kmfoda commited on
Commit
c9156f4
·
verified ·
1 Parent(s): 1977e21

Run 4. Outer Step 22. Inner Step 0. Peers 96.

Browse files
Files changed (4) hide show
  1. config.json +36 -36
  2. inner_optimizer.pt +1 -1
  3. model.safetensors +1 -1
  4. outer_optimizer.pt +1 -1
config.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "_name_or_path": "wandb01/c66-h16",
3
  "activation_function": "gelu_new",
4
  "all_reduce_scores": {
5
  "0": "NON_PARTICIPATING",
@@ -11,12 +11,12 @@
11
  "103": "NON_PARTICIPATING",
12
  "104": "NON_PARTICIPATING",
13
  "105": "NON_PARTICIPATING",
14
- "106": "SUCCESS",
15
  "107": "SUCCESS",
16
  "108": "NON_PARTICIPATING",
17
- "109": "SUCCESS",
18
  "11": "NON_PARTICIPATING",
19
- "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "NON_PARTICIPATING",
22
  "113": "NON_PARTICIPATING",
@@ -24,13 +24,13 @@
24
  "115": "NON_PARTICIPATING",
25
  "116": "NON_PARTICIPATING",
26
  "117": "NON_PARTICIPATING",
27
- "118": "NON_PARTICIPATING",
28
  "119": "NON_PARTICIPATING",
29
  "12": "SUCCESS",
30
- "120": "SUCCESS",
31
- "121": "NON_PARTICIPATING",
32
  "122": "NON_PARTICIPATING",
33
- "123": "NON_PARTICIPATING",
34
  "124": "NON_PARTICIPATING",
35
  "125": "NON_PARTICIPATING",
36
  "126": "NON_PARTICIPATING",
@@ -43,7 +43,7 @@
43
  "132": "SUCCESS",
44
  "133": "NON_PARTICIPATING",
45
  "134": "NON_PARTICIPATING",
46
- "135": "NON_PARTICIPATING",
47
  "136": "NON_PARTICIPATING",
48
  "137": "NON_PARTICIPATING",
49
  "138": "NON_PARTICIPATING",
@@ -57,10 +57,10 @@
57
  "145": "NON_PARTICIPATING",
58
  "146": "NON_PARTICIPATING",
59
  "147": "SUCCESS",
60
- "148": "SUCCESS",
61
  "149": "SUCCESS",
62
  "15": "SUCCESS",
63
- "150": "SUCCESS",
64
  "151": "NON_PARTICIPATING",
65
  "152": "NON_PARTICIPATING",
66
  "153": "NON_PARTICIPATING",
@@ -69,25 +69,25 @@
69
  "156": "NON_PARTICIPATING",
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": "NON_PARTICIPATING",
79
  "165": "NON_PARTICIPATING",
80
  "166": "NON_PARTICIPATING",
81
- "167": "NON_PARTICIPATING",
82
- "168": "FAIL",
83
  "169": "NON_PARTICIPATING",
84
  "17": "SUCCESS",
85
- "170": "NON_PARTICIPATING",
86
  "171": "NON_PARTICIPATING",
87
  "172": "SUCCESS",
88
  "173": "NON_PARTICIPATING",
89
  "174": "NON_PARTICIPATING",
90
- "175": "NON_PARTICIPATING",
91
  "176": "NON_PARTICIPATING",
92
  "177": "NON_PARTICIPATING",
93
  "178": "NON_PARTICIPATING",
@@ -107,9 +107,9 @@
107
  "190": "NON_PARTICIPATING",
108
  "191": "SUCCESS",
109
  "192": "SUCCESS",
110
- "193": "SUCCESS",
111
  "194": "NON_PARTICIPATING",
112
- "195": "NON_PARTICIPATING",
113
  "196": "NON_PARTICIPATING",
114
  "197": "NON_PARTICIPATING",
115
  "198": "NON_PARTICIPATING",
@@ -117,9 +117,9 @@
117
  "2": "SUCCESS",
118
  "20": "SUCCESS",
119
  "200": "SUCCESS",
120
- "201": "SUCCESS",
121
  "202": "NON_PARTICIPATING",
122
- "203": "NON_PARTICIPATING",
123
  "204": "NON_PARTICIPATING",
124
  "205": "NON_PARTICIPATING",
125
  "206": "NON_PARTICIPATING",
@@ -127,12 +127,12 @@
127
  "208": "NON_PARTICIPATING",
128
  "209": "NON_PARTICIPATING",
129
  "21": "SUCCESS",
130
- "210": "SUCCESS",
131
- "211": "SUCCESS",
132
  "212": "NON_PARTICIPATING",
133
- "213": "SUCCESS",
134
  "214": "SUCCESS",
135
- "215": "NON_PARTICIPATING",
136
  "216": "NON_PARTICIPATING",
137
  "217": "NON_PARTICIPATING",
138
  "218": "NON_PARTICIPATING",
@@ -143,14 +143,14 @@
143
  "222": "NON_PARTICIPATING",
144
  "223": "NON_PARTICIPATING",
145
  "224": "NON_PARTICIPATING",
146
- "225": "NON_PARTICIPATING",
147
  "226": "SUCCESS",
148
  "227": "NON_PARTICIPATING",
149
  "228": "NON_PARTICIPATING",
150
  "229": "NON_PARTICIPATING",
151
  "23": "SUCCESS",
152
- "230": "SUCCESS",
153
- "231": "SUCCESS",
154
  "232": "NON_PARTICIPATING",
155
  "233": "NON_PARTICIPATING",
156
  "234": "NON_PARTICIPATING",
@@ -161,13 +161,13 @@
161
  "239": "NON_PARTICIPATING",
162
  "24": "SUCCESS",
163
  "240": "NON_PARTICIPATING",
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": "SUCCESS",
171
  "248": "NON_PARTICIPATING",
172
  "249": "SUCCESS",
173
  "25": "SUCCESS",
@@ -180,7 +180,7 @@
180
  "26": "SUCCESS",
181
  "27": "SUCCESS",
182
  "28": "SUCCESS",
183
- "29": "NON_PARTICIPATING",
184
  "3": "SUCCESS",
185
  "30": "NON_PARTICIPATING",
186
  "31": "NON_PARTICIPATING",
@@ -218,7 +218,7 @@
218
  "60": "SUCCESS",
219
  "61": "SUCCESS",
220
  "62": "SUCCESS",
221
- "63": "SUCCESS",
222
  "64": "NON_PARTICIPATING",
223
  "65": "SUCCESS",
224
  "66": "NON_PARTICIPATING",
@@ -236,19 +236,19 @@
236
  "77": "NON_PARTICIPATING",
237
  "78": "NON_PARTICIPATING",
238
  "79": "NON_PARTICIPATING",
239
- "8": "FAIL",
240
  "80": "NON_PARTICIPATING",
241
  "81": "NON_PARTICIPATING",
242
  "82": "SUCCESS",
243
  "83": "NON_PARTICIPATING",
244
  "84": "SUCCESS",
245
- "85": "NON_PARTICIPATING",
246
  "86": "NON_PARTICIPATING",
247
  "87": "SUCCESS",
248
  "88": "SUCCESS",
249
  "89": "SUCCESS",
250
  "9": "NON_PARTICIPATING",
251
- "90": "NON_PARTICIPATING",
252
  "91": "SUCCESS",
253
  "92": "NON_PARTICIPATING",
254
  "93": "NON_PARTICIPATING",
@@ -275,7 +275,7 @@
275
  "initializer_range": 0.02,
276
  "inner_step": 0,
277
  "inner_steps": 0,
278
- "last_allreduce_block": 5525369,
279
  "layer_norm_epsilon": 1e-05,
280
  "model_type": "gpt_optimized",
281
  "n_embd": 1280,
 
1
  {
2
+ "_name_or_path": "penguin102/c66-h20",
3
  "activation_function": "gelu_new",
4
  "all_reduce_scores": {
5
  "0": "NON_PARTICIPATING",
 
11
  "103": "NON_PARTICIPATING",
12
  "104": "NON_PARTICIPATING",
13
  "105": "NON_PARTICIPATING",
14
+ "106": "NON_PARTICIPATING",
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",
 
24
  "115": "NON_PARTICIPATING",
25
  "116": "NON_PARTICIPATING",
26
  "117": "NON_PARTICIPATING",
27
+ "118": "FAIL",
28
  "119": "NON_PARTICIPATING",
29
  "12": "SUCCESS",
30
+ "120": "NON_PARTICIPATING",
31
+ "121": "SUCCESS",
32
  "122": "NON_PARTICIPATING",
33
+ "123": "SUCCESS",
34
  "124": "NON_PARTICIPATING",
35
  "125": "NON_PARTICIPATING",
36
  "126": "NON_PARTICIPATING",
 
43
  "132": "SUCCESS",
44
  "133": "NON_PARTICIPATING",
45
  "134": "NON_PARTICIPATING",
46
+ "135": "SUCCESS",
47
  "136": "NON_PARTICIPATING",
48
  "137": "NON_PARTICIPATING",
49
  "138": "NON_PARTICIPATING",
 
57
  "145": "NON_PARTICIPATING",
58
  "146": "NON_PARTICIPATING",
59
  "147": "SUCCESS",
60
+ "148": "NON_PARTICIPATING",
61
  "149": "SUCCESS",
62
  "15": "SUCCESS",
63
+ "150": "NON_PARTICIPATING",
64
  "151": "NON_PARTICIPATING",
65
  "152": "NON_PARTICIPATING",
66
  "153": "NON_PARTICIPATING",
 
69
  "156": "NON_PARTICIPATING",
70
  "157": "NON_PARTICIPATING",
71
  "158": "NON_PARTICIPATING",
72
+ "159": "NON_PARTICIPATING",
73
  "16": "SUCCESS",
74
+ "160": "SUCCESS",
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": "FAIL",
82
+ "168": "NON_PARTICIPATING",
83
  "169": "NON_PARTICIPATING",
84
  "17": "SUCCESS",
85
+ "170": "SUCCESS",
86
  "171": "NON_PARTICIPATING",
87
  "172": "SUCCESS",
88
  "173": "NON_PARTICIPATING",
89
  "174": "NON_PARTICIPATING",
90
+ "175": "SUCCESS",
91
  "176": "NON_PARTICIPATING",
92
  "177": "NON_PARTICIPATING",
93
  "178": "NON_PARTICIPATING",
 
107
  "190": "NON_PARTICIPATING",
108
  "191": "SUCCESS",
109
  "192": "SUCCESS",
110
+ "193": "NON_PARTICIPATING",
111
  "194": "NON_PARTICIPATING",
112
+ "195": "SUCCESS",
113
  "196": "NON_PARTICIPATING",
114
  "197": "NON_PARTICIPATING",
115
  "198": "NON_PARTICIPATING",
 
117
  "2": "SUCCESS",
118
  "20": "SUCCESS",
119
  "200": "SUCCESS",
120
+ "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
+ "203": "SUCCESS",
123
  "204": "NON_PARTICIPATING",
124
  "205": "NON_PARTICIPATING",
125
  "206": "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": "NON_PARTICIPATING",
 
143
  "222": "NON_PARTICIPATING",
144
  "223": "NON_PARTICIPATING",
145
  "224": "NON_PARTICIPATING",
146
+ "225": "SUCCESS",
147
  "226": "SUCCESS",
148
  "227": "NON_PARTICIPATING",
149
  "228": "NON_PARTICIPATING",
150
  "229": "NON_PARTICIPATING",
151
  "23": "SUCCESS",
152
+ "230": "NON_PARTICIPATING",
153
+ "231": "NON_PARTICIPATING",
154
  "232": "NON_PARTICIPATING",
155
  "233": "NON_PARTICIPATING",
156
  "234": "NON_PARTICIPATING",
 
161
  "239": "NON_PARTICIPATING",
162
  "24": "SUCCESS",
163
  "240": "NON_PARTICIPATING",
164
+ "241": "SUCCESS",
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": "SUCCESS",
173
  "25": "SUCCESS",
 
180
  "26": "SUCCESS",
181
  "27": "SUCCESS",
182
  "28": "SUCCESS",
183
+ "29": "SUCCESS",
184
  "3": "SUCCESS",
185
  "30": "NON_PARTICIPATING",
186
  "31": "NON_PARTICIPATING",
 
218
  "60": "SUCCESS",
219
  "61": "SUCCESS",
220
  "62": "SUCCESS",
221
+ "63": "FAIL",
222
  "64": "NON_PARTICIPATING",
223
  "65": "SUCCESS",
224
  "66": "NON_PARTICIPATING",
 
236
  "77": "NON_PARTICIPATING",
237
  "78": "NON_PARTICIPATING",
238
  "79": "NON_PARTICIPATING",
239
+ "8": "SUCCESS",
240
  "80": "NON_PARTICIPATING",
241
  "81": "NON_PARTICIPATING",
242
  "82": "SUCCESS",
243
  "83": "NON_PARTICIPATING",
244
  "84": "SUCCESS",
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",
 
275
  "initializer_range": 0.02,
276
  "inner_step": 0,
277
  "inner_steps": 0,
278
+ "last_allreduce_block": 5526620,
279
  "layer_norm_epsilon": 1e-05,
280
  "model_type": "gpt_optimized",
281
  "n_embd": 1280,
inner_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:67bd6e0db6c2a34afd9b8e88815d323e13dfc2e83d5e0a37bc6ec62cb348935d
3
  size 8081782026
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71c62fa06874e2a87be2b983b7330d68ba9daa7336fc8fa550d59d235ea91b65
3
  size 8081782026
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:18c125bbbaa5b77926501f473aefab3d7712828ca96f0a164f72c97ca09e8f6d
3
  size 4040701744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:356442642c28c614d259e56af5f0d183c5f0520a47a69ea7940da76e4eae812f
3
  size 4040701744
outer_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:13bc49470bcd86d617886d2162ad6ad6a96e60b5f78c16850bf5525e1a88f3be
3
  size 4040805354
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1a8c7aa4b7c426b9d2ffd39da113e252bce583712817ed98fcde22b0c337562
3
  size 4040805354