250 |
+
| No log | 198.0 | 198 | 0.1290 | 98.5812 | 6.5 |
|
251 |
+
| No log | 199.0 | 199 | 0.1275 | 98.5812 | 6.5 |
|
252 |
+
| No log | 200.0 | 200 | 0.1261 | 98.5812 | 6.5 |
|
253 |
+
| No log | 201.0 | 201 | 0.1247 | 98.5812 | 6.5 |
|
254 |
+
| No log | 202.0 | 202 | 0.1233 | 98.5812 | 6.5 |
|
255 |
+
| No log | 203.0 | 203 | 0.1220 | 98.5812 | 6.5 |
|
256 |
+
| No log | 204.0 | 204 | 0.1208 | 98.5812 | 6.5 |
|
257 |
+
| No log | 205.0 | 205 | 0.1196 | 98.5812 | 6.5 |
|
258 |
+
| No log | 206.0 | 206 | 0.1184 | 98.5812 | 6.5 |
|
259 |
+
| No log | 207.0 | 207 | 0.1173 | 98.5812 | 6.5 |
|
260 |
+
| No log | 208.0 | 208 | 0.1162 | 98.5812 | 6.5 |
|
261 |
+
| No log | 209.0 | 209 | 0.1152 | 98.5812 | 6.5 |
|
262 |
+
| No log | 210.0 | 210 | 0.1141 | 98.5812 | 6.5 |
|
263 |
+
| No log | 211.0 | 211 | 0.1132 | 98.5812 | 6.5 |
|
264 |
+
| No log | 212.0 | 212 | 0.1123 | 98.5812 | 6.5 |
|
265 |
+
| No log | 213.0 | 213 | 0.1114 | 98.5812 | 6.5 |
|
266 |
+
| No log | 214.0 | 214 | 0.1105 | 98.5812 | 6.5 |
|
267 |
+
| No log | 215.0 | 215 | 0.1096 | 98.5812 | 6.5 |
|
268 |
+
| No log | 216.0 | 216 | 0.1087 | 98.5812 | 6.5 |
|
269 |
+
| No log | 217.0 | 217 | 0.1079 | 98.5812 | 6.5 |
|
270 |
+
| No log | 218.0 | 218 | 0.1071 | 98.5812 | 6.5 |
|
271 |
+
| No log | 219.0 | 219 | 0.1064 | 98.5812 | 6.5 |
|
272 |
+
| No log | 220.0 | 220 | 0.1056 | 98.5812 | 6.5 |
|
273 |
+
| No log | 221.0 | 221 | 0.1049 | 98.5812 | 6.5 |
|
274 |
+
| No log | 222.0 | 222 | 0.1042 | 98.5812 | 6.5 |
|
275 |
+
| No log | 223.0 | 223 | 0.1035 | 98.5812 | 6.5 |
|
276 |
+
| No log | 224.0 | 224 | 0.1028 | 98.5812 | 6.5 |
|
277 |
+
| No log | 225.0 | 225 | 0.1021 | 98.5812 | 6.5 |
|
278 |
+
| No log | 226.0 | 226 | 0.1015 | 98.5812 | 6.5 |
|
279 |
+
| No log | 227.0 | 227 | 0.1008 | 98.5812 | 6.5 |
|
280 |
+
| No log | 228.0 | 228 | 0.1001 | 98.5812 | 6.5 |
|
281 |
+
| No log | 229.0 | 229 | 0.0994 | 98.5812 | 6.5 |
|
282 |
+
| No log | 230.0 | 230 | 0.0987 | 98.5812 | 6.5 |
|
283 |
+
| No log | 231.0 | 231 | 0.0980 | 98.5812 | 6.5 |
|
284 |
+
| No log | 232.0 | 232 | 0.0974 | 98.5812 | 6.5 |
|
285 |
+
| No log | 233.0 | 233 | 0.0968 | 98.5812 | 6.5 |
|
286 |
+
| No log | 234.0 | 234 | 0.0962 | 98.5812 | 6.5 |
|
287 |
+
| No log | 235.0 | 235 | 0.0956 | 98.5812 | 6.5 |
|
288 |
+
| No log | 236.0 | 236 | 0.0950 | 98.5812 | 6.5 |
|
289 |
+
| No log | 237.0 | 237 | 0.0944 | 98.5812 | 6.5 |
|
290 |
+
| No log | 238.0 | 238 | 0.0938 | 98.5812 | 6.5 |
|
291 |
+
| No log | 239.0 | 239 | 0.0932 | 98.5812 | 6.5 |
|
292 |
+
| No log | 240.0 | 240 | 0.0926 | 98.5812 | 6.5 |
|
293 |
+
| No log | 241.0 | 241 | 0.0921 | 98.5812 | 6.5 |
|
294 |
+
| No log | 242.0 | 242 | 0.0916 | 98.5812 | 6.5 |
|
295 |
+
| No log | 243.0 | 243 | 0.0910 | 98.5812 | 6.5 |
|
296 |
+
| No log | 244.0 | 244 | 0.0905 | 98.5812 | 6.5 |
|
297 |
+
| No log | 245.0 | 245 | 0.0900 | 98.5812 | 6.5 |
|
298 |
+
| No log | 246.0 | 246 | 0.0896 | 98.5812 | 6.5 |
|
299 |
+
| No log | 247.0 | 247 | 0.0891 | 98.5812 | 6.5 |
|
300 |
+
| No log | 248.0 | 248 | 0.0887 | 98.5812 | 6.5 |
|
301 |
+
| No log | 249.0 | 249 | 0.0882 | 98.5812 | 6.5 |
|
302 |
+
| No log | 250.0 | 250 | 0.0878 | 98.5812 | 6.5 |
|
303 |
+
| No log | 251.0 | 251 | 0.0873 | 98.5812 | 6.5 |
|
304 |
+
| No log | 252.0 | 252 | 0.0869 | 98.5812 | 6.5 |
|
305 |
+
| No log | 253.0 | 253 | 0.0864 | 98.5812 | 6.5 |
|
306 |
+
| No log | 254.0 | 254 | 0.0860 | 98.5812 | 6.5 |
|
307 |
+
| No log | 255.0 | 255 | 0.0856 | 98.5812 | 6.5 |
|
308 |
+
| No log | 256.0 | 256 | 0.0852 | 98.5812 | 6.5 |
|
309 |
+
| No log | 257.0 | 257 | 0.0848 | 98.5812 | 6.5 |
|
310 |
+
| No log | 258.0 | 258 | 0.0844 | 98.5812 | 6.5 |
|
311 |
+
| No log | 259.0 | 259 | 0.0840 | 98.5812 | 6.5 |
|
312 |
+
| No log | 260.0 | 260 | 0.0836 | 98.5812 | 6.5 |
|
313 |
+
| No log | 261.0 | 261 | 0.0832 | 98.5812 | 6.5 |
|
314 |
+
| No log | 262.0 | 262 | 0.0829 | 98.5812 | 6.5 |
|
315 |
+
| No log | 263.0 | 263 | 0.0826 | 98.5812 | 6.5 |
|
316 |
+
| No log | 264.0 | 264 | 0.0822 | 98.5812 | 6.5 |
|
317 |
+
| No log | 265.0 | 265 | 0.0819 | 98.5812 | 6.5 |
|
318 |
+
| No log | 266.0 | 266 | 0.0816 | 98.5812 | 6.5 |
|
319 |
+
| No log | 267.0 | 267 | 0.0813 | 98.5812 | 6.5 |
|
320 |
+
| No log | 268.0 | 268 | 0.0811 | 98.5812 | 6.5 |
|
321 |
+
| No log | 269.0 | 269 | 0.0808 | 98.5812 | 6.5 |
|
322 |
+
| No log | 270.0 | 270 | 0.0805 | 98.5812 | 6.5 |
|
323 |
+
| No log | 271.0 | 271 | 0.0803 | 98.5812 | 6.5 |
|
324 |
+
| No log | 272.0 | 272 | 0.0801 | 98.5812 | 6.5 |
|
325 |
+
| No log | 273.0 | 273 | 0.0799 | 98.5812 | 6.5 |
|
326 |
+
| No log | 274.0 | 274 | 0.0797 | 98.5812 | 6.5 |
|
327 |
+
| No log | 275.0 | 275 | 0.0795 | 98.5812 | 6.5 |
|
328 |
+
| No log | 276.0 | 276 | 0.0793 | 98.5812 | 6.5 |
|
329 |
+
| No log | 277.0 | 277 | 0.0792 | 98.5812 | 6.5 |
|
330 |
+
| No log | 278.0 | 278 | 0.0790 | 98.5812 | 6.5 |
|
331 |
+
| No log | 279.0 | 279 | 0.0788 | 98.5812 | 6.5 |
|
332 |
+
| No log | 280.0 | 280 | 0.0787 | 98.5812 | 6.5 |
|
333 |
+
| No log | 281.0 | 281 | 0.0785 | 98.5812 | 6.5 |
|
334 |
+
| No log | 282.0 | 282 | 0.0784 | 98.5812 | 6.5 |
|
335 |
+
| No log | 283.0 | 283 | 0.0783 | 98.5812 | 6.5 |
|
336 |
+
| No log | 284.0 | 284 | 0.0781 | 98.5812 | 6.5 |
|
337 |
+
| No log | 285.0 | 285 | 0.0780 | 98.5812 | 6.5 |
|
338 |
+
| No log | 286.0 | 286 | 0.0779 | 98.5812 | 6.5 |
|
339 |
+
| No log | 287.0 | 287 | 0.0778 | 98.5812 | 6.5 |
|
340 |
+
| No log | 288.0 | 288 | 0.0777 | 98.5812 | 6.5 |
|
341 |
+
| No log | 289.0 | 289 | 0.0776 | 98.5812 | 6.5 |
|
342 |
+
| No log | 290.0 | 290 | 0.0775 | 98.5812 | 6.5 |
|
343 |
+
| No log | 291.0 | 291 | 0.0774 | 98.5812 | 6.5 |
|
344 |
+
| No log | 292.0 | 292 | 0.0773 | 98.5812 | 6.5 |
|
345 |
+
| No log | 293.0 | 293 | 0.0772 | 98.5812 | 6.5 |
|
346 |
+
| No log | 294.0 | 294 | 0.0772 | 98.5812 | 6.5 |
|
347 |
+
| No log | 295.0 | 295 | 0.0771 | 98.5812 | 6.5 |
|
348 |
+
| No log | 296.0 | 296 | 0.0771 | 98.5812 | 6.5 |
|
349 |
+
| No log | 297.0 | 297 | 0.0770 | 98.5812 | 6.5 |
|
350 |
+
| No log | 298.0 | 298 | 0.0770 | 98.5812 | 6.5 |
|
351 |
+
| No log | 299.0 | 299 | 0.0770 | 98.5812 | 6.5 |
|
352 |
+
| No log | 300.0 | 300 | 0.0770 | 98.5812 | 6.5 |
|
353 |
|
354 |
|
355 |
### Framework versions
|
356 |
|
357 |
+
- Transformers 4.33.1
|
358 |
- Pytorch 2.0.1
|
359 |
- Datasets 2.13.0
|
360 |
- Tokenizers 0.13.2
|
config.json
CHANGED
@@ -3,6 +3,7 @@
|
|
3 |
"architectures": [
|
4 |
"T5ForConditionalGeneration"
|
5 |
],
|
|
|
6 |
"d_ff": 2048,
|
7 |
"d_kv": 64,
|
8 |
"d_model": 512,
|
@@ -54,7 +55,7 @@
|
|
54 |
}
|
55 |
},
|
56 |
"torch_dtype": "float32",
|
57 |
-
"transformers_version": "4.
|
58 |
"use_cache": true,
|
59 |
"vocab_size": 32128
|
60 |
}
|
|
|
3 |
"architectures": [
|
4 |
"T5ForConditionalGeneration"
|
5 |
],
|
6 |
+
"classifier_dropout": 0.0,
|
7 |
"d_ff": 2048,
|
8 |
"d_kv": 64,
|
9 |
"d_model": 512,
|
|
|
55 |
}
|
56 |
},
|
57 |
"torch_dtype": "float32",
|
58 |
+
"transformers_version": "4.33.1",
|
59 |
"use_cache": true,
|
60 |
"vocab_size": 32128
|
61 |
}
|
generation_config.json
CHANGED
@@ -2,5 +2,5 @@
|
|
2 |
"decoder_start_token_id": 0,
|
3 |
"eos_token_id": 1,
|
4 |
"pad_token_id": 0,
|
5 |
-
"transformers_version": "4.
|
6 |
}
|
|
|
2 |
"decoder_start_token_id": 0,
|
3 |
"eos_token_id": 1,
|
4 |
"pad_token_id": 0,
|
5 |
+
"transformers_version": "4.33.1"
|
6 |
}
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 242069785
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dd54c09bb69fa2fa418d8a221b0c9d621d50131c1b4bed6781d0a84e62be0068
|
3 |
size 242069785
|
tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
tokenizer_config.json
CHANGED
@@ -106,7 +106,6 @@
|
|
106 |
"extra_ids": 100,
|
107 |
"model_max_length": 512,
|
108 |
"pad_token": "<pad>",
|
109 |
-
"sp_model_kwargs": {},
|
110 |
"tokenizer_class": "T5Tokenizer",
|
111 |
"unk_token": "<unk>"
|
112 |
}
|
|
|
106 |
"extra_ids": 100,
|
107 |
"model_max_length": 512,
|
108 |
"pad_token": "<pad>",
|
|
|
109 |
"tokenizer_class": "T5Tokenizer",
|
110 |
"unk_token": "<unk>"
|
111 |
}
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4f0a1610d94c751827418f51e9b6f3bb9e1c18dc6e42454cf61a1b3cb10c7918
|
3 |
+
size 4219
|