kmfoda commited on
Commit
b76dbfc
·
verified ·
1 Parent(s): ca3c2e3

Epoch 1673. Batch Size 54262. Peers 101.

Browse files
config.json CHANGED
@@ -3,25 +3,25 @@
3
  "activation_function": "gelu_new",
4
  "all_reduce_scores": {
5
  "0": "NON_PARTICIPATING",
6
- "1": "SUCCESS",
7
- "10": "SUCCESS",
8
  "100": "NON_PARTICIPATING",
9
  "101": "SUCCESS",
10
- "102": "SUCCESS",
11
  "103": "NON_PARTICIPATING",
12
  "104": "NON_PARTICIPATING",
13
- "105": "NON_PARTICIPATING",
14
  "106": "SUCCESS",
15
- "107": "NON_PARTICIPATING",
16
- "108": "SUCCESS",
17
  "109": "SUCCESS",
18
- "11": "NON_PARTICIPATING",
19
  "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "SUCCESS",
22
- "113": "SUCCESS",
23
  "114": "NON_PARTICIPATING",
24
- "115": "NON_PARTICIPATING",
25
  "116": "SUCCESS",
26
  "117": "NON_PARTICIPATING",
27
  "118": "NON_PARTICIPATING",
@@ -29,34 +29,34 @@
29
  "12": "SUCCESS",
30
  "120": "NON_PARTICIPATING",
31
  "121": "NON_PARTICIPATING",
32
- "122": "NON_PARTICIPATING",
33
- "123": "SUCCESS",
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": "NON_PARTICIPATING",
41
  "130": "NON_PARTICIPATING",
42
  "131": "NON_PARTICIPATING",
43
- "132": "NON_PARTICIPATING",
44
  "133": "NON_PARTICIPATING",
45
- "134": "NON_PARTICIPATING",
46
  "135": "NON_PARTICIPATING",
47
  "136": "SUCCESS",
48
  "137": "SUCCESS",
49
- "138": "NON_PARTICIPATING",
50
  "139": "SUCCESS",
51
- "14": "NON_PARTICIPATING",
52
- "140": "SUCCESS",
53
  "141": "NON_PARTICIPATING",
54
  "142": "NON_PARTICIPATING",
55
  "143": "NON_PARTICIPATING",
56
  "144": "NON_PARTICIPATING",
57
- "145": "SUCCESS",
58
- "146": "SUCCESS",
59
- "147": "NON_PARTICIPATING",
60
  "148": "NON_PARTICIPATING",
61
  "149": "NON_PARTICIPATING",
62
  "15": "SUCCESS",
@@ -68,35 +68,35 @@
68
  "155": "NON_PARTICIPATING",
69
  "156": "NON_PARTICIPATING",
70
  "157": "SUCCESS",
71
- "158": "NON_PARTICIPATING",
72
  "159": "NON_PARTICIPATING",
73
  "16": "SUCCESS",
74
  "160": "NON_PARTICIPATING",
75
- "161": "SUCCESS",
76
  "162": "SUCCESS",
77
  "163": "NON_PARTICIPATING",
78
- "164": "NON_PARTICIPATING",
79
  "165": "NON_PARTICIPATING",
80
  "166": "NON_PARTICIPATING",
81
- "167": "SUCCESS",
82
  "168": "NON_PARTICIPATING",
83
  "169": "NON_PARTICIPATING",
84
  "17": "SUCCESS",
85
- "170": "SUCCESS",
86
- "171": "SUCCESS",
87
  "172": "NON_PARTICIPATING",
88
  "173": "SUCCESS",
89
- "174": "SUCCESS",
90
  "175": "SUCCESS",
91
- "176": "NON_PARTICIPATING",
92
- "177": "NON_PARTICIPATING",
93
  "178": "NON_PARTICIPATING",
94
  "179": "NON_PARTICIPATING",
95
  "18": "SUCCESS",
96
- "180": "SUCCESS",
97
  "181": "NON_PARTICIPATING",
98
- "182": "NON_PARTICIPATING",
99
- "183": "SUCCESS",
100
  "184": "SUCCESS",
101
  "185": "SUCCESS",
102
  "186": "NON_PARTICIPATING",
@@ -105,97 +105,97 @@
105
  "189": "SUCCESS",
106
  "19": "SUCCESS",
107
  "190": "NON_PARTICIPATING",
108
- "191": "SUCCESS",
109
  "192": "NON_PARTICIPATING",
110
  "193": "NON_PARTICIPATING",
111
- "194": "NON_PARTICIPATING",
112
- "195": "NON_PARTICIPATING",
113
- "196": "SUCCESS",
114
  "197": "SUCCESS",
115
- "198": "NON_PARTICIPATING",
116
  "199": "SUCCESS",
117
  "2": "NON_PARTICIPATING",
118
- "20": "SUCCESS",
119
  "200": "NON_PARTICIPATING",
120
  "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
- "203": "SUCCESS",
123
  "204": "NON_PARTICIPATING",
124
  "205": "SUCCESS",
125
- "206": "SUCCESS",
126
  "207": "NON_PARTICIPATING",
127
- "208": "NON_PARTICIPATING",
128
- "209": "NON_PARTICIPATING",
129
  "21": "NON_PARTICIPATING",
130
- "210": "SUCCESS",
131
- "211": "SUCCESS",
132
  "212": "SUCCESS",
133
  "213": "NON_PARTICIPATING",
134
- "214": "NON_PARTICIPATING",
135
  "215": "SUCCESS",
136
  "216": "SUCCESS",
137
- "217": "NON_PARTICIPATING",
138
  "218": "NON_PARTICIPATING",
139
  "219": "NON_PARTICIPATING",
140
- "22": "SUCCESS",
141
- "220": "NON_PARTICIPATING",
142
- "221": "NON_PARTICIPATING",
143
  "222": "NON_PARTICIPATING",
144
  "223": "NON_PARTICIPATING",
145
  "224": "NON_PARTICIPATING",
146
  "225": "NON_PARTICIPATING",
147
  "226": "NON_PARTICIPATING",
148
- "227": "NON_PARTICIPATING",
149
  "228": "NON_PARTICIPATING",
150
  "229": "NON_PARTICIPATING",
151
  "23": "NON_PARTICIPATING",
152
  "230": "NON_PARTICIPATING",
153
- "231": "SUCCESS",
154
- "232": "NON_PARTICIPATING",
155
  "233": "SUCCESS",
156
  "234": "NON_PARTICIPATING",
157
- "235": "NON_PARTICIPATING",
158
  "236": "SUCCESS",
159
  "237": "NON_PARTICIPATING",
160
- "238": "SUCCESS",
161
  "239": "NON_PARTICIPATING",
162
- "24": "NON_PARTICIPATING",
163
  "240": "SUCCESS",
164
- "241": "SUCCESS",
165
  "242": "NON_PARTICIPATING",
166
- "243": "NON_PARTICIPATING",
167
  "244": "NON_PARTICIPATING",
168
- "245": "SUCCESS",
169
- "246": "NON_PARTICIPATING",
170
- "247": "NON_PARTICIPATING",
171
- "248": "SUCCESS",
172
- "249": "SUCCESS",
173
  "25": "SUCCESS",
174
  "250": "SUCCESS",
175
- "251": "SUCCESS",
176
  "252": "SUCCESS",
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": "SUCCESS",
185
  "30": "NON_PARTICIPATING",
186
- "31": "NON_PARTICIPATING",
187
- "32": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
  "34": "NON_PARTICIPATING",
190
  "35": "NON_PARTICIPATING",
191
  "36": "SUCCESS",
192
- "37": "NON_PARTICIPATING",
193
  "38": "NON_PARTICIPATING",
194
  "39": "NON_PARTICIPATING",
195
  "4": "SUCCESS",
196
- "40": "NON_PARTICIPATING",
197
  "41": "NON_PARTICIPATING",
198
- "42": "NON_PARTICIPATING",
199
  "43": "NON_PARTICIPATING",
200
  "44": "NON_PARTICIPATING",
201
  "45": "NON_PARTICIPATING",
@@ -205,59 +205,59 @@
205
  "49": "NON_PARTICIPATING",
206
  "5": "SUCCESS",
207
  "50": "NON_PARTICIPATING",
208
- "51": "SUCCESS",
209
  "52": "SUCCESS",
210
  "53": "SUCCESS",
211
  "54": "NON_PARTICIPATING",
212
  "55": "NON_PARTICIPATING",
213
- "56": "SUCCESS",
214
  "57": "NON_PARTICIPATING",
215
  "58": "NON_PARTICIPATING",
216
- "59": "NON_PARTICIPATING",
217
  "6": "SUCCESS",
218
  "60": "NON_PARTICIPATING",
219
  "61": "SUCCESS",
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": "SUCCESS",
228
- "7": "SUCCESS",
229
- "70": "NON_PARTICIPATING",
230
  "71": "NON_PARTICIPATING",
231
- "72": "SUCCESS",
232
- "73": "SUCCESS",
233
- "74": "SUCCESS",
234
- "75": "SUCCESS",
235
  "76": "SUCCESS",
236
  "77": "SUCCESS",
237
  "78": "NON_PARTICIPATING",
238
- "79": "SUCCESS",
239
- "8": "SUCCESS",
240
  "80": "NON_PARTICIPATING",
241
  "81": "NON_PARTICIPATING",
242
  "82": "SUCCESS",
243
  "83": "SUCCESS",
244
  "84": "NON_PARTICIPATING",
245
- "85": "SUCCESS",
246
  "86": "NON_PARTICIPATING",
247
  "87": "NON_PARTICIPATING",
248
- "88": "NON_PARTICIPATING",
249
- "89": "NON_PARTICIPATING",
250
- "9": "SUCCESS",
251
  "90": "NON_PARTICIPATING",
252
  "91": "SUCCESS",
253
  "92": "SUCCESS",
254
  "93": "NON_PARTICIPATING",
255
- "94": "SUCCESS",
256
- "95": "SUCCESS",
257
  "96": "NON_PARTICIPATING",
258
- "97": "SUCCESS",
259
  "98": "NON_PARTICIPATING",
260
- "99": "NON_PARTICIPATING"
261
  },
262
  "architectures": [
263
  "GPTOptim"
 
3
  "activation_function": "gelu_new",
4
  "all_reduce_scores": {
5
  "0": "NON_PARTICIPATING",
6
+ "1": "NON_PARTICIPATING",
7
+ "10": "NON_PARTICIPATING",
8
  "100": "NON_PARTICIPATING",
9
  "101": "SUCCESS",
10
+ "102": "NON_PARTICIPATING",
11
  "103": "NON_PARTICIPATING",
12
  "104": "NON_PARTICIPATING",
13
+ "105": "SUCCESS",
14
  "106": "SUCCESS",
15
+ "107": "SUCCESS",
16
+ "108": "NON_PARTICIPATING",
17
  "109": "SUCCESS",
18
+ "11": "SUCCESS",
19
  "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "SUCCESS",
22
+ "113": "NON_PARTICIPATING",
23
  "114": "NON_PARTICIPATING",
24
+ "115": "SUCCESS",
25
  "116": "SUCCESS",
26
  "117": "NON_PARTICIPATING",
27
  "118": "NON_PARTICIPATING",
 
29
  "12": "SUCCESS",
30
  "120": "NON_PARTICIPATING",
31
  "121": "NON_PARTICIPATING",
32
+ "122": "SUCCESS",
33
+ "123": "NON_PARTICIPATING",
34
  "124": "NON_PARTICIPATING",
35
+ "125": "SUCCESS",
36
+ "126": "SUCCESS",
37
+ "127": "SUCCESS",
38
+ "128": "SUCCESS",
39
  "129": "SUCCESS",
40
  "13": "NON_PARTICIPATING",
41
  "130": "NON_PARTICIPATING",
42
  "131": "NON_PARTICIPATING",
43
+ "132": "SUCCESS",
44
  "133": "NON_PARTICIPATING",
45
+ "134": "SUCCESS",
46
  "135": "NON_PARTICIPATING",
47
  "136": "SUCCESS",
48
  "137": "SUCCESS",
49
+ "138": "SUCCESS",
50
  "139": "SUCCESS",
51
+ "14": "SUCCESS",
52
+ "140": "NON_PARTICIPATING",
53
  "141": "NON_PARTICIPATING",
54
  "142": "NON_PARTICIPATING",
55
  "143": "NON_PARTICIPATING",
56
  "144": "NON_PARTICIPATING",
57
+ "145": "NON_PARTICIPATING",
58
+ "146": "NON_PARTICIPATING",
59
+ "147": "SUCCESS",
60
  "148": "NON_PARTICIPATING",
61
  "149": "NON_PARTICIPATING",
62
  "15": "SUCCESS",
 
68
  "155": "NON_PARTICIPATING",
69
  "156": "NON_PARTICIPATING",
70
  "157": "SUCCESS",
71
+ "158": "SUCCESS",
72
  "159": "NON_PARTICIPATING",
73
  "16": "SUCCESS",
74
  "160": "NON_PARTICIPATING",
75
+ "161": "NON_PARTICIPATING",
76
  "162": "SUCCESS",
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": "SUCCESS",
85
+ "170": "NON_PARTICIPATING",
86
+ "171": "NON_PARTICIPATING",
87
  "172": "NON_PARTICIPATING",
88
  "173": "SUCCESS",
89
+ "174": "NON_PARTICIPATING",
90
  "175": "SUCCESS",
91
+ "176": "SUCCESS",
92
+ "177": "SUCCESS",
93
  "178": "NON_PARTICIPATING",
94
  "179": "NON_PARTICIPATING",
95
  "18": "SUCCESS",
96
+ "180": "NON_PARTICIPATING",
97
  "181": "NON_PARTICIPATING",
98
+ "182": "SUCCESS",
99
+ "183": "NON_PARTICIPATING",
100
  "184": "SUCCESS",
101
  "185": "SUCCESS",
102
  "186": "NON_PARTICIPATING",
 
105
  "189": "SUCCESS",
106
  "19": "SUCCESS",
107
  "190": "NON_PARTICIPATING",
108
+ "191": "NON_PARTICIPATING",
109
  "192": "NON_PARTICIPATING",
110
  "193": "NON_PARTICIPATING",
111
+ "194": "SUCCESS",
112
+ "195": "SUCCESS",
113
+ "196": "NON_PARTICIPATING",
114
  "197": "SUCCESS",
115
+ "198": "SUCCESS",
116
  "199": "SUCCESS",
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": "SUCCESS",
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": "SUCCESS",
133
  "213": "NON_PARTICIPATING",
134
+ "214": "SUCCESS",
135
  "215": "SUCCESS",
136
  "216": "SUCCESS",
137
+ "217": "SUCCESS",
138
  "218": "NON_PARTICIPATING",
139
  "219": "NON_PARTICIPATING",
140
+ "22": "NON_PARTICIPATING",
141
+ "220": "SUCCESS",
142
+ "221": "SUCCESS",
143
  "222": "NON_PARTICIPATING",
144
  "223": "NON_PARTICIPATING",
145
  "224": "NON_PARTICIPATING",
146
  "225": "NON_PARTICIPATING",
147
  "226": "NON_PARTICIPATING",
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": "SUCCESS",
155
  "233": "SUCCESS",
156
  "234": "NON_PARTICIPATING",
157
+ "235": "SUCCESS",
158
  "236": "SUCCESS",
159
  "237": "NON_PARTICIPATING",
160
+ "238": "NON_PARTICIPATING",
161
  "239": "NON_PARTICIPATING",
162
+ "24": "SUCCESS",
163
  "240": "SUCCESS",
164
+ "241": "NON_PARTICIPATING",
165
  "242": "NON_PARTICIPATING",
166
+ "243": "SUCCESS",
167
  "244": "NON_PARTICIPATING",
168
+ "245": "NON_PARTICIPATING",
169
+ "246": "SUCCESS",
170
+ "247": "SUCCESS",
171
+ "248": "NON_PARTICIPATING",
172
+ "249": "NON_PARTICIPATING",
173
  "25": "SUCCESS",
174
  "250": "SUCCESS",
175
+ "251": "NON_PARTICIPATING",
176
  "252": "SUCCESS",
177
  "253": "NON_PARTICIPATING",
178
  "254": "NON_PARTICIPATING",
179
  "255": "SUCCESS",
180
  "26": "NON_PARTICIPATING",
181
  "27": "NON_PARTICIPATING",
182
+ "28": "SUCCESS",
183
  "29": "NON_PARTICIPATING",
184
  "3": "SUCCESS",
185
  "30": "NON_PARTICIPATING",
186
+ "31": "SUCCESS",
187
+ "32": "NON_PARTICIPATING",
188
  "33": "NON_PARTICIPATING",
189
  "34": "NON_PARTICIPATING",
190
  "35": "NON_PARTICIPATING",
191
  "36": "SUCCESS",
192
+ "37": "SUCCESS",
193
  "38": "NON_PARTICIPATING",
194
  "39": "NON_PARTICIPATING",
195
  "4": "SUCCESS",
196
+ "40": "SUCCESS",
197
  "41": "NON_PARTICIPATING",
198
+ "42": "SUCCESS",
199
  "43": "NON_PARTICIPATING",
200
  "44": "NON_PARTICIPATING",
201
  "45": "NON_PARTICIPATING",
 
205
  "49": "NON_PARTICIPATING",
206
  "5": "SUCCESS",
207
  "50": "NON_PARTICIPATING",
208
+ "51": "NON_PARTICIPATING",
209
  "52": "SUCCESS",
210
  "53": "SUCCESS",
211
  "54": "NON_PARTICIPATING",
212
  "55": "NON_PARTICIPATING",
213
+ "56": "NON_PARTICIPATING",
214
  "57": "NON_PARTICIPATING",
215
  "58": "NON_PARTICIPATING",
216
+ "59": "SUCCESS",
217
  "6": "SUCCESS",
218
  "60": "NON_PARTICIPATING",
219
  "61": "SUCCESS",
220
  "62": "NON_PARTICIPATING",
221
+ "63": "SUCCESS",
222
  "64": "NON_PARTICIPATING",
223
+ "65": "SUCCESS",
224
  "66": "SUCCESS",
225
  "67": "NON_PARTICIPATING",
226
+ "68": "SUCCESS",
227
+ "69": "NON_PARTICIPATING",
228
+ "7": "NON_PARTICIPATING",
229
+ "70": "SUCCESS",
230
  "71": "NON_PARTICIPATING",
231
+ "72": "NON_PARTICIPATING",
232
+ "73": "NON_PARTICIPATING",
233
+ "74": "NON_PARTICIPATING",
234
+ "75": "NON_PARTICIPATING",
235
  "76": "SUCCESS",
236
  "77": "SUCCESS",
237
  "78": "NON_PARTICIPATING",
238
+ "79": "NON_PARTICIPATING",
239
+ "8": "NON_PARTICIPATING",
240
  "80": "NON_PARTICIPATING",
241
  "81": "NON_PARTICIPATING",
242
  "82": "SUCCESS",
243
  "83": "SUCCESS",
244
  "84": "NON_PARTICIPATING",
245
+ "85": "NON_PARTICIPATING",
246
  "86": "NON_PARTICIPATING",
247
  "87": "NON_PARTICIPATING",
248
+ "88": "SUCCESS",
249
+ "89": "SUCCESS",
250
+ "9": "NON_PARTICIPATING",
251
  "90": "NON_PARTICIPATING",
252
  "91": "SUCCESS",
253
  "92": "SUCCESS",
254
  "93": "NON_PARTICIPATING",
255
+ "94": "NON_PARTICIPATING",
256
+ "95": "NON_PARTICIPATING",
257
  "96": "NON_PARTICIPATING",
258
+ "97": "NON_PARTICIPATING",
259
  "98": "NON_PARTICIPATING",
260
+ "99": "SUCCESS"
261
  },
262
  "architectures": [
263
  "GPTOptim"
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:411b21f2ef5119b961e6d40356a69000e4d59c5aba312d07a054fded9861c301
3
  size 4975530824
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e45ff65d4447452169782dfa202e32033287ddc3fa375ee089f8e7ba07df81f6
3
  size 4975530824
model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f2e8ee273c8dca4a7b8ef4a7f4599cd48e8141059e50faad10816b522c7d755b
3
  size 3075942944
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17d5d9a6992c15819e65e294c9d8dc0fdc46b835eae194fd0f2794b4880fac80
3
  size 3075942944
optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0e03baa86d9b8ceee30059e42988bb9dc7e55543d62cd75f512fb8e5f3631217
3
  size 4675755870
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31c39e07a1b3d97749cc24402be589fe7b0e4c989b2aee4679ddff00d20b982d
3
  size 4675755870