jockibaba commited on
Commit
97975a9
·
verified ·
1 Parent(s): a5eac8b

Run 4. Outer Step 70. Inner Step 0.

Browse files
Files changed (3) hide show
  1. config.json +33 -33
  2. inner_optimizer.pt +1 -1
  3. model.safetensors +1 -1
config.json CHANGED
@@ -6,7 +6,7 @@
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",
@@ -27,23 +27,23 @@
27
  "118": "NON_PARTICIPATING",
28
  "119": "NON_PARTICIPATING",
29
  "12": "SUCCESS",
30
- "120": "SUCCESS",
31
- "121": "SUCCESS",
32
  "122": "NON_PARTICIPATING",
33
  "123": "NON_PARTICIPATING",
34
  "124": "NON_PARTICIPATING",
35
- "125": "SUCCESS",
36
  "126": "NON_PARTICIPATING",
37
- "127": "SUCCESS",
38
  "128": "NON_PARTICIPATING",
39
- "129": "SUCCESS",
40
  "13": "SUCCESS",
41
  "130": "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": "NON_PARTICIPATING",
@@ -74,20 +74,20 @@
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": "SUCCESS",
82
  "168": "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": "SUCCESS",
91
  "176": "NON_PARTICIPATING",
92
  "177": "NON_PARTICIPATING",
93
  "178": "NON_PARTICIPATING",
@@ -106,9 +106,9 @@
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": "NON_PARTICIPATING",
114
  "197": "NON_PARTICIPATING",
@@ -119,24 +119,24 @@
119
  "200": "NON_PARTICIPATING",
120
  "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
- "203": "NON_PARTICIPATING",
123
  "204": "NON_PARTICIPATING",
124
- "205": "NON_PARTICIPATING",
125
  "206": "NON_PARTICIPATING",
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": "SUCCESS",
133
  "213": "NON_PARTICIPATING",
134
  "214": "NON_PARTICIPATING",
135
  "215": "NON_PARTICIPATING",
136
  "216": "NON_PARTICIPATING",
137
  "217": "NON_PARTICIPATING",
138
  "218": "NON_PARTICIPATING",
139
- "219": "SUCCESS",
140
  "22": "SUCCESS",
141
  "220": "NON_PARTICIPATING",
142
  "221": "NON_PARTICIPATING",
@@ -145,14 +145,14 @@
145
  "224": "NON_PARTICIPATING",
146
  "225": "NON_PARTICIPATING",
147
  "226": "NON_PARTICIPATING",
148
- "227": "SUCCESS",
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",
157
  "235": "NON_PARTICIPATING",
158
  "236": "NON_PARTICIPATING",
@@ -162,14 +162,14 @@
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": "NON_PARTICIPATING",
171
  "248": "NON_PARTICIPATING",
172
- "249": "NON_PARTICIPATING",
173
  "25": "SUCCESS",
174
  "250": "NON_PARTICIPATING",
175
  "251": "NON_PARTICIPATING",
@@ -219,7 +219,7 @@
219
  "61": "SUCCESS",
220
  "62": "SUCCESS",
221
  "63": "SUCCESS",
222
- "64": "SUCCESS",
223
  "65": "SUCCESS",
224
  "66": "SUCCESS",
225
  "67": "SUCCESS",
@@ -268,20 +268,20 @@
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
- 5656133,
272
- 5656137,
273
- 5656141,
274
- 5656155,
275
- 5656159
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": 36,
283
  "inner_steps": 0,
284
- "last_allreduce_block": 5654739,
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": "SUCCESS",
10
  "102": "NON_PARTICIPATING",
11
  "103": "NON_PARTICIPATING",
12
  "104": "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": "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": "NON_PARTICIPATING",
45
+ "134": "NON_PARTICIPATING",
46
+ "135": "SUCCESS",
47
  "136": "NON_PARTICIPATING",
48
  "137": "NON_PARTICIPATING",
49
  "138": "NON_PARTICIPATING",
 
74
  "160": "NON_PARTICIPATING",
75
  "161": "NON_PARTICIPATING",
76
  "162": "NON_PARTICIPATING",
77
+ "163": "SUCCESS",
78
  "164": "NON_PARTICIPATING",
79
  "165": "NON_PARTICIPATING",
80
  "166": "NON_PARTICIPATING",
81
+ "167": "NON_PARTICIPATING",
82
  "168": "NON_PARTICIPATING",
83
+ "169": "SUCCESS",
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",
 
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",
 
119
  "200": "NON_PARTICIPATING",
120
  "201": "NON_PARTICIPATING",
121
  "202": "NON_PARTICIPATING",
122
+ "203": "SUCCESS",
123
  "204": "NON_PARTICIPATING",
124
+ "205": "SUCCESS",
125
  "206": "NON_PARTICIPATING",
126
  "207": "NON_PARTICIPATING",
127
  "208": "NON_PARTICIPATING",
128
  "209": "NON_PARTICIPATING",
129
  "21": "SUCCESS",
130
+ "210": "SUCCESS",
131
  "211": "NON_PARTICIPATING",
132
+ "212": "NON_PARTICIPATING",
133
  "213": "NON_PARTICIPATING",
134
  "214": "NON_PARTICIPATING",
135
  "215": "NON_PARTICIPATING",
136
  "216": "NON_PARTICIPATING",
137
  "217": "NON_PARTICIPATING",
138
  "218": "NON_PARTICIPATING",
139
+ "219": "NON_PARTICIPATING",
140
  "22": "SUCCESS",
141
  "220": "NON_PARTICIPATING",
142
  "221": "NON_PARTICIPATING",
 
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": "SUCCESS",
153
  "231": "NON_PARTICIPATING",
154
  "232": "NON_PARTICIPATING",
155
+ "233": "SUCCESS",
156
  "234": "NON_PARTICIPATING",
157
  "235": "NON_PARTICIPATING",
158
  "236": "NON_PARTICIPATING",
 
162
  "24": "SUCCESS",
163
  "240": "NON_PARTICIPATING",
164
  "241": "NON_PARTICIPATING",
165
+ "242": "SUCCESS",
166
  "243": "SUCCESS",
167
  "244": "NON_PARTICIPATING",
168
  "245": "NON_PARTICIPATING",
169
  "246": "NON_PARTICIPATING",
170
  "247": "NON_PARTICIPATING",
171
  "248": "NON_PARTICIPATING",
172
+ "249": "SUCCESS",
173
  "25": "SUCCESS",
174
  "250": "NON_PARTICIPATING",
175
  "251": "NON_PARTICIPATING",
 
219
  "61": "SUCCESS",
220
  "62": "SUCCESS",
221
  "63": "SUCCESS",
222
+ "64": "NON_PARTICIPATING",
223
  "65": "SUCCESS",
224
  "66": "SUCCESS",
225
  "67": "SUCCESS",
 
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
+ 5656339,
272
+ 5656353,
273
+ 5656357,
274
+ 5656362,
275
+ 5656366
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": 0,
283
  "inner_steps": 0,
284
+ "last_allreduce_block": 5656212,
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:aadf8246a47e7ed1be3454f0bc62fbf692f7dd881f573a9fe5ded54a7019b3a4
3
  size 8081782026
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa73acd18ec4b6746fc2763e417195c83e18f0a01c1eb3efea981b2144f28968
3
  size 8081782026
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1f6e75e70b372e70b49a4288533b9f8679c4a9f3a74bfc38a0fefc654b85c3de
3
  size 4040701744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f21556206cb761325e8119860c070bbcbc6f5e75bcda51489b0e434baa53d6e
3
  size 4040701744