kmfoda commited on
Commit
b7e207c
·
verified ·
1 Parent(s): 6b93da6

Run 4. Outer Step 65. Inner Step 0. Peers 97.

Browse files
Files changed (4) hide show
  1. config.json +37 -37
  2. inner_optimizer.pt +1 -1
  3. model.safetensors +1 -1
  4. outer_optimizer.pt +1 -1
config.json CHANGED
@@ -3,20 +3,20 @@
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": "SUCCESS",
15
  "107": "NON_PARTICIPATING",
16
  "108": "NON_PARTICIPATING",
17
  "109": "NON_PARTICIPATING",
18
  "11": "SUCCESS",
19
- "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "NON_PARTICIPATING",
22
  "113": "NON_PARTICIPATING",
@@ -27,21 +27,21 @@
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": "NON_PARTICIPATING",
40
- "13": "SUCCESS",
41
- "130": "NON_PARTICIPATING",
42
  "131": "NON_PARTICIPATING",
43
  "132": "NON_PARTICIPATING",
44
- "133": "SUCCESS",
45
  "134": "NON_PARTICIPATING",
46
  "135": "NON_PARTICIPATING",
47
  "136": "NON_PARTICIPATING",
@@ -52,7 +52,7 @@
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",
@@ -63,7 +63,7 @@
63
  "150": "NON_PARTICIPATING",
64
  "151": "NON_PARTICIPATING",
65
  "152": "NON_PARTICIPATING",
66
- "153": "NON_PARTICIPATING",
67
  "154": "NON_PARTICIPATING",
68
  "155": "NON_PARTICIPATING",
69
  "156": "NON_PARTICIPATING",
@@ -83,13 +83,13 @@
83
  "169": "NON_PARTICIPATING",
84
  "17": "SUCCESS",
85
  "170": "NON_PARTICIPATING",
86
- "171": "SUCCESS",
87
  "172": "NON_PARTICIPATING",
88
  "173": "NON_PARTICIPATING",
89
  "174": "NON_PARTICIPATING",
90
- "175": "SUCCESS",
91
  "176": "NON_PARTICIPATING",
92
- "177": "SUCCESS",
93
  "178": "NON_PARTICIPATING",
94
  "179": "NON_PARTICIPATING",
95
  "18": "SUCCESS",
@@ -100,20 +100,20 @@
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": "SUCCESS",
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": "NON_PARTICIPATING",
114
  "197": "NON_PARTICIPATING",
115
  "198": "NON_PARTICIPATING",
116
- "199": "SUCCESS",
117
  "2": "SUCCESS",
118
  "20": "SUCCESS",
119
  "200": "NON_PARTICIPATING",
@@ -125,9 +125,9 @@
125
  "206": "NON_PARTICIPATING",
126
  "207": "NON_PARTICIPATING",
127
  "208": "SUCCESS",
128
- "209": "NON_PARTICIPATING",
129
  "21": "SUCCESS",
130
- "210": "NON_PARTICIPATING",
131
  "211": "NON_PARTICIPATING",
132
  "212": "NON_PARTICIPATING",
133
  "213": "NON_PARTICIPATING",
@@ -140,10 +140,10 @@
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": "SUCCESS",
@@ -151,32 +151,32 @@
151
  "23": "SUCCESS",
152
  "230": "NON_PARTICIPATING",
153
  "231": "NON_PARTICIPATING",
154
- "232": "SUCCESS",
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": "SUCCESS",
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": "NON_PARTICIPATING",
172
- "249": "SUCCESS",
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": "SUCCESS",
182
  "28": "SUCCESS",
@@ -186,7 +186,7 @@
186
  "31": "NON_PARTICIPATING",
187
  "32": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
- "34": "NON_PARTICIPATING",
190
  "35": "SUCCESS",
191
  "36": "SUCCESS",
192
  "37": "SUCCESS",
@@ -200,12 +200,12 @@
200
  "44": "SUCCESS",
201
  "45": "SUCCESS",
202
  "46": "NON_PARTICIPATING",
203
- "47": "SUCCESS",
204
  "48": "NON_PARTICIPATING",
205
  "49": "SUCCESS",
206
  "5": "SUCCESS",
207
  "50": "SUCCESS",
208
- "51": "FAIL",
209
  "52": "NON_PARTICIPATING",
210
  "53": "SUCCESS",
211
  "54": "SUCCESS",
@@ -218,7 +218,7 @@
218
  "60": "SUCCESS",
219
  "61": "NON_PARTICIPATING",
220
  "62": "SUCCESS",
221
- "63": "SUCCESS",
222
  "64": "SUCCESS",
223
  "65": "SUCCESS",
224
  "66": "SUCCESS",
@@ -232,7 +232,7 @@
232
  "73": "NON_PARTICIPATING",
233
  "74": "SUCCESS",
234
  "75": "SUCCESS",
235
- "76": "NON_PARTICIPATING",
236
  "77": "SUCCESS",
237
  "78": "SUCCESS",
238
  "79": "NON_PARTICIPATING",
@@ -247,7 +247,7 @@
247
  "87": "SUCCESS",
248
  "88": "SUCCESS",
249
  "89": "SUCCESS",
250
- "9": "SUCCESS",
251
  "90": "SUCCESS",
252
  "91": "SUCCESS",
253
  "92": "SUCCESS",
@@ -255,7 +255,7 @@
255
  "94": "SUCCESS",
256
  "95": "SUCCESS",
257
  "96": "SUCCESS",
258
- "97": "SUCCESS",
259
  "98": "SUCCESS",
260
  "99": "SUCCESS"
261
  },
@@ -275,7 +275,7 @@
275
  "initializer_range": 0.02,
276
  "inner_step": 0,
277
  "inner_steps": 0,
278
- "last_allreduce_block": 5613899,
279
  "layer_norm_epsilon": 1e-05,
280
  "model_type": "gpt_optimized",
281
  "n_embd": 1280,
 
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": "NON_PARTICIPATING",
12
+ "104": "NON_PARTICIPATING",
13
  "105": "NON_PARTICIPATING",
14
+ "106": "NON_PARTICIPATING",
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",
 
27
  "118": "NON_PARTICIPATING",
28
  "119": "NON_PARTICIPATING",
29
  "12": "SUCCESS",
30
+ "120": "SUCCESS",
31
  "121": "NON_PARTICIPATING",
32
+ "122": "SUCCESS",
33
  "123": "NON_PARTICIPATING",
34
  "124": "NON_PARTICIPATING",
35
  "125": "NON_PARTICIPATING",
36
+ "126": "SUCCESS",
37
  "127": "NON_PARTICIPATING",
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",
 
52
  "140": "NON_PARTICIPATING",
53
  "141": "NON_PARTICIPATING",
54
  "142": "NON_PARTICIPATING",
55
+ "143": "SUCCESS",
56
  "144": "NON_PARTICIPATING",
57
  "145": "NON_PARTICIPATING",
58
  "146": "NON_PARTICIPATING",
 
63
  "150": "NON_PARTICIPATING",
64
  "151": "NON_PARTICIPATING",
65
  "152": "NON_PARTICIPATING",
66
+ "153": "SUCCESS",
67
  "154": "NON_PARTICIPATING",
68
  "155": "NON_PARTICIPATING",
69
  "156": "NON_PARTICIPATING",
 
83
  "169": "NON_PARTICIPATING",
84
  "17": "SUCCESS",
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": "NON_PARTICIPATING",
94
  "179": "NON_PARTICIPATING",
95
  "18": "SUCCESS",
 
100
  "184": "NON_PARTICIPATING",
101
  "185": "NON_PARTICIPATING",
102
  "186": "NON_PARTICIPATING",
103
+ "187": "SUCCESS",
104
  "188": "NON_PARTICIPATING",
105
  "189": "NON_PARTICIPATING",
106
  "19": "SUCCESS",
107
  "190": "NON_PARTICIPATING",
108
  "191": "NON_PARTICIPATING",
109
+ "192": "SUCCESS",
110
  "193": "NON_PARTICIPATING",
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": "SUCCESS",
118
  "20": "SUCCESS",
119
  "200": "NON_PARTICIPATING",
 
125
  "206": "NON_PARTICIPATING",
126
  "207": "NON_PARTICIPATING",
127
  "208": "SUCCESS",
128
+ "209": "SUCCESS",
129
  "21": "SUCCESS",
130
+ "210": "SUCCESS",
131
  "211": "NON_PARTICIPATING",
132
  "212": "NON_PARTICIPATING",
133
  "213": "NON_PARTICIPATING",
 
140
  "22": "SUCCESS",
141
  "220": "NON_PARTICIPATING",
142
  "221": "NON_PARTICIPATING",
143
+ "222": "SUCCESS",
144
  "223": "NON_PARTICIPATING",
145
  "224": "NON_PARTICIPATING",
146
+ "225": "SUCCESS",
147
  "226": "NON_PARTICIPATING",
148
  "227": "NON_PARTICIPATING",
149
  "228": "SUCCESS",
 
151
  "23": "SUCCESS",
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": "SUCCESS",
159
  "237": "NON_PARTICIPATING",
160
  "238": "NON_PARTICIPATING",
161
  "239": "SUCCESS",
162
  "24": "SUCCESS",
163
  "240": "NON_PARTICIPATING",
164
+ "241": "NON_PARTICIPATING",
165
  "242": "NON_PARTICIPATING",
166
  "243": "NON_PARTICIPATING",
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": "SUCCESS",
181
  "27": "SUCCESS",
182
  "28": "SUCCESS",
 
186
  "31": "NON_PARTICIPATING",
187
  "32": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
+ "34": "SUCCESS",
190
  "35": "SUCCESS",
191
  "36": "SUCCESS",
192
  "37": "SUCCESS",
 
200
  "44": "SUCCESS",
201
  "45": "SUCCESS",
202
  "46": "NON_PARTICIPATING",
203
+ "47": "NON_PARTICIPATING",
204
  "48": "NON_PARTICIPATING",
205
  "49": "SUCCESS",
206
  "5": "SUCCESS",
207
  "50": "SUCCESS",
208
+ "51": "SUCCESS",
209
  "52": "NON_PARTICIPATING",
210
  "53": "SUCCESS",
211
  "54": "SUCCESS",
 
218
  "60": "SUCCESS",
219
  "61": "NON_PARTICIPATING",
220
  "62": "SUCCESS",
221
+ "63": "NON_PARTICIPATING",
222
  "64": "SUCCESS",
223
  "65": "SUCCESS",
224
  "66": "SUCCESS",
 
232
  "73": "NON_PARTICIPATING",
233
  "74": "SUCCESS",
234
  "75": "SUCCESS",
235
+ "76": "SUCCESS",
236
  "77": "SUCCESS",
237
  "78": "SUCCESS",
238
  "79": "NON_PARTICIPATING",
 
247
  "87": "SUCCESS",
248
  "88": "SUCCESS",
249
  "89": "SUCCESS",
250
+ "9": "NON_PARTICIPATING",
251
  "90": "SUCCESS",
252
  "91": "SUCCESS",
253
  "92": "SUCCESS",
 
255
  "94": "SUCCESS",
256
  "95": "SUCCESS",
257
  "96": "SUCCESS",
258
+ "97": "NON_PARTICIPATING",
259
  "98": "SUCCESS",
260
  "99": "SUCCESS"
261
  },
 
275
  "initializer_range": 0.02,
276
  "inner_step": 0,
277
  "inner_steps": 0,
278
+ "last_allreduce_block": 5617822,
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:ebc41618cd9a1ff2fc7df56831f0863ea0bb88d53a3c679cb1f6cc3f3999d75d
3
  size 8081782026
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ffd7401e8934a55c7d0e0f2b84f19e63d9bc15283999fb2fb655db192e23025
3
  size 8081782026
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:29f25e4e1f5ad183d22736c38a618157c09202e62e790401715d03e9b9392f5f
3
  size 4040701744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef686e036684854e7925c34f48eb4bffa71b26fd892eaa097bc5f58ab68bb121
3
  size 4040701744
outer_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:34864d261e24f9dd234ee990528e4df61b6d763238a7a2b339795d2f8d3b3576
3
  size 4040805354
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6207ec203b9d383945b613e359feb4f42db7fd86de234b8d3f33f84b44468b3c
3
  size 4040805354