kmfoda commited on
Commit
e25094c
·
verified ·
1 Parent(s): 6342ede

Epoch 3. Batch Size 32516. Peers 80.

Browse files
config.json CHANGED
@@ -3,38 +3,38 @@
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": "SUCCESS",
11
  "103": "SUCCESS",
12
- "104": "NON_PARTICIPATING",
13
  "105": "SUCCESS",
14
  "106": "SUCCESS",
15
  "107": "SUCCESS",
16
  "108": "NON_PARTICIPATING",
17
  "109": "NON_PARTICIPATING",
18
- "11": "SUCCESS",
19
- "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
- "112": "SUCCESS",
22
- "113": "SUCCESS",
23
  "114": "NON_PARTICIPATING",
24
  "115": "SUCCESS",
25
- "116": "SUCCESS",
26
  "117": "NON_PARTICIPATING",
27
- "118": "SUCCESS",
28
  "119": "NON_PARTICIPATING",
29
  "12": "NON_PARTICIPATING",
30
- "120": "SUCCESS",
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": "NON_PARTICIPATING",
40
  "13": "NON_PARTICIPATING",
@@ -42,45 +42,45 @@
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": "NON_PARTICIPATING",
49
  "138": "SUCCESS",
50
  "139": "NON_PARTICIPATING",
51
- "14": "NON_PARTICIPATING",
52
  "140": "NON_PARTICIPATING",
53
  "141": "NON_PARTICIPATING",
54
  "142": "NON_PARTICIPATING",
55
  "143": "NON_PARTICIPATING",
56
- "144": "SUCCESS",
57
  "145": "NON_PARTICIPATING",
58
  "146": "SUCCESS",
59
- "147": "SUCCESS",
60
  "148": "NON_PARTICIPATING",
61
- "149": "NON_PARTICIPATING",
62
- "15": "NON_PARTICIPATING",
63
  "150": "NON_PARTICIPATING",
64
  "151": "NON_PARTICIPATING",
65
  "152": "SUCCESS",
66
- "153": "NON_PARTICIPATING",
67
  "154": "NON_PARTICIPATING",
68
- "155": "NON_PARTICIPATING",
69
  "156": "NON_PARTICIPATING",
70
- "157": "SUCCESS",
71
  "158": "SUCCESS",
72
  "159": "NON_PARTICIPATING",
73
- "16": "SUCCESS",
74
- "160": "SUCCESS",
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": "SUCCESS",
83
- "169": "NON_PARTICIPATING",
84
  "17": "NON_PARTICIPATING",
85
  "170": "SUCCESS",
86
  "171": "NON_PARTICIPATING",
@@ -89,26 +89,26 @@
89
  "174": "NON_PARTICIPATING",
90
  "175": "NON_PARTICIPATING",
91
  "176": "SUCCESS",
92
- "177": "NON_PARTICIPATING",
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": "SUCCESS",
100
- "184": "SUCCESS",
101
  "185": "NON_PARTICIPATING",
102
  "186": "NON_PARTICIPATING",
103
  "187": "SUCCESS",
104
  "188": "NON_PARTICIPATING",
105
- "189": "NON_PARTICIPATING",
106
- "19": "NON_PARTICIPATING",
107
  "190": "NON_PARTICIPATING",
108
- "191": "SUCCESS",
109
  "192": "NON_PARTICIPATING",
110
- "193": "SUCCESS",
111
- "194": "NON_PARTICIPATING",
112
  "195": "NON_PARTICIPATING",
113
  "196": "SUCCESS",
114
  "197": "NON_PARTICIPATING",
@@ -116,7 +116,7 @@
116
  "199": "NON_PARTICIPATING",
117
  "2": "NON_PARTICIPATING",
118
  "20": "NON_PARTICIPATING",
119
- "200": "SUCCESS",
120
  "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
  "203": "NON_PARTICIPATING",
@@ -126,16 +126,16 @@
126
  "207": "NON_PARTICIPATING",
127
  "208": "NON_PARTICIPATING",
128
  "209": "NON_PARTICIPATING",
129
- "21": "SUCCESS",
130
  "210": "NON_PARTICIPATING",
131
- "211": "NON_PARTICIPATING",
132
- "212": "NON_PARTICIPATING",
133
  "213": "NON_PARTICIPATING",
134
  "214": "SUCCESS",
135
- "215": "SUCCESS",
136
  "216": "NON_PARTICIPATING",
137
  "217": "NON_PARTICIPATING",
138
- "218": "SUCCESS",
139
  "219": "NON_PARTICIPATING",
140
  "22": "NON_PARTICIPATING",
141
  "220": "SUCCESS",
@@ -143,33 +143,33 @@
143
  "222": "SUCCESS",
144
  "223": "NON_PARTICIPATING",
145
  "224": "NON_PARTICIPATING",
146
- "225": "SUCCESS",
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": "NON_PARTICIPATING",
154
  "232": "NON_PARTICIPATING",
155
- "233": "NON_PARTICIPATING",
156
  "234": "SUCCESS",
157
  "235": "NON_PARTICIPATING",
158
  "236": "NON_PARTICIPATING",
159
- "237": "NON_PARTICIPATING",
160
- "238": "NON_PARTICIPATING",
161
- "239": "NON_PARTICIPATING",
162
- "24": "SUCCESS",
163
  "240": "NON_PARTICIPATING",
164
- "241": "SUCCESS",
165
- "242": "NON_PARTICIPATING",
166
- "243": "SUCCESS",
167
  "244": "NON_PARTICIPATING",
168
- "245": "SUCCESS",
169
  "246": "SUCCESS",
170
  "247": "NON_PARTICIPATING",
171
  "248": "NON_PARTICIPATING",
172
- "249": "SUCCESS",
173
  "25": "SUCCESS",
174
  "250": "NON_PARTICIPATING",
175
  "251": "NON_PARTICIPATING",
@@ -177,13 +177,13 @@
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": "NON_PARTICIPATING",
186
- "31": "NON_PARTICIPATING",
187
  "32": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
  "34": "NON_PARTICIPATING",
@@ -198,24 +198,24 @@
198
  "42": "SUCCESS",
199
  "43": "NON_PARTICIPATING",
200
  "44": "NON_PARTICIPATING",
201
- "45": "NON_PARTICIPATING",
202
  "46": "NON_PARTICIPATING",
203
  "47": "NON_PARTICIPATING",
204
- "48": "SUCCESS",
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": "NON_PARTICIPATING",
212
  "55": "NON_PARTICIPATING",
213
- "56": "NON_PARTICIPATING",
214
  "57": "NON_PARTICIPATING",
215
  "58": "NON_PARTICIPATING",
216
- "59": "NON_PARTICIPATING",
217
  "6": "NON_PARTICIPATING",
218
- "60": "NON_PARTICIPATING",
219
  "61": "SUCCESS",
220
  "62": "NON_PARTICIPATING",
221
  "63": "NON_PARTICIPATING",
@@ -223,39 +223,39 @@
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": "NON_PARTICIPATING",
231
- "72": "NON_PARTICIPATING",
232
  "73": "NON_PARTICIPATING",
233
  "74": "NON_PARTICIPATING",
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": "SUCCESS",
241
- "81": "NON_PARTICIPATING",
242
- "82": "SUCCESS",
243
  "83": "SUCCESS",
244
  "84": "NON_PARTICIPATING",
245
- "85": "SUCCESS",
246
  "86": "NON_PARTICIPATING",
247
- "87": "SUCCESS",
248
- "88": "SUCCESS",
249
- "89": "SUCCESS",
250
  "9": "NON_PARTICIPATING",
251
- "90": "SUCCESS",
252
  "91": "SUCCESS",
253
- "92": "NON_PARTICIPATING",
254
  "93": "NON_PARTICIPATING",
255
- "94": "SUCCESS",
256
- "95": "NON_PARTICIPATING",
257
- "96": "NON_PARTICIPATING",
258
- "97": "NON_PARTICIPATING",
259
  "98": "NON_PARTICIPATING",
260
  "99": "NON_PARTICIPATING"
261
  },
 
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": "SUCCESS",
12
+ "104": "SUCCESS",
13
  "105": "SUCCESS",
14
  "106": "SUCCESS",
15
  "107": "SUCCESS",
16
  "108": "NON_PARTICIPATING",
17
  "109": "NON_PARTICIPATING",
18
+ "11": "NON_PARTICIPATING",
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": "NON_PARTICIPATING",
27
+ "118": "NON_PARTICIPATING",
28
  "119": "NON_PARTICIPATING",
29
  "12": "NON_PARTICIPATING",
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": "SUCCESS",
38
  "128": "NON_PARTICIPATING",
39
  "129": "NON_PARTICIPATING",
40
  "13": "NON_PARTICIPATING",
 
42
  "131": "NON_PARTICIPATING",
43
  "132": "NON_PARTICIPATING",
44
  "133": "NON_PARTICIPATING",
45
+ "134": "SUCCESS",
46
  "135": "NON_PARTICIPATING",
47
+ "136": "NON_PARTICIPATING",
48
  "137": "NON_PARTICIPATING",
49
  "138": "SUCCESS",
50
  "139": "NON_PARTICIPATING",
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": "SUCCESS",
59
+ "147": "NON_PARTICIPATING",
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": "SUCCESS",
69
  "156": "NON_PARTICIPATING",
70
+ "157": "NON_PARTICIPATING",
71
  "158": "SUCCESS",
72
  "159": "NON_PARTICIPATING",
73
+ "16": "NON_PARTICIPATING",
74
+ "160": "NON_PARTICIPATING",
75
  "161": "NON_PARTICIPATING",
76
+ "162": "SUCCESS",
77
  "163": "SUCCESS",
78
+ "164": "NON_PARTICIPATING",
79
+ "165": "SUCCESS",
80
  "166": "NON_PARTICIPATING",
81
+ "167": "SUCCESS",
82
  "168": "SUCCESS",
83
+ "169": "SUCCESS",
84
  "17": "NON_PARTICIPATING",
85
  "170": "SUCCESS",
86
  "171": "NON_PARTICIPATING",
 
89
  "174": "NON_PARTICIPATING",
90
  "175": "NON_PARTICIPATING",
91
  "176": "SUCCESS",
92
+ "177": "SUCCESS",
93
  "178": "NON_PARTICIPATING",
94
  "179": "NON_PARTICIPATING",
95
  "18": "NON_PARTICIPATING",
96
+ "180": "SUCCESS",
97
  "181": "NON_PARTICIPATING",
98
+ "182": "SUCCESS",
99
  "183": "SUCCESS",
100
+ "184": "NON_PARTICIPATING",
101
  "185": "NON_PARTICIPATING",
102
  "186": "NON_PARTICIPATING",
103
  "187": "SUCCESS",
104
  "188": "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": "NON_PARTICIPATING",
113
  "196": "SUCCESS",
114
  "197": "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",
 
126
  "207": "NON_PARTICIPATING",
127
  "208": "NON_PARTICIPATING",
128
  "209": "NON_PARTICIPATING",
129
+ "21": "NON_PARTICIPATING",
130
  "210": "NON_PARTICIPATING",
131
+ "211": "SUCCESS",
132
+ "212": "SUCCESS",
133
  "213": "NON_PARTICIPATING",
134
  "214": "SUCCESS",
135
+ "215": "NON_PARTICIPATING",
136
  "216": "NON_PARTICIPATING",
137
  "217": "NON_PARTICIPATING",
138
+ "218": "NON_PARTICIPATING",
139
  "219": "NON_PARTICIPATING",
140
  "22": "NON_PARTICIPATING",
141
  "220": "SUCCESS",
 
143
  "222": "SUCCESS",
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": "SUCCESS",
151
  "23": "NON_PARTICIPATING",
152
  "230": "NON_PARTICIPATING",
153
+ "231": "SUCCESS",
154
  "232": "NON_PARTICIPATING",
155
+ "233": "SUCCESS",
156
  "234": "SUCCESS",
157
  "235": "NON_PARTICIPATING",
158
  "236": "NON_PARTICIPATING",
159
+ "237": "SUCCESS",
160
+ "238": "SUCCESS",
161
+ "239": "SUCCESS",
162
+ "24": "NON_PARTICIPATING",
163
  "240": "NON_PARTICIPATING",
164
+ "241": "NON_PARTICIPATING",
165
+ "242": "SUCCESS",
166
+ "243": "NON_PARTICIPATING",
167
  "244": "NON_PARTICIPATING",
168
+ "245": "NON_PARTICIPATING",
169
  "246": "SUCCESS",
170
  "247": "NON_PARTICIPATING",
171
  "248": "NON_PARTICIPATING",
172
+ "249": "NON_PARTICIPATING",
173
  "25": "SUCCESS",
174
  "250": "NON_PARTICIPATING",
175
  "251": "NON_PARTICIPATING",
 
177
  "253": "NON_PARTICIPATING",
178
  "254": "NON_PARTICIPATING",
179
  "255": "SUCCESS",
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": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
  "34": "NON_PARTICIPATING",
 
198
  "42": "SUCCESS",
199
  "43": "NON_PARTICIPATING",
200
  "44": "NON_PARTICIPATING",
201
+ "45": "SUCCESS",
202
  "46": "NON_PARTICIPATING",
203
  "47": "NON_PARTICIPATING",
204
+ "48": "NON_PARTICIPATING",
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": "NON_PARTICIPATING",
213
+ "56": "SUCCESS",
214
  "57": "NON_PARTICIPATING",
215
  "58": "NON_PARTICIPATING",
216
+ "59": "SUCCESS",
217
  "6": "NON_PARTICIPATING",
218
+ "60": "SUCCESS",
219
  "61": "SUCCESS",
220
  "62": "NON_PARTICIPATING",
221
  "63": "NON_PARTICIPATING",
 
223
  "65": "NON_PARTICIPATING",
224
  "66": "SUCCESS",
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": "NON_PARTICIPATING",
233
  "74": "NON_PARTICIPATING",
234
  "75": "NON_PARTICIPATING",
235
+ "76": "NON_PARTICIPATING",
236
+ "77": "SUCCESS",
237
  "78": "NON_PARTICIPATING",
238
+ "79": "SUCCESS",
239
  "8": "NON_PARTICIPATING",
240
  "80": "SUCCESS",
241
+ "81": "SUCCESS",
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": "NON_PARTICIPATING",
251
+ "90": "NON_PARTICIPATING",
252
  "91": "SUCCESS",
253
+ "92": "SUCCESS",
254
  "93": "NON_PARTICIPATING",
255
+ "94": "NON_PARTICIPATING",
256
+ "95": "SUCCESS",
257
+ "96": "SUCCESS",
258
+ "97": "SUCCESS",
259
  "98": "NON_PARTICIPATING",
260
  "99": "NON_PARTICIPATING"
261
  },
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6fa184d16f413c5ce059e341847589531dfa085d9d2f0700b5ed6bfab19c2bfa
3
  size 4975530824
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dd6136915d69211ea4b5d18ed0e2858ecea8ef1838ab59140b8f8ddeb340684
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:ad2d72e6fd7c9a51d7ff9730ed6faaf7a245c0a512d527d19b07b2fa48d39c78
3
  size 3075942944
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be27f118ca0901d17c747aa24014726f233d8fecefcfdaa6bfae2f838f82ee64
3
  size 3075942944
optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0903dd2f218dfa40c9a659be7d0d63d35972856e10bb0da5ec4676abcca13ab7
3
- size 4675070692
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b2b0dff1980a1158c53deaab24e1a493b326b861d0dcbbb059c367c33a9237f8
3
+ size 4675755358