ai-forever commited on
Commit
09ab83d
1 Parent(s): a6d30cb

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +30 -31
README.md CHANGED
@@ -12,7 +12,7 @@ metrics:
12
  - f1
13
  library_name: transformers
14
  model-index:
15
- - name: sage-fredt5-large
16
  results:
17
  - task:
18
  type: text-generation
@@ -22,15 +22,15 @@ model-index:
22
  metrics:
23
  - name: F1 (spell)
24
  type: f1_spell
25
- value: 62.2
26
  verified: false
27
  - name: F1 (punct)
28
  type: f1_punct
29
- value: 60.2
30
  verified: false
31
  - name: F1 (case)
32
  type: f1_case
33
- value: 78.1
34
  verified: false
35
  - task:
36
  type: text-generation
@@ -40,15 +40,15 @@ model-index:
40
  metrics:
41
  - name: F1 (spell)
42
  type: f1_spell
43
- value: 46.3
44
  verified: false
45
  - name: F1 (punct)
46
  type: f1_punct
47
- value: 21.6
48
  verified: false
49
  - name: F1 (case)
50
  type: f1_case
51
- value: 34.0
52
  verified: false
53
  - task:
54
  type: text-generation
@@ -58,15 +58,15 @@ model-index:
58
  metrics:
59
  - name: F1 (spell)
60
  type: f1_spell
61
- value: 42.7
62
  verified: false
63
  - name: F1 (punct)
64
  type: f1_punct
65
- value: 15.7
66
  verified: false
67
  - name: F1 (case)
68
  type: f1_case
69
- value: 41.9
70
  verified: false
71
  - task:
72
  type: text-generation
@@ -76,25 +76,25 @@ model-index:
76
  metrics:
77
  - name: F1 (spell)
78
  type: f1_spell
79
- value: 46.3
80
  verified: false
81
  - name: F1 (punct)
82
  type: f1_punct
83
- value: 20.2
84
  verified: false
85
  - name: F1 (case)
86
  type: f1_case
87
- value: 12.6
88
  verified: false
89
  ---
90
- # sage-fredt5-large
91
 
92
  ![banner](images/sage_banner.jpg)
93
 
94
  ## Summary
95
 
96
  The model corrects spelling and punctuation errors and typos by bringing all the words in the text to the norm of the Russian language.
97
- Corrector had been trained based on the model [FRED-T5-large](https://huggingface.co/ai-forever/FRED-T5-large).
98
  An extensive dataset with “artificial” errors was taken as a training corpus: the corpus was assembled on the basis of the Russian-language Wikipedia and transcripts of Russian-language videos, then typos and spelling errors were automatically introduced into it using the library [SAGE](https://github.com/ai-forever/sage).
99
 
100
  ## Public references
@@ -106,9 +106,9 @@ An extensive dataset with “artificial” errors was taken as a training corpus
106
  ## Examples
107
  | Input | Output |
108
  | --- | --- |
109
- | И не чсно прохожим в этот день непогожйи почему я веселый такйо | И не ясно прохожим в этот день непогожий, почему я веселый такой. |
110
- | Каждй день воттак делой, и спена балеть нибудет. А вотак каждый день ниделай | Каждый день вот так делай и спина болеть не будет. А вот так каждый день не делай. |
111
- | Основая цель мероприятия практическая отработка навыков по оказанию помощи гражданам, попавшим в ДТП а также повышение и совершенствование уровня профессиональной подготовки сотрудников МЧС при проведении аварийно-спасательных работ по ликвидации последствий дорожно-транспортных проишествий сокращение временных показателей реагирования. | Основная цель мероприятия практическая отработка навыков по оказанию помощи гражданам, попавшим в ДТП, а также повышение и совершенствование уровня профессиональной подготовки сотрудников МЧС при проведении аварийно-спасательных работ по ликвидации последствий дорожно-транспортных происшествий, сокращение временных показателей реагирования |
112
  | | |
113
 
114
  ## Metrics
@@ -123,8 +123,7 @@ We compare our solution with both open automatic spell checkers and the ChatGPT
123
  **RUSpellRU**
124
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
125
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
126
- | sage-fredt5-large | 57.3 | 68.0 | 62.2 | 86.7 | 46.1 | 60.2 | 92.1 | 67.8 | 78.1 |
127
- | sage-fredt5-large (ft) | 88.4 | 80.9 | 84.5 | 88.2 | 85.3 | 86.8 | 95.5 | 94.0 | 94.7 |
128
  | sage-ai-service | 90.3 | 86.3 | 88.2 | 90.3 | 86.6 | 88.4 | 95.2 | 95.9 | 95.6 |
129
  | gpt-3.5-turbo | 33.6 | 58.5 | 42.7 | 85.9 | 64.6 | 73.7 | 84.9 | 73.9 | 79.0 |
130
  | gpt-4 | 54.9 | 76.7 | 64.0 | 84.0 | 82.3 | 83.2 | 91.5 | 90.2 | 90.9 |
@@ -133,8 +132,7 @@ We compare our solution with both open automatic spell checkers and the ChatGPT
133
  **MultidomainGold**
134
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
135
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
136
- | sage-fredt5-large | 43.4 | 49.7 | 46.3 | 21.8 | 21.3 | 21.6 | 58.8 | 23.9 | 34.0 |
137
- | sage-fredt5-large (ft) | 80.3 | 75.1 | 77.6 | 69.0 | 66.5 | 67.7 | 78.6 | 80.0 | 79.3 |
138
  | sage-ai-service | 81.6 | 77.7 | 79.6 | 70.2 | 67.5 | 68.8 | 80.5 | 80.5 | 80.5 |
139
  | gpt-3.5-turbo | 18.8 | 48.1 | 27.1 | 42.0 | 31.8 | 36.2 | 47.1 | 51.3 | 49.1 |
140
  | gpt-4 | 25.4 | 68.0 | 37.0 | 57.8 | 54.3 | 56.0 | 54.0 | 67.5 | 60.0 |
@@ -143,8 +141,7 @@ We compare our solution with both open automatic spell checkers and the ChatGPT
143
  **MedSpellChecker**
144
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
145
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
146
- | sage-fredt5-large | 35.2 | 54.5 | 42.8 | 19.2 | 13.2 | 15.7 | 48.7 | 36.8 | 41.9 |
147
- | sage-fredt5-large (ft) | 72.5 | 72.2 | 72.3 | 74.6 | 66.4 | 70.3 | 79.3 | 85.1 | 82.1 |
148
  | sage-ai-service | 71.3 | 73.5 | 72.4 | 75.1 | 69.2 | 72.0 | 80.9 | 72.8 | 76.6|
149
  | gpt-3.5-turbo | 14.7 | 45.9 | 22.3 | 69.9 | 52.3 | 59.8 | 26.4 | 41.8 | 32.3 |
150
  | gpt-4 | 37.8 | 72.3 | 49.6 | 81.4 | 64.3 | 71.9 | 73.0 | 62.1 | 67.1 |
@@ -153,18 +150,20 @@ We compare our solution with both open automatic spell checkers and the ChatGPT
153
  **GitHubTypoCorpusRu**
154
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
155
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
156
- | sage-fredt5-large | 46.0 | 46.6 | 46.3 | 22.7 | 18.3 | 20.2 | 12.0 | 13.2 | 12.6 |
157
- | sage-fredt5-large (ft) | 67.5 | 53.2 | 59.5 | 48.5 | 38.0 | 42.6 | 37.3 | 50.0 | 42.7 |
158
  | sage-ai-service | 70.8 | 56.3 | 62.7 | 48.9 | 35.8 | 41.4 | 32.9 | 45.3 | 38.1|
159
  | gpt-3.5-turbo | 23.7 | 38.7 | 29.4 | 37.6 | 23.3 | 28.7 | 19.6 | 35.9 | 25.3 |
160
  | gpt-4 | 27.0 | 52.8 | 35.7 | 45.9 | 32.6 | 38.2 | 25.7 | 36.8 | 30.2 |
161
 
 
162
  ## How to use
163
  ```python
164
  from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
165
- tokenizer = AutoTokenizer.from_pretrained("ai-forever/sage-fredt5-large")
166
- model = AutoModelForSeq2SeqLM.from_pretrained("ai-forever/sage-fredt5-large")
 
167
  model.to("cuda:0")
 
168
  sentence = "И не чсно прохожим в этот день непогожйи почему я веселый такйо"
169
  text = "<LM>" + sentence
170
  with torch.inference_mode():
@@ -178,12 +177,12 @@ with torch.inference_mode():
178
  )
179
  res = res.cpu().tolist()
180
  res = tokenizer.batch_decode(res, skip_special_tokens=True)
 
181
  print(res)
182
- # ["И не ясно прохожим в этот день непогожий, почему я веселый такой."]
183
  ```
184
 
185
  ## Limitations
186
- - The model is intended to be fine-tuned on sets with natural errors for better performance. The realized model is a pre-train and pre-train task is different from the usual spell checking in terms of density of the noise in a corpus and its origin;
187
  - Complex formatting may cause some trouble in output generation.
188
 
189
  ## Resources
@@ -194,7 +193,7 @@ print(res)
194
  - [sage-mt5-large](https://huggingface.co/ai-forever/sage-mt5-large), HuggingFace
195
 
196
  ## License
197
- Model [FRED-T5-large](https://huggingface.co/ai-forever/FRED-T5-large), on the basis of which our solution is made, and its source code are supplied under the MIT license.
198
  Our solution comes with MIT license also.
199
 
200
  ## Specifications
 
12
  - f1
13
  library_name: transformers
14
  model-index:
15
+ - name: sage-fredt5-distilled-95m
16
  results:
17
  - task:
18
  type: text-generation
 
22
  metrics:
23
  - name: F1 (spell)
24
  type: f1_spell
25
+ value: 78.9
26
  verified: false
27
  - name: F1 (punct)
28
  type: f1_punct
29
+ value: 83.6
30
  verified: false
31
  - name: F1 (case)
32
  type: f1_case
33
+ value: 93.5
34
  verified: false
35
  - task:
36
  type: text-generation
 
40
  metrics:
41
  - name: F1 (spell)
42
  type: f1_spell
43
+ value: 73.4
44
  verified: false
45
  - name: F1 (punct)
46
  type: f1_punct
47
+ value: 65.0
48
  verified: false
49
  - name: F1 (case)
50
  type: f1_case
51
+ value: 77.9
52
  verified: false
53
  - task:
54
  type: text-generation
 
58
  metrics:
59
  - name: F1 (spell)
60
  type: f1_spell
61
+ value: 64.9
62
  verified: false
63
  - name: F1 (punct)
64
  type: f1_punct
65
+ value: 70.0
66
  verified: false
67
  - name: F1 (case)
68
  type: f1_case
69
+ value: 68.7
70
  verified: false
71
  - task:
72
  type: text-generation
 
76
  metrics:
77
  - name: F1 (spell)
78
  type: f1_spell
79
+ value: 52.7
80
  verified: false
81
  - name: F1 (punct)
82
  type: f1_punct
83
+ value: 42.1
84
  verified: false
85
  - name: F1 (case)
86
  type: f1_case
87
+ value: 36.3
88
  verified: false
89
  ---
90
+ # sage-fredt5-distilled-95m
91
 
92
  ![banner](images/sage_banner.jpg)
93
 
94
  ## Summary
95
 
96
  The model corrects spelling and punctuation errors and typos by bringing all the words in the text to the norm of the Russian language.
97
+ Corrector is a distilled version of the original model that had been trained based on the [FRED-T5-1.7B](https://huggingface.co/ai-forever/FRED-T5-1.7B) architecture.
98
  An extensive dataset with “artificial” errors was taken as a training corpus: the corpus was assembled on the basis of the Russian-language Wikipedia and transcripts of Russian-language videos, then typos and spelling errors were automatically introduced into it using the library [SAGE](https://github.com/ai-forever/sage).
99
 
100
  ## Public references
 
106
  ## Examples
107
  | Input | Output |
108
  | --- | --- |
109
+ | И не чсно прохожим в этот день непогожйи почему я веселый такйо | И не ясно прохожим в этот день непогожий, почему я весёлый такой? |
110
+ | Каждй день воттак делой, и спена балеть нибудет. А вотак каждый день ниделай | Каждый день вот так делай, и спена болеть не будет. А вот так каждый день ни делай. |
111
+ | Основая цель мероприятия практическая отработка навыков по оказанию помощи гражданам, попавшим в ДТП а также повышение и совершенствование уровня профессиональной подготовки сотрудников МЧС при проведении аварийно-спасательных работ по ликвидации последствий дорожно-транспортных проишествий сокращение временных показателей реагирования. | Основная цель мероприятия - практическая отработка навыков по оказанию помощи гражданам, попавшим в ДТП, а также повышение и совершенствование уровня профессиональной подготовки сотрудников МЧС при проведении аварийно-спасательных работ по ликвидации последствий дорожно-транспортных происшествий, сокращение временных показателей реагирования. |
112
  | | |
113
 
114
  ## Metrics
 
123
  **RUSpellRU**
124
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
125
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
126
+ | sage-fredt5-distilled-95m | 83.5 | 74.8 | 78.9 | 86.8 | 80.6 | 83.6 | 94.4 | 92.5 | 93.5 |
 
127
  | sage-ai-service | 90.3 | 86.3 | 88.2 | 90.3 | 86.6 | 88.4 | 95.2 | 95.9 | 95.6 |
128
  | gpt-3.5-turbo | 33.6 | 58.5 | 42.7 | 85.9 | 64.6 | 73.7 | 84.9 | 73.9 | 79.0 |
129
  | gpt-4 | 54.9 | 76.7 | 64.0 | 84.0 | 82.3 | 83.2 | 91.5 | 90.2 | 90.9 |
 
132
  **MultidomainGold**
133
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
134
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
135
+ | sage-fredt5-distilled-95m | 77.2 | 69.9 | 73.4 | 66.8 | 63.4 | 65.0 | 76.8 | 79.1 | 77.9 |
 
136
  | sage-ai-service | 81.6 | 77.7 | 79.6 | 70.2 | 67.5 | 68.8 | 80.5 | 80.5 | 80.5 |
137
  | gpt-3.5-turbo | 18.8 | 48.1 | 27.1 | 42.0 | 31.8 | 36.2 | 47.1 | 51.3 | 49.1 |
138
  | gpt-4 | 25.4 | 68.0 | 37.0 | 57.8 | 54.3 | 56.0 | 54.0 | 67.5 | 60.0 |
 
141
  **MedSpellChecker**
142
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
143
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
144
+ | sage-fredt5-distilled-95m | 65.1 | 64.8 | 64.9 | 78.6 | 63.1 | 70.0 | 63.5 | 74.7 | 68.7 |
 
145
  | sage-ai-service | 71.3 | 73.5 | 72.4 | 75.1 | 69.2 | 72.0 | 80.9 | 72.8 | 76.6|
146
  | gpt-3.5-turbo | 14.7 | 45.9 | 22.3 | 69.9 | 52.3 | 59.8 | 26.4 | 41.8 | 32.3 |
147
  | gpt-4 | 37.8 | 72.3 | 49.6 | 81.4 | 64.3 | 71.9 | 73.0 | 62.1 | 67.1 |
 
150
  **GitHubTypoCorpusRu**
151
  | Model | Pr. (spell) | Rec. (spell) | F1 (spell) | Pr. (punc) | Rec. (punc) | F1 (punc) | Pr. (case) | Rec. (case) | F1 (case) |
152
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
153
+ | sage-fredt5-distilled-95m | 57.8 | 48.5 | 52.7 | 45.2 | 39.5 | 42.1 | 29.9 | 46.2 | 36.3 |
 
154
  | sage-ai-service | 70.8 | 56.3 | 62.7 | 48.9 | 35.8 | 41.4 | 32.9 | 45.3 | 38.1|
155
  | gpt-3.5-turbo | 23.7 | 38.7 | 29.4 | 37.6 | 23.3 | 28.7 | 19.6 | 35.9 | 25.3 |
156
  | gpt-4 | 27.0 | 52.8 | 35.7 | 45.9 | 32.6 | 38.2 | 25.7 | 36.8 | 30.2 |
157
 
158
+
159
  ## How to use
160
  ```python
161
  from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
162
+
163
+ tokenizer = AutoTokenizer.from_pretrained("ai-forever/sage-fredt5-distilled-95m")
164
+ model = AutoModelForSeq2SeqLM.from_pretrained("ai-forever/sage-fredt5-distilled-95m")
165
  model.to("cuda:0")
166
+
167
  sentence = "И не чсно прохожим в этот день непогожйи почему я веселый такйо"
168
  text = "<LM>" + sentence
169
  with torch.inference_mode():
 
177
  )
178
  res = res.cpu().tolist()
179
  res = tokenizer.batch_decode(res, skip_special_tokens=True)
180
+
181
  print(res)
182
+ # ["И не ясно прохожим в этот день непогожий, почему я весёлый такой?"]
183
  ```
184
 
185
  ## Limitations
 
186
  - Complex formatting may cause some trouble in output generation.
187
 
188
  ## Resources
 
193
  - [sage-mt5-large](https://huggingface.co/ai-forever/sage-mt5-large), HuggingFace
194
 
195
  ## License
196
+ Model [FRED-T5-1.7B](https://huggingface.co/ai-forever/FRED-T5-1.7B), on the basis of which our solution is made, and its source code are supplied under the MIT license.
197
  Our solution comes with MIT license also.
198
 
199
  ## Specifications