kmfoda commited on
Commit
771d741
·
verified ·
1 Parent(s): 2a438b2

Epoch 1603. Batch Size 53562. Peers 94.

Browse files
config.json CHANGED
@@ -3,27 +3,27 @@
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": "NON_PARTICIPATING",
10
  "102": "NON_PARTICIPATING",
11
- "103": "SUCCESS",
12
  "104": "SUCCESS",
13
  "105": "NON_PARTICIPATING",
14
  "106": "NON_PARTICIPATING",
15
- "107": "NON_PARTICIPATING",
16
  "108": "SUCCESS",
17
  "109": "SUCCESS",
18
- "11": "SUCCESS",
19
  "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "NON_PARTICIPATING",
22
  "113": "NON_PARTICIPATING",
23
  "114": "SUCCESS",
24
  "115": "NON_PARTICIPATING",
25
- "116": "SUCCESS",
26
- "117": "NON_PARTICIPATING",
27
  "118": "NON_PARTICIPATING",
28
  "119": "SUCCESS",
29
  "12": "NON_PARTICIPATING",
@@ -31,113 +31,113 @@
31
  "121": "NON_PARTICIPATING",
32
  "122": "NON_PARTICIPATING",
33
  "123": "NON_PARTICIPATING",
34
- "124": "SUCCESS",
35
- "125": "SUCCESS",
36
- "126": "SUCCESS",
37
- "127": "SUCCESS",
38
- "128": "NON_PARTICIPATING",
39
  "129": "NON_PARTICIPATING",
40
  "13": "NON_PARTICIPATING",
41
- "130": "SUCCESS",
42
  "131": "NON_PARTICIPATING",
43
  "132": "NON_PARTICIPATING",
44
- "133": "NON_PARTICIPATING",
45
- "134": "NON_PARTICIPATING",
46
  "135": "NON_PARTICIPATING",
47
- "136": "NON_PARTICIPATING",
48
- "137": "SUCCESS",
49
  "138": "NON_PARTICIPATING",
50
  "139": "NON_PARTICIPATING",
51
  "14": "NON_PARTICIPATING",
52
- "140": "NON_PARTICIPATING",
53
- "141": "SUCCESS",
54
- "142": "NON_PARTICIPATING",
55
  "143": "NON_PARTICIPATING",
56
  "144": "NON_PARTICIPATING",
57
  "145": "SUCCESS",
58
  "146": "SUCCESS",
59
  "147": "SUCCESS",
60
  "148": "NON_PARTICIPATING",
61
- "149": "NON_PARTICIPATING",
62
  "15": "SUCCESS",
63
  "150": "NON_PARTICIPATING",
64
  "151": "NON_PARTICIPATING",
65
- "152": "NON_PARTICIPATING",
66
  "153": "SUCCESS",
67
  "154": "NON_PARTICIPATING",
68
- "155": "SUCCESS",
69
  "156": "NON_PARTICIPATING",
70
- "157": "SUCCESS",
71
  "158": "NON_PARTICIPATING",
72
- "159": "SUCCESS",
73
  "16": "NON_PARTICIPATING",
74
  "160": "NON_PARTICIPATING",
75
- "161": "SUCCESS",
76
- "162": "SUCCESS",
77
- "163": "SUCCESS",
78
  "164": "NON_PARTICIPATING",
79
  "165": "NON_PARTICIPATING",
80
- "166": "SUCCESS",
81
  "167": "NON_PARTICIPATING",
82
- "168": "NON_PARTICIPATING",
83
  "169": "NON_PARTICIPATING",
84
  "17": "NON_PARTICIPATING",
85
  "170": "SUCCESS",
86
- "171": "SUCCESS",
87
  "172": "NON_PARTICIPATING",
88
  "173": "SUCCESS",
89
  "174": "NON_PARTICIPATING",
90
  "175": "NON_PARTICIPATING",
91
  "176": "NON_PARTICIPATING",
92
- "177": "NON_PARTICIPATING",
93
- "178": "SUCCESS",
94
  "179": "NON_PARTICIPATING",
95
  "18": "NON_PARTICIPATING",
96
- "180": "SUCCESS",
97
- "181": "SUCCESS",
98
  "182": "NON_PARTICIPATING",
99
  "183": "NON_PARTICIPATING",
100
  "184": "SUCCESS",
101
- "185": "SUCCESS",
102
- "186": "NON_PARTICIPATING",
103
  "187": "NON_PARTICIPATING",
104
- "188": "SUCCESS",
105
  "189": "NON_PARTICIPATING",
106
- "19": "SUCCESS",
107
- "190": "SUCCESS",
108
  "191": "NON_PARTICIPATING",
109
- "192": "SUCCESS",
110
- "193": "SUCCESS",
111
- "194": "SUCCESS",
112
- "195": "SUCCESS",
113
- "196": "NON_PARTICIPATING",
114
- "197": "NON_PARTICIPATING",
115
- "198": "NON_PARTICIPATING",
116
- "199": "SUCCESS",
117
- "2": "SUCCESS",
118
- "20": "SUCCESS",
119
  "200": "SUCCESS",
120
  "201": "SUCCESS",
121
- "202": "SUCCESS",
122
- "203": "SUCCESS",
123
  "204": "NON_PARTICIPATING",
124
- "205": "NON_PARTICIPATING",
125
  "206": "SUCCESS",
126
  "207": "NON_PARTICIPATING",
127
  "208": "SUCCESS",
128
  "209": "NON_PARTICIPATING",
129
- "21": "SUCCESS",
130
  "210": "NON_PARTICIPATING",
131
- "211": "SUCCESS",
132
- "212": "SUCCESS",
133
- "213": "NON_PARTICIPATING",
134
  "214": "NON_PARTICIPATING",
135
  "215": "NON_PARTICIPATING",
136
  "216": "SUCCESS",
137
- "217": "NON_PARTICIPATING",
138
  "218": "NON_PARTICIPATING",
139
  "219": "SUCCESS",
140
- "22": "SUCCESS",
141
  "220": "NON_PARTICIPATING",
142
  "221": "NON_PARTICIPATING",
143
  "222": "NON_PARTICIPATING",
@@ -145,61 +145,61 @@
145
  "224": "NON_PARTICIPATING",
146
  "225": "NON_PARTICIPATING",
147
  "226": "NON_PARTICIPATING",
148
- "227": "NON_PARTICIPATING",
149
  "228": "SUCCESS",
150
  "229": "NON_PARTICIPATING",
151
  "23": "SUCCESS",
152
- "230": "NON_PARTICIPATING",
153
- "231": "SUCCESS",
154
  "232": "NON_PARTICIPATING",
155
  "233": "NON_PARTICIPATING",
156
  "234": "NON_PARTICIPATING",
157
- "235": "SUCCESS",
158
  "236": "NON_PARTICIPATING",
159
  "237": "NON_PARTICIPATING",
160
  "238": "SUCCESS",
161
- "239": "NON_PARTICIPATING",
162
  "24": "SUCCESS",
163
  "240": "NON_PARTICIPATING",
164
- "241": "SUCCESS",
165
  "242": "NON_PARTICIPATING",
166
- "243": "NON_PARTICIPATING",
167
  "244": "NON_PARTICIPATING",
168
  "245": "NON_PARTICIPATING",
169
- "246": "SUCCESS",
170
- "247": "NON_PARTICIPATING",
171
- "248": "SUCCESS",
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": "NON_PARTICIPATING",
180
  "26": "SUCCESS",
181
- "27": "NON_PARTICIPATING",
182
  "28": "NON_PARTICIPATING",
183
- "29": "NON_PARTICIPATING",
184
- "3": "NON_PARTICIPATING",
185
  "30": "NON_PARTICIPATING",
186
  "31": "SUCCESS",
187
  "32": "NON_PARTICIPATING",
188
  "33": "NON_PARTICIPATING",
189
  "34": "SUCCESS",
190
  "35": "NON_PARTICIPATING",
191
- "36": "SUCCESS",
192
  "37": "NON_PARTICIPATING",
193
  "38": "NON_PARTICIPATING",
194
  "39": "NON_PARTICIPATING",
195
- "4": "NON_PARTICIPATING",
196
  "40": "SUCCESS",
197
  "41": "NON_PARTICIPATING",
198
- "42": "SUCCESS",
199
  "43": "NON_PARTICIPATING",
200
  "44": "NON_PARTICIPATING",
201
  "45": "SUCCESS",
202
- "46": "NON_PARTICIPATING",
203
  "47": "SUCCESS",
204
  "48": "NON_PARTICIPATING",
205
  "49": "SUCCESS",
@@ -207,7 +207,7 @@
207
  "50": "NON_PARTICIPATING",
208
  "51": "SUCCESS",
209
  "52": "SUCCESS",
210
- "53": "SUCCESS",
211
  "54": "SUCCESS",
212
  "55": "NON_PARTICIPATING",
213
  "56": "SUCCESS",
@@ -217,47 +217,47 @@
217
  "6": "NON_PARTICIPATING",
218
  "60": "NON_PARTICIPATING",
219
  "61": "NON_PARTICIPATING",
220
- "62": "SUCCESS",
221
  "63": "NON_PARTICIPATING",
222
  "64": "NON_PARTICIPATING",
223
  "65": "NON_PARTICIPATING",
224
  "66": "NON_PARTICIPATING",
225
- "67": "NON_PARTICIPATING",
226
  "68": "SUCCESS",
227
  "69": "NON_PARTICIPATING",
228
  "7": "NON_PARTICIPATING",
229
  "70": "NON_PARTICIPATING",
230
  "71": "NON_PARTICIPATING",
231
  "72": "SUCCESS",
232
- "73": "SUCCESS",
233
- "74": "SUCCESS",
234
  "75": "NON_PARTICIPATING",
235
- "76": "SUCCESS",
236
- "77": "NON_PARTICIPATING",
237
- "78": "NON_PARTICIPATING",
238
  "79": "NON_PARTICIPATING",
239
  "8": "NON_PARTICIPATING",
240
  "80": "NON_PARTICIPATING",
241
  "81": "NON_PARTICIPATING",
242
  "82": "NON_PARTICIPATING",
243
- "83": "NON_PARTICIPATING",
244
  "84": "NON_PARTICIPATING",
245
  "85": "NON_PARTICIPATING",
246
  "86": "NON_PARTICIPATING",
247
  "87": "NON_PARTICIPATING",
248
  "88": "NON_PARTICIPATING",
249
- "89": "SUCCESS",
250
- "9": "NON_PARTICIPATING",
251
- "90": "SUCCESS",
252
- "91": "SUCCESS",
253
- "92": "NON_PARTICIPATING",
254
- "93": "SUCCESS",
255
  "94": "NON_PARTICIPATING",
256
  "95": "NON_PARTICIPATING",
257
- "96": "NON_PARTICIPATING",
258
  "97": "NON_PARTICIPATING",
259
  "98": "SUCCESS",
260
- "99": "SUCCESS"
261
  },
262
  "architectures": [
263
  "GPTOptim"
 
3
  "activation_function": "gelu_new",
4
  "all_reduce_scores": {
5
  "0": "NON_PARTICIPATING",
6
+ "1": "SUCCESS",
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": "NON_PARTICIPATING",
14
  "106": "NON_PARTICIPATING",
15
+ "107": "SUCCESS",
16
  "108": "SUCCESS",
17
  "109": "SUCCESS",
18
+ "11": "NON_PARTICIPATING",
19
  "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "NON_PARTICIPATING",
22
  "113": "NON_PARTICIPATING",
23
  "114": "SUCCESS",
24
  "115": "NON_PARTICIPATING",
25
+ "116": "NON_PARTICIPATING",
26
+ "117": "SUCCESS",
27
  "118": "NON_PARTICIPATING",
28
  "119": "SUCCESS",
29
  "12": "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": "SUCCESS",
39
  "129": "NON_PARTICIPATING",
40
  "13": "NON_PARTICIPATING",
41
+ "130": "NON_PARTICIPATING",
42
  "131": "NON_PARTICIPATING",
43
  "132": "NON_PARTICIPATING",
44
+ "133": "SUCCESS",
45
+ "134": "SUCCESS",
46
  "135": "NON_PARTICIPATING",
47
+ "136": "SUCCESS",
48
+ "137": "NON_PARTICIPATING",
49
  "138": "NON_PARTICIPATING",
50
  "139": "NON_PARTICIPATING",
51
  "14": "NON_PARTICIPATING",
52
+ "140": "SUCCESS",
53
+ "141": "NON_PARTICIPATING",
54
+ "142": "SUCCESS",
55
  "143": "NON_PARTICIPATING",
56
  "144": "NON_PARTICIPATING",
57
  "145": "SUCCESS",
58
  "146": "SUCCESS",
59
  "147": "SUCCESS",
60
  "148": "NON_PARTICIPATING",
61
+ "149": "SUCCESS",
62
  "15": "SUCCESS",
63
  "150": "NON_PARTICIPATING",
64
  "151": "NON_PARTICIPATING",
65
+ "152": "SUCCESS",
66
  "153": "SUCCESS",
67
  "154": "NON_PARTICIPATING",
68
+ "155": "NON_PARTICIPATING",
69
  "156": "NON_PARTICIPATING",
70
+ "157": "NON_PARTICIPATING",
71
  "158": "NON_PARTICIPATING",
72
+ "159": "NON_PARTICIPATING",
73
  "16": "NON_PARTICIPATING",
74
  "160": "NON_PARTICIPATING",
75
+ "161": "NON_PARTICIPATING",
76
+ "162": "NON_PARTICIPATING",
77
+ "163": "NON_PARTICIPATING",
78
  "164": "NON_PARTICIPATING",
79
  "165": "NON_PARTICIPATING",
80
+ "166": "NON_PARTICIPATING",
81
  "167": "NON_PARTICIPATING",
82
+ "168": "SUCCESS",
83
  "169": "NON_PARTICIPATING",
84
  "17": "NON_PARTICIPATING",
85
  "170": "SUCCESS",
86
+ "171": "NON_PARTICIPATING",
87
  "172": "NON_PARTICIPATING",
88
  "173": "SUCCESS",
89
  "174": "NON_PARTICIPATING",
90
  "175": "NON_PARTICIPATING",
91
  "176": "NON_PARTICIPATING",
92
+ "177": "SUCCESS",
93
+ "178": "NON_PARTICIPATING",
94
  "179": "NON_PARTICIPATING",
95
  "18": "NON_PARTICIPATING",
96
+ "180": "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": "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": "NON_PARTICIPATING",
113
+ "196": "SUCCESS",
114
+ "197": "SUCCESS",
115
+ "198": "SUCCESS",
116
+ "199": "NON_PARTICIPATING",
117
+ "2": "NON_PARTICIPATING",
118
+ "20": "NON_PARTICIPATING",
119
  "200": "SUCCESS",
120
  "201": "SUCCESS",
121
+ "202": "NON_PARTICIPATING",
122
+ "203": "NON_PARTICIPATING",
123
  "204": "NON_PARTICIPATING",
124
+ "205": "SUCCESS",
125
  "206": "SUCCESS",
126
  "207": "NON_PARTICIPATING",
127
  "208": "SUCCESS",
128
  "209": "NON_PARTICIPATING",
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": "SUCCESS",
140
+ "22": "NON_PARTICIPATING",
141
  "220": "NON_PARTICIPATING",
142
  "221": "NON_PARTICIPATING",
143
  "222": "NON_PARTICIPATING",
 
145
  "224": "NON_PARTICIPATING",
146
  "225": "NON_PARTICIPATING",
147
  "226": "NON_PARTICIPATING",
148
+ "227": "SUCCESS",
149
  "228": "SUCCESS",
150
  "229": "NON_PARTICIPATING",
151
  "23": "SUCCESS",
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": "NON_PARTICIPATING",
159
  "237": "NON_PARTICIPATING",
160
  "238": "SUCCESS",
161
+ "239": "SUCCESS",
162
  "24": "SUCCESS",
163
  "240": "NON_PARTICIPATING",
164
+ "241": "NON_PARTICIPATING",
165
  "242": "NON_PARTICIPATING",
166
+ "243": "SUCCESS",
167
  "244": "NON_PARTICIPATING",
168
  "245": "NON_PARTICIPATING",
169
+ "246": "NON_PARTICIPATING",
170
+ "247": "SUCCESS",
171
+ "248": "NON_PARTICIPATING",
172
  "249": "NON_PARTICIPATING",
173
  "25": "SUCCESS",
174
+ "250": "SUCCESS",
175
+ "251": "SUCCESS",
176
  "252": "NON_PARTICIPATING",
177
+ "253": "SUCCESS",
178
  "254": "NON_PARTICIPATING",
179
+ "255": "SUCCESS",
180
  "26": "SUCCESS",
181
+ "27": "SUCCESS",
182
  "28": "NON_PARTICIPATING",
183
+ "29": "SUCCESS",
184
+ "3": "SUCCESS",
185
  "30": "NON_PARTICIPATING",
186
  "31": "SUCCESS",
187
  "32": "NON_PARTICIPATING",
188
  "33": "NON_PARTICIPATING",
189
  "34": "SUCCESS",
190
  "35": "NON_PARTICIPATING",
191
+ "36": "NON_PARTICIPATING",
192
  "37": "NON_PARTICIPATING",
193
  "38": "NON_PARTICIPATING",
194
  "39": "NON_PARTICIPATING",
195
+ "4": "SUCCESS",
196
  "40": "SUCCESS",
197
  "41": "NON_PARTICIPATING",
198
+ "42": "NON_PARTICIPATING",
199
  "43": "NON_PARTICIPATING",
200
  "44": "NON_PARTICIPATING",
201
  "45": "SUCCESS",
202
+ "46": "SUCCESS",
203
  "47": "SUCCESS",
204
  "48": "NON_PARTICIPATING",
205
  "49": "SUCCESS",
 
207
  "50": "NON_PARTICIPATING",
208
  "51": "SUCCESS",
209
  "52": "SUCCESS",
210
+ "53": "NON_PARTICIPATING",
211
  "54": "SUCCESS",
212
  "55": "NON_PARTICIPATING",
213
  "56": "SUCCESS",
 
217
  "6": "NON_PARTICIPATING",
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": "NON_PARTICIPATING",
225
+ "67": "SUCCESS",
226
  "68": "SUCCESS",
227
  "69": "NON_PARTICIPATING",
228
  "7": "NON_PARTICIPATING",
229
  "70": "NON_PARTICIPATING",
230
  "71": "NON_PARTICIPATING",
231
  "72": "SUCCESS",
232
+ "73": "NON_PARTICIPATING",
233
+ "74": "NON_PARTICIPATING",
234
  "75": "NON_PARTICIPATING",
235
+ "76": "NON_PARTICIPATING",
236
+ "77": "SUCCESS",
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": "NON_PARTICIPATING",
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": "NON_PARTICIPATING",
253
+ "92": "SUCCESS",
254
+ "93": "NON_PARTICIPATING",
255
  "94": "NON_PARTICIPATING",
256
  "95": "NON_PARTICIPATING",
257
+ "96": "SUCCESS",
258
  "97": "NON_PARTICIPATING",
259
  "98": "SUCCESS",
260
+ "99": "NON_PARTICIPATING"
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 sha256:91f54a6472f939dc5cfd060957d1cf3bbebbb24332cfa7b5e3a8df06bdf1b73b
3
  size 4975530824
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4419878668f8f98c3c69ae3a5cfb731a23f66ec9e461a1194c0b2486e2f64fac
3
  size 4975530824
model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7a80458b5a56a104334b745907dcd1820579de2748cbec398f5b8ebc259f9012
3
  size 3075942944
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54929883fe193de868ef384a1030f81aceeef6d8a9914bef8dddb30d4ee0d700
3
  size 3075942944
optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f21e8b49f6bab2b97d7e50daabde37b5707f586ccdd7d95ac5d3fb7ac6f9fb41
3
  size 4675755870
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92ecc791a0d0c08718a662819e439898f3a1387feb52be394c63a3dde73bf4d3
3
  size 4675755870