Hugging Face's logo

distributed
/
optimized-gpt2-1b

Text Generation
Transformers
Safetensors
gpt_optimized
custom_code
Model card Files Community
kmfoda commited on
Commit
ef9535d
·
verified ·
1 Parent(s): f9b718d

Epoch 683. Batch Size 194014. Peers 58.

Browse files
Files changed (2) hide show
  1. config.json +93 -93
  2. model.safetensors +1 -1
config.json CHANGED
@@ -4,40 +4,40 @@
4
  "all_reduce_scores": {
5
  "0": "NON_PARTICIPATING",
6
  "1": "NON_PARTICIPATING",
7
- "10": "SUCCESS",
8
  "100": "NON_PARTICIPATING",
9
- "101": "SUCCESS",
10
  "102": "NON_PARTICIPATING",
11
  "103": "NON_PARTICIPATING",
12
- "104": "NON_PARTICIPATING",
13
- "105": "NON_PARTICIPATING",
14
- "106": "SUCCESS",
15
- "107": "NON_PARTICIPATING",
16
  "108": "NON_PARTICIPATING",
17
- "109": "NON_PARTICIPATING",
18
- "11": "SUCCESS",
19
- "110": "SUCCESS",
20
- "111": "NON_PARTICIPATING",
21
  "112": "NON_PARTICIPATING",
22
  "113": "NON_PARTICIPATING",
23
  "114": "NON_PARTICIPATING",
24
- "115": "SUCCESS",
25
- "116": "NON_PARTICIPATING",
26
- "117": "SUCCESS",
27
- "118": "NON_PARTICIPATING",
28
- "119": "NON_PARTICIPATING",
29
- "12": "SUCCESS",
30
  "120": "NON_PARTICIPATING",
31
- "121": "NON_PARTICIPATING",
32
- "122": "NON_PARTICIPATING",
33
  "123": "NON_PARTICIPATING",
34
  "124": "NON_PARTICIPATING",
35
  "125": "NON_PARTICIPATING",
36
  "126": "NON_PARTICIPATING",
37
  "127": "NON_PARTICIPATING",
38
  "128": "NON_PARTICIPATING",
39
- "129": "SUCCESS",
40
- "13": "SUCCESS",
41
  "130": "NON_PARTICIPATING",
42
  "131": "NON_PARTICIPATING",
43
  "132": "NON_PARTICIPATING",
@@ -50,7 +50,7 @@
50
  "139": "NON_PARTICIPATING",
51
  "14": "NON_PARTICIPATING",
52
  "140": "NON_PARTICIPATING",
53
- "141": "NON_PARTICIPATING",
54
  "142": "SUCCESS",
55
  "143": "SUCCESS",
56
  "144": "NON_PARTICIPATING",
@@ -65,30 +65,30 @@
65
  "152": "NON_PARTICIPATING",
66
  "153": "NON_PARTICIPATING",
67
  "154": "NON_PARTICIPATING",
68
- "155": "NON_PARTICIPATING",
69
- "156": "SUCCESS",
70
- "157": "NON_PARTICIPATING",
71
- "158": "NON_PARTICIPATING",
72
- "159": "SUCCESS",
73
- "16": "SUCCESS",
74
  "160": "NON_PARTICIPATING",
75
  "161": "NON_PARTICIPATING",
76
  "162": "NON_PARTICIPATING",
77
- "163": "NON_PARTICIPATING",
78
  "164": "SUCCESS",
79
  "165": "NON_PARTICIPATING",
80
  "166": "NON_PARTICIPATING",
81
  "167": "NON_PARTICIPATING",
82
  "168": "NON_PARTICIPATING",
83
- "169": "NON_PARTICIPATING",
84
  "17": "NON_PARTICIPATING",
85
  "170": "NON_PARTICIPATING",
86
  "171": "NON_PARTICIPATING",
87
  "172": "NON_PARTICIPATING",
88
  "173": "NON_PARTICIPATING",
89
- "174": "NON_PARTICIPATING",
90
- "175": "NON_PARTICIPATING",
91
- "176": "NON_PARTICIPATING",
92
  "177": "NON_PARTICIPATING",
93
  "178": "SUCCESS",
94
  "179": "NON_PARTICIPATING",
@@ -97,44 +97,44 @@
97
  "181": "NON_PARTICIPATING",
98
  "182": "NON_PARTICIPATING",
99
  "183": "NON_PARTICIPATING",
100
- "184": "NON_PARTICIPATING",
101
  "185": "NON_PARTICIPATING",
102
- "186": "NON_PARTICIPATING",
103
  "187": "NON_PARTICIPATING",
104
- "188": "NON_PARTICIPATING",
105
- "189": "NON_PARTICIPATING",
106
  "19": "NON_PARTICIPATING",
107
- "190": "NON_PARTICIPATING",
108
  "191": "NON_PARTICIPATING",
109
  "192": "NON_PARTICIPATING",
110
- "193": "NON_PARTICIPATING",
111
  "194": "NON_PARTICIPATING",
112
- "195": "SUCCESS",
113
  "196": "NON_PARTICIPATING",
114
  "197": "NON_PARTICIPATING",
115
  "198": "NON_PARTICIPATING",
116
  "199": "NON_PARTICIPATING",
117
- "2": "SUCCESS",
118
  "20": "NON_PARTICIPATING",
119
  "200": "NON_PARTICIPATING",
120
- "201": "SUCCESS",
121
  "202": "NON_PARTICIPATING",
122
  "203": "NON_PARTICIPATING",
123
- "204": "SUCCESS",
124
  "205": "NON_PARTICIPATING",
125
  "206": "NON_PARTICIPATING",
126
  "207": "NON_PARTICIPATING",
127
- "208": "SUCCESS",
128
- "209": "SUCCESS",
129
  "21": "NON_PARTICIPATING",
130
  "210": "NON_PARTICIPATING",
131
  "211": "NON_PARTICIPATING",
132
  "212": "NON_PARTICIPATING",
133
- "213": "SUCCESS",
134
  "214": "NON_PARTICIPATING",
135
  "215": "NON_PARTICIPATING",
136
- "216": "SUCCESS",
137
- "217": "SUCCESS",
138
  "218": "NON_PARTICIPATING",
139
  "219": "NON_PARTICIPATING",
140
  "22": "NON_PARTICIPATING",
@@ -142,57 +142,57 @@
142
  "221": "NON_PARTICIPATING",
143
  "222": "NON_PARTICIPATING",
144
  "223": "NON_PARTICIPATING",
145
- "224": "SUCCESS",
146
- "225": "SUCCESS",
147
- "226": "NON_PARTICIPATING",
148
  "227": "SUCCESS",
149
  "228": "NON_PARTICIPATING",
150
- "229": "SUCCESS",
151
  "23": "NON_PARTICIPATING",
152
- "230": "SUCCESS",
153
  "231": "NON_PARTICIPATING",
154
  "232": "NON_PARTICIPATING",
155
  "233": "NON_PARTICIPATING",
156
  "234": "NON_PARTICIPATING",
157
  "235": "NON_PARTICIPATING",
158
- "236": "SUCCESS",
159
  "237": "NON_PARTICIPATING",
160
  "238": "NON_PARTICIPATING",
161
- "239": "SUCCESS",
162
  "24": "NON_PARTICIPATING",
163
- "240": "SUCCESS",
164
  "241": "NON_PARTICIPATING",
165
- "242": "NON_PARTICIPATING",
166
- "243": "NON_PARTICIPATING",
167
- "244": "SUCCESS",
168
  "245": "NON_PARTICIPATING",
169
  "246": "NON_PARTICIPATING",
170
  "247": "NON_PARTICIPATING",
171
  "248": "NON_PARTICIPATING",
172
  "249": "NON_PARTICIPATING",
173
- "25": "NON_PARTICIPATING",
174
  "250": "NON_PARTICIPATING",
175
  "251": "NON_PARTICIPATING",
176
- "252": "SUCCESS",
177
  "253": "NON_PARTICIPATING",
178
  "254": "NON_PARTICIPATING",
179
- "255": "NON_PARTICIPATING",
180
  "26": "NON_PARTICIPATING",
181
  "27": "NON_PARTICIPATING",
182
  "28": "NON_PARTICIPATING",
183
  "29": "NON_PARTICIPATING",
184
  "3": "NON_PARTICIPATING",
185
- "30": "NON_PARTICIPATING",
186
- "31": "NON_PARTICIPATING",
187
- "32": "NON_PARTICIPATING",
188
  "33": "NON_PARTICIPATING",
189
- "34": "NON_PARTICIPATING",
190
  "35": "NON_PARTICIPATING",
191
- "36": "NON_PARTICIPATING",
192
- "37": "SUCCESS",
193
  "38": "NON_PARTICIPATING",
194
- "39": "NON_PARTICIPATING",
195
- "4": "SUCCESS",
196
  "40": "NON_PARTICIPATING",
197
  "41": "NON_PARTICIPATING",
198
  "42": "NON_PARTICIPATING",
@@ -201,62 +201,62 @@
201
  "45": "NON_PARTICIPATING",
202
  "46": "NON_PARTICIPATING",
203
  "47": "NON_PARTICIPATING",
204
- "48": "SUCCESS",
205
  "49": "NON_PARTICIPATING",
206
- "5": "NON_PARTICIPATING",
207
  "50": "NON_PARTICIPATING",
208
  "51": "NON_PARTICIPATING",
209
  "52": "NON_PARTICIPATING",
210
  "53": "NON_PARTICIPATING",
211
- "54": "NON_PARTICIPATING",
212
- "55": "SUCCESS",
213
  "56": "NON_PARTICIPATING",
214
  "57": "NON_PARTICIPATING",
215
- "58": "SUCCESS",
216
  "59": "NON_PARTICIPATING",
217
- "6": "NON_PARTICIPATING",
218
  "60": "NON_PARTICIPATING",
219
  "61": "NON_PARTICIPATING",
220
- "62": "SUCCESS",
221
- "63": "SUCCESS",
222
  "64": "NON_PARTICIPATING",
223
- "65": "SUCCESS",
224
- "66": "NON_PARTICIPATING",
225
  "67": "NON_PARTICIPATING",
226
  "68": "NON_PARTICIPATING",
227
  "69": "NON_PARTICIPATING",
228
- "7": "SUCCESS",
229
  "70": "NON_PARTICIPATING",
230
- "71": "NON_PARTICIPATING",
231
- "72": "NON_PARTICIPATING",
232
  "73": "SUCCESS",
233
- "74": "SUCCESS",
234
  "75": "NON_PARTICIPATING",
235
- "76": "NON_PARTICIPATING",
236
  "77": "NON_PARTICIPATING",
237
- "78": "NON_PARTICIPATING",
238
  "79": "NON_PARTICIPATING",
239
  "8": "NON_PARTICIPATING",
240
- "80": "SUCCESS",
241
  "81": "NON_PARTICIPATING",
242
  "82": "NON_PARTICIPATING",
243
- "83": "NON_PARTICIPATING",
244
  "84": "NON_PARTICIPATING",
245
  "85": "SUCCESS",
246
- "86": "NON_PARTICIPATING",
247
  "87": "NON_PARTICIPATING",
248
- "88": "SUCCESS",
249
- "89": "NON_PARTICIPATING",
250
- "9": "NON_PARTICIPATING",
251
  "90": "NON_PARTICIPATING",
252
  "91": "NON_PARTICIPATING",
253
  "92": "NON_PARTICIPATING",
254
  "93": "SUCCESS",
255
  "94": "SUCCESS",
256
  "95": "NON_PARTICIPATING",
257
- "96": "SUCCESS",
258
  "97": "NON_PARTICIPATING",
259
- "98": "SUCCESS",
260
  "99": "NON_PARTICIPATING"
261
  },
262
  "architectures": [
 
4
  "all_reduce_scores": {
5
  "0": "NON_PARTICIPATING",
6
  "1": "NON_PARTICIPATING",
7
+ "10": "NON_PARTICIPATING",
8
  "100": "NON_PARTICIPATING",
9
+ "101": "NON_PARTICIPATING",
10
  "102": "NON_PARTICIPATING",
11
  "103": "NON_PARTICIPATING",
12
+ "104": "SUCCESS",
13
+ "105": "SUCCESS",
14
+ "106": "NON_PARTICIPATING",
15
+ "107": "SUCCESS",
16
  "108": "NON_PARTICIPATING",
17
+ "109": "SUCCESS",
18
+ "11": "NON_PARTICIPATING",
19
+ "110": "NON_PARTICIPATING",
20
+ "111": "SUCCESS",
21
  "112": "NON_PARTICIPATING",
22
  "113": "NON_PARTICIPATING",
23
  "114": "NON_PARTICIPATING",
24
+ "115": "NON_PARTICIPATING",
25
+ "116": "SUCCESS",
26
+ "117": "NON_PARTICIPATING",
27
+ "118": "SUCCESS",
28
+ "119": "SUCCESS",
29
+ "12": "NON_PARTICIPATING",
30
  "120": "NON_PARTICIPATING",
31
+ "121": "SUCCESS",
32
+ "122": "SUCCESS",
33
  "123": "NON_PARTICIPATING",
34
  "124": "NON_PARTICIPATING",
35
  "125": "NON_PARTICIPATING",
36
  "126": "NON_PARTICIPATING",
37
  "127": "NON_PARTICIPATING",
38
  "128": "NON_PARTICIPATING",
39
+ "129": "NON_PARTICIPATING",
40
+ "13": "NON_PARTICIPATING",
41
  "130": "NON_PARTICIPATING",
42
  "131": "NON_PARTICIPATING",
43
  "132": "NON_PARTICIPATING",
 
50
  "139": "NON_PARTICIPATING",
51
  "14": "NON_PARTICIPATING",
52
  "140": "NON_PARTICIPATING",
53
+ "141": "SUCCESS",
54
  "142": "SUCCESS",
55
  "143": "SUCCESS",
56
  "144": "NON_PARTICIPATING",
 
65
  "152": "NON_PARTICIPATING",
66
  "153": "NON_PARTICIPATING",
67
  "154": "NON_PARTICIPATING",
68
+ "155": "SUCCESS",
69
+ "156": "NON_PARTICIPATING",
70
+ "157": "SUCCESS",
71
+ "158": "SUCCESS",
72
+ "159": "NON_PARTICIPATING",
73
+ "16": "NON_PARTICIPATING",
74
  "160": "NON_PARTICIPATING",
75
  "161": "NON_PARTICIPATING",
76
  "162": "NON_PARTICIPATING",
77
+ "163": "SUCCESS",
78
  "164": "SUCCESS",
79
  "165": "NON_PARTICIPATING",
80
  "166": "NON_PARTICIPATING",
81
  "167": "NON_PARTICIPATING",
82
  "168": "NON_PARTICIPATING",
83
+ "169": "SUCCESS",
84
  "17": "NON_PARTICIPATING",
85
  "170": "NON_PARTICIPATING",
86
  "171": "NON_PARTICIPATING",
87
  "172": "NON_PARTICIPATING",
88
  "173": "NON_PARTICIPATING",
89
+ "174": "SUCCESS",
90
+ "175": "SUCCESS",
91
+ "176": "SUCCESS",
92
  "177": "NON_PARTICIPATING",
93
  "178": "SUCCESS",
94
  "179": "NON_PARTICIPATING",
 
97
  "181": "NON_PARTICIPATING",
98
  "182": "NON_PARTICIPATING",
99
  "183": "NON_PARTICIPATING",
100
+ "184": "SUCCESS",
101
  "185": "NON_PARTICIPATING",
102
+ "186": "SUCCESS",
103
  "187": "NON_PARTICIPATING",
104
+ "188": "SUCCESS",
105
+ "189": "SUCCESS",
106
  "19": "NON_PARTICIPATING",
107
+ "190": "SUCCESS",
108
  "191": "NON_PARTICIPATING",
109
  "192": "NON_PARTICIPATING",
110
+ "193": "SUCCESS",
111
  "194": "NON_PARTICIPATING",
112
+ "195": "NON_PARTICIPATING",
113
  "196": "NON_PARTICIPATING",
114
  "197": "NON_PARTICIPATING",
115
  "198": "NON_PARTICIPATING",
116
  "199": "NON_PARTICIPATING",
117
+ "2": "NON_PARTICIPATING",
118
  "20": "NON_PARTICIPATING",
119
  "200": "NON_PARTICIPATING",
120
+ "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
  "203": "NON_PARTICIPATING",
123
+ "204": "NON_PARTICIPATING",
124
  "205": "NON_PARTICIPATING",
125
  "206": "NON_PARTICIPATING",
126
  "207": "NON_PARTICIPATING",
127
+ "208": "NON_PARTICIPATING",
128
+ "209": "NON_PARTICIPATING",
129
  "21": "NON_PARTICIPATING",
130
  "210": "NON_PARTICIPATING",
131
  "211": "NON_PARTICIPATING",
132
  "212": "NON_PARTICIPATING",
133
+ "213": "NON_PARTICIPATING",
134
  "214": "NON_PARTICIPATING",
135
  "215": "NON_PARTICIPATING",
136
+ "216": "NON_PARTICIPATING",
137
+ "217": "NON_PARTICIPATING",
138
  "218": "NON_PARTICIPATING",
139
  "219": "NON_PARTICIPATING",
140
  "22": "NON_PARTICIPATING",
 
142
  "221": "NON_PARTICIPATING",
143
  "222": "NON_PARTICIPATING",
144
  "223": "NON_PARTICIPATING",
145
+ "224": "NON_PARTICIPATING",
146
+ "225": "NON_PARTICIPATING",
147
+ "226": "SUCCESS",
148
  "227": "SUCCESS",
149
  "228": "NON_PARTICIPATING",
150
+ "229": "NON_PARTICIPATING",
151
  "23": "NON_PARTICIPATING",
152
+ "230": "NON_PARTICIPATING",
153
  "231": "NON_PARTICIPATING",
154
  "232": "NON_PARTICIPATING",
155
  "233": "NON_PARTICIPATING",
156
  "234": "NON_PARTICIPATING",
157
  "235": "NON_PARTICIPATING",
158
+ "236": "NON_PARTICIPATING",
159
  "237": "NON_PARTICIPATING",
160
  "238": "NON_PARTICIPATING",
161
+ "239": "NON_PARTICIPATING",
162
  "24": "NON_PARTICIPATING",
163
+ "240": "NON_PARTICIPATING",
164
  "241": "NON_PARTICIPATING",
165
+ "242": "SUCCESS",
166
+ "243": "SUCCESS",
167
+ "244": "NON_PARTICIPATING",
168
  "245": "NON_PARTICIPATING",
169
  "246": "NON_PARTICIPATING",
170
  "247": "NON_PARTICIPATING",
171
  "248": "NON_PARTICIPATING",
172
  "249": "NON_PARTICIPATING",
173
+ "25": "SUCCESS",
174
  "250": "NON_PARTICIPATING",
175
  "251": "NON_PARTICIPATING",
176
+ "252": "NON_PARTICIPATING",
177
  "253": "NON_PARTICIPATING",
178
  "254": "NON_PARTICIPATING",
179
+ "255": "SUCCESS",
180
  "26": "NON_PARTICIPATING",
181
  "27": "NON_PARTICIPATING",
182
  "28": "NON_PARTICIPATING",
183
  "29": "NON_PARTICIPATING",
184
  "3": "NON_PARTICIPATING",
185
+ "30": "SUCCESS",
186
+ "31": "SUCCESS",
187
+ "32": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
+ "34": "SUCCESS",
190
  "35": "NON_PARTICIPATING",
191
+ "36": "SUCCESS",
192
+ "37": "NON_PARTICIPATING",
193
  "38": "NON_PARTICIPATING",
194
+ "39": "SUCCESS",
195
+ "4": "NON_PARTICIPATING",
196
  "40": "NON_PARTICIPATING",
197
  "41": "NON_PARTICIPATING",
198
  "42": "NON_PARTICIPATING",
 
201
  "45": "NON_PARTICIPATING",
202
  "46": "NON_PARTICIPATING",
203
  "47": "NON_PARTICIPATING",
204
+ "48": "NON_PARTICIPATING",
205
  "49": "NON_PARTICIPATING",
206
+ "5": "SUCCESS",
207
  "50": "NON_PARTICIPATING",
208
  "51": "NON_PARTICIPATING",
209
  "52": "NON_PARTICIPATING",
210
  "53": "NON_PARTICIPATING",
211
+ "54": "SUCCESS",
212
+ "55": "NON_PARTICIPATING",
213
  "56": "NON_PARTICIPATING",
214
  "57": "NON_PARTICIPATING",
215
+ "58": "NON_PARTICIPATING",
216
  "59": "NON_PARTICIPATING",
217
+ "6": "SUCCESS",
218
  "60": "NON_PARTICIPATING",
219
  "61": "NON_PARTICIPATING",
220
+ "62": "NON_PARTICIPATING",
221
+ "63": "NON_PARTICIPATING",
222
  "64": "NON_PARTICIPATING",
223
+ "65": "NON_PARTICIPATING",
224
+ "66": "SUCCESS",
225
  "67": "NON_PARTICIPATING",
226
  "68": "NON_PARTICIPATING",
227
  "69": "NON_PARTICIPATING",
228
+ "7": "NON_PARTICIPATING",
229
  "70": "NON_PARTICIPATING",
230
+ "71": "SUCCESS",
231
+ "72": "SUCCESS",
232
  "73": "SUCCESS",
233
+ "74": "NON_PARTICIPATING",
234
  "75": "NON_PARTICIPATING",
235
+ "76": "SUCCESS",
236
  "77": "NON_PARTICIPATING",
237
+ "78": "SUCCESS",
238
  "79": "NON_PARTICIPATING",
239
  "8": "NON_PARTICIPATING",
240
+ "80": "NON_PARTICIPATING",
241
  "81": "NON_PARTICIPATING",
242
  "82": "NON_PARTICIPATING",
243
+ "83": "SUCCESS",
244
  "84": "NON_PARTICIPATING",
245
  "85": "SUCCESS",
246
+ "86": "SUCCESS",
247
  "87": "NON_PARTICIPATING",
248
+ "88": "NON_PARTICIPATING",
249
+ "89": "SUCCESS",
250
+ "9": "SUCCESS",
251
  "90": "NON_PARTICIPATING",
252
  "91": "NON_PARTICIPATING",
253
  "92": "NON_PARTICIPATING",
254
  "93": "SUCCESS",
255
  "94": "SUCCESS",
256
  "95": "NON_PARTICIPATING",
257
+ "96": "NON_PARTICIPATING",
258
  "97": "NON_PARTICIPATING",
259
+ "98": "NON_PARTICIPATING",
260
  "99": "NON_PARTICIPATING"
261
  },
262
  "architectures": [
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c9f1922ff37ee182f0c23d7d8c61ef53866fafb3b1c462bd9167b0407c3c0d7b
3
  size 4040701744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73923f959754007779184a505e1197f53640e1c9f385c86d2f7f25387fd32867
3
  size 4040701744