philip324mark commited on
Commit
38119c3
·
verified ·
1 Parent(s): d10b186

Run 4. Outer Step 58. Inner Step 1.

Browse files
Files changed (3) hide show
  1. config.json +52 -52
  2. inner_optimizer.pt +2 -2
  3. model.safetensors +1 -1
config.json CHANGED
@@ -6,76 +6,76 @@
6
  "1": "SUCCESS",
7
  "10": "NON_PARTICIPATING",
8
  "100": "NON_PARTICIPATING",
9
- "101": "SUCCESS",
10
  "102": "NON_PARTICIPATING",
11
  "103": "NON_PARTICIPATING",
12
- "104": "SUCCESS",
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": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "NON_PARTICIPATING",
22
- "113": "NON_PARTICIPATING",
23
  "114": "NON_PARTICIPATING",
24
  "115": "NON_PARTICIPATING",
25
  "116": "NON_PARTICIPATING",
26
  "117": "NON_PARTICIPATING",
27
- "118": "SUCCESS",
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",
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": "SUCCESS",
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": "NON_PARTICIPATING",
54
- "142": "SUCCESS",
55
  "143": "NON_PARTICIPATING",
56
  "144": "NON_PARTICIPATING",
57
  "145": "NON_PARTICIPATING",
58
  "146": "NON_PARTICIPATING",
59
  "147": "NON_PARTICIPATING",
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": "SUCCESS",
68
  "155": "NON_PARTICIPATING",
69
  "156": "NON_PARTICIPATING",
70
  "157": "NON_PARTICIPATING",
71
  "158": "NON_PARTICIPATING",
72
- "159": "NON_PARTICIPATING",
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": "SUCCESS",
80
  "166": "NON_PARTICIPATING",
81
  "167": "NON_PARTICIPATING",
@@ -94,18 +94,18 @@
94
  "179": "NON_PARTICIPATING",
95
  "18": "SUCCESS",
96
  "180": "NON_PARTICIPATING",
97
- "181": "NON_PARTICIPATING",
98
  "182": "NON_PARTICIPATING",
99
  "183": "NON_PARTICIPATING",
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",
@@ -116,7 +116,7 @@
116
  "199": "NON_PARTICIPATING",
117
  "2": "SUCCESS",
118
  "20": "SUCCESS",
119
- "200": "SUCCESS",
120
  "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
  "203": "NON_PARTICIPATING",
@@ -138,40 +138,40 @@
138
  "218": "NON_PARTICIPATING",
139
  "219": "NON_PARTICIPATING",
140
  "22": "SUCCESS",
141
- "220": "SUCCESS",
142
  "221": "NON_PARTICIPATING",
143
  "222": "NON_PARTICIPATING",
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": "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": "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": "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",
@@ -182,15 +182,15 @@
182
  "28": "SUCCESS",
183
  "29": "SUCCESS",
184
  "3": "SUCCESS",
185
- "30": "SUCCESS",
186
  "31": "NON_PARTICIPATING",
187
  "32": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
  "34": "SUCCESS",
190
- "35": "SUCCESS",
191
  "36": "SUCCESS",
192
  "37": "SUCCESS",
193
- "38": "NON_PARTICIPATING",
194
  "39": "SUCCESS",
195
  "4": "SUCCESS",
196
  "40": "SUCCESS",
@@ -200,7 +200,7 @@
200
  "44": "SUCCESS",
201
  "45": "SUCCESS",
202
  "46": "NON_PARTICIPATING",
203
- "47": "NON_PARTICIPATING",
204
  "48": "NON_PARTICIPATING",
205
  "49": "SUCCESS",
206
  "5": "SUCCESS",
@@ -209,8 +209,8 @@
209
  "52": "SUCCESS",
210
  "53": "SUCCESS",
211
  "54": "SUCCESS",
212
- "55": "NON_PARTICIPATING",
213
- "56": "NON_PARTICIPATING",
214
  "57": "SUCCESS",
215
  "58": "NON_PARTICIPATING",
216
  "59": "SUCCESS",
@@ -225,7 +225,7 @@
225
  "67": "SUCCESS",
226
  "68": "SUCCESS",
227
  "69": "SUCCESS",
228
- "7": "SUCCESS",
229
  "70": "NON_PARTICIPATING",
230
  "71": "NON_PARTICIPATING",
231
  "72": "SUCCESS",
@@ -247,9 +247,9 @@
247
  "87": "SUCCESS",
248
  "88": "SUCCESS",
249
  "89": "SUCCESS",
250
- "9": "SUCCESS",
251
- "90": "SUCCESS",
252
- "91": "NON_PARTICIPATING",
253
  "92": "SUCCESS",
254
  "93": "NON_PARTICIPATING",
255
  "94": "SUCCESS",
@@ -257,7 +257,7 @@
257
  "96": "SUCCESS",
258
  "97": "SUCCESS",
259
  "98": "SUCCESS",
260
- "99": "NON_PARTICIPATING"
261
  },
262
  "architectures": [
263
  "GPTOptim"
@@ -268,20 +268,20 @@
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
- 5587006,
272
- 5587007,
273
- 5587011,
274
- 5587014,
275
- 5587016
276
  ],
277
  "block_size": 1024,
278
  "bos_token_id": 50256,
279
  "embd_pdrop": 0.1,
280
  "eos_token_id": 50256,
281
  "initializer_range": 0.02,
282
- "inner_step": 51,
283
  "inner_steps": 0,
284
- "last_allreduce_block": 5583387,
285
  "layer_norm_epsilon": 1e-05,
286
  "model_type": "gpt_optimized",
287
  "n_embd": 1280,
 
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": "NON_PARTICIPATING",
13
  "105": "NON_PARTICIPATING",
14
  "106": "NON_PARTICIPATING",
15
+ "107": "SUCCESS",
16
  "108": "NON_PARTICIPATING",
17
+ "109": "SUCCESS",
18
  "11": "SUCCESS",
19
  "110": "NON_PARTICIPATING",
20
  "111": "NON_PARTICIPATING",
21
  "112": "NON_PARTICIPATING",
22
+ "113": "SUCCESS",
23
  "114": "NON_PARTICIPATING",
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": "NON_PARTICIPATING",
31
+ "121": "NON_PARTICIPATING",
32
  "122": "NON_PARTICIPATING",
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": "SUCCESS",
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": "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": "NON_PARTICIPATING",
59
  "147": "NON_PARTICIPATING",
60
+ "148": "SUCCESS",
61
  "149": "NON_PARTICIPATING",
62
+ "15": "NON_PARTICIPATING",
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",
70
  "157": "NON_PARTICIPATING",
71
  "158": "NON_PARTICIPATING",
72
+ "159": "SUCCESS",
73
  "16": "SUCCESS",
74
  "160": "NON_PARTICIPATING",
75
  "161": "NON_PARTICIPATING",
76
+ "162": "SUCCESS",
77
  "163": "NON_PARTICIPATING",
78
+ "164": "SUCCESS",
79
  "165": "SUCCESS",
80
  "166": "NON_PARTICIPATING",
81
  "167": "NON_PARTICIPATING",
 
94
  "179": "NON_PARTICIPATING",
95
  "18": "SUCCESS",
96
  "180": "NON_PARTICIPATING",
97
+ "181": "SUCCESS",
98
  "182": "NON_PARTICIPATING",
99
  "183": "NON_PARTICIPATING",
100
  "184": "NON_PARTICIPATING",
101
+ "185": "SUCCESS",
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": "SUCCESS",
109
  "192": "NON_PARTICIPATING",
110
  "193": "NON_PARTICIPATING",
111
  "194": "NON_PARTICIPATING",
 
116
  "199": "NON_PARTICIPATING",
117
  "2": "SUCCESS",
118
  "20": "SUCCESS",
119
+ "200": "NON_PARTICIPATING",
120
  "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
  "203": "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": "SUCCESS",
151
  "23": "SUCCESS",
152
+ "230": "SUCCESS",
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": "SUCCESS",
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": "SUCCESS",
169
  "246": "NON_PARTICIPATING",
170
  "247": "NON_PARTICIPATING",
171
  "248": "NON_PARTICIPATING",
172
  "249": "NON_PARTICIPATING",
173
  "25": "SUCCESS",
174
+ "250": "SUCCESS",
175
  "251": "NON_PARTICIPATING",
176
  "252": "NON_PARTICIPATING",
177
  "253": "NON_PARTICIPATING",
 
182
  "28": "SUCCESS",
183
  "29": "SUCCESS",
184
  "3": "SUCCESS",
185
+ "30": "NON_PARTICIPATING",
186
  "31": "NON_PARTICIPATING",
187
  "32": "SUCCESS",
188
  "33": "NON_PARTICIPATING",
189
  "34": "SUCCESS",
190
+ "35": "NON_PARTICIPATING",
191
  "36": "SUCCESS",
192
  "37": "SUCCESS",
193
+ "38": "SUCCESS",
194
  "39": "SUCCESS",
195
  "4": "SUCCESS",
196
  "40": "SUCCESS",
 
200
  "44": "SUCCESS",
201
  "45": "SUCCESS",
202
  "46": "NON_PARTICIPATING",
203
+ "47": "SUCCESS",
204
  "48": "NON_PARTICIPATING",
205
  "49": "SUCCESS",
206
  "5": "SUCCESS",
 
209
  "52": "SUCCESS",
210
  "53": "SUCCESS",
211
  "54": "SUCCESS",
212
+ "55": "SUCCESS",
213
+ "56": "SUCCESS",
214
  "57": "SUCCESS",
215
  "58": "NON_PARTICIPATING",
216
  "59": "SUCCESS",
 
225
  "67": "SUCCESS",
226
  "68": "SUCCESS",
227
  "69": "SUCCESS",
228
+ "7": "NON_PARTICIPATING",
229
  "70": "NON_PARTICIPATING",
230
  "71": "NON_PARTICIPATING",
231
  "72": "SUCCESS",
 
247
  "87": "SUCCESS",
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": "SUCCESS",
 
257
  "96": "SUCCESS",
258
  "97": "SUCCESS",
259
  "98": "SUCCESS",
260
+ "99": "SUCCESS"
261
  },
262
  "architectures": [
263
  "GPTOptim"
 
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
+ 5587040,
272
+ 5587099,
273
+ 5587102,
274
+ 5587104,
275
+ 5587108
276
  ],
277
  "block_size": 1024,
278
  "bos_token_id": 50256,
279
  "embd_pdrop": 0.1,
280
  "eos_token_id": 50256,
281
  "initializer_range": 0.02,
282
+ "inner_step": 1,
283
  "inner_steps": 0,
284
+ "last_allreduce_block": 5576204,
285
  "layer_norm_epsilon": 1e-05,
286
  "model_type": "gpt_optimized",
287
  "n_embd": 1280,
inner_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6da8bcca46d64181ccdff72524bcd68c438161d238efc2aa47ef886f7a152d08
3
- size 8081782503
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c67b9ea2aa6db578bf4dfca2e689e7ab1e90f1a25fa4e372af7142aaf7cc208
3
+ size 8081782439
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bab1eeaf5240a1175a434d91486fc98406f0781382b20884f697892fded77c79
3
  size 4040701744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ba35fe72ed12b54c4e91ff5e916450689614454f76f3fc50436b55c099335a8
3
  size 4040701744