markalan324 commited on
Commit
8d59ebc
·
verified ·
1 Parent(s): 00533ef

Run 5. Outer Step 17. Inner Step 2.

Browse files
Files changed (3) hide show
  1. config.json +22 -22
  2. inner_optimizer.pt +1 -1
  3. model.safetensors +1 -1
config.json CHANGED
@@ -11,7 +11,7 @@
11
  "104": "NON_PARTICIPATING",
12
  "105": "NON_PARTICIPATING",
13
  "106": "NON_PARTICIPATING",
14
- "107": "SUCCESS",
15
  "108": "NON_PARTICIPATING",
16
  "109": "NON_PARTICIPATING",
17
  "11": "SUCCESS",
@@ -28,11 +28,11 @@
28
  "12": "SUCCESS",
29
  "120": "NON_PARTICIPATING",
30
  "121": "NON_PARTICIPATING",
31
- "122": "SUCCESS",
32
  "123": "NON_PARTICIPATING",
33
  "124": "NON_PARTICIPATING",
34
  "125": "NON_PARTICIPATING",
35
- "126": "NON_PARTICIPATING",
36
  "127": "NON_PARTICIPATING",
37
  "128": "NON_PARTICIPATING",
38
  "129": "NON_PARTICIPATING",
@@ -44,13 +44,13 @@
44
  "134": "NON_PARTICIPATING",
45
  "135": "NON_PARTICIPATING",
46
  "136": "NON_PARTICIPATING",
47
- "137": "NON_PARTICIPATING",
48
  "138": "NON_PARTICIPATING",
49
  "139": "NON_PARTICIPATING",
50
  "14": "SUCCESS",
51
  "140": "NON_PARTICIPATING",
52
  "141": "NON_PARTICIPATING",
53
- "142": "SUCCESS",
54
  "143": "NON_PARTICIPATING",
55
  "144": "NON_PARTICIPATING",
56
  "145": "NON_PARTICIPATING",
@@ -71,7 +71,7 @@
71
  "159": "NON_PARTICIPATING",
72
  "16": "SUCCESS",
73
  "160": "NON_PARTICIPATING",
74
- "161": "SUCCESS",
75
  "162": "NON_PARTICIPATING",
76
  "163": "NON_PARTICIPATING",
77
  "164": "NON_PARTICIPATING",
@@ -96,7 +96,7 @@
96
  "181": "NON_PARTICIPATING",
97
  "182": "NON_PARTICIPATING",
98
  "183": "NON_PARTICIPATING",
99
- "184": "NON_PARTICIPATING",
100
  "185": "NON_PARTICIPATING",
101
  "186": "NON_PARTICIPATING",
102
  "187": "NON_PARTICIPATING",
@@ -107,7 +107,7 @@
107
  "191": "NON_PARTICIPATING",
108
  "192": "NON_PARTICIPATING",
109
  "193": "NON_PARTICIPATING",
110
- "194": "SUCCESS",
111
  "195": "NON_PARTICIPATING",
112
  "196": "NON_PARTICIPATING",
113
  "197": "NON_PARTICIPATING",
@@ -131,14 +131,14 @@
131
  "212": "NON_PARTICIPATING",
132
  "213": "NON_PARTICIPATING",
133
  "214": "NON_PARTICIPATING",
134
- "215": "SUCCESS",
135
  "216": "NON_PARTICIPATING",
136
  "217": "NON_PARTICIPATING",
137
- "218": "NON_PARTICIPATING",
138
- "219": "NON_PARTICIPATING",
139
  "22": "SUCCESS",
140
  "220": "NON_PARTICIPATING",
141
- "221": "SUCCESS",
142
  "222": "NON_PARTICIPATING",
143
  "223": "NON_PARTICIPATING",
144
  "224": "NON_PARTICIPATING",
@@ -159,7 +159,7 @@
159
  "238": "NON_PARTICIPATING",
160
  "239": "NON_PARTICIPATING",
161
  "24": "SUCCESS",
162
- "240": "NON_PARTICIPATING",
163
  "241": "NON_PARTICIPATING",
164
  "242": "NON_PARTICIPATING",
165
  "243": "NON_PARTICIPATING",
@@ -176,10 +176,10 @@
176
  "253": "NON_PARTICIPATING",
177
  "254": "NON_PARTICIPATING",
178
  "255": "NON_PARTICIPATING",
179
- "26": "NON_PARTICIPATING",
180
  "27": "SUCCESS",
181
  "28": "SUCCESS",
182
- "29": "FAIL",
183
  "3": "SUCCESS",
184
  "30": "SUCCESS",
185
  "31": "SUCCESS",
@@ -210,7 +210,7 @@
210
  "54": "SUCCESS",
211
  "55": "SUCCESS",
212
  "56": "SUCCESS",
213
- "57": "SUCCESS",
214
  "58": "NON_PARTICIPATING",
215
  "59": "SUCCESS",
216
  "6": "SUCCESS",
@@ -218,9 +218,9 @@
218
  "61": "SUCCESS",
219
  "62": "SUCCESS",
220
  "63": "SUCCESS",
221
- "64": "SUCCESS",
222
  "65": "SUCCESS",
223
- "66": "SUCCESS",
224
  "67": "SUCCESS",
225
  "68": "SUCCESS",
226
  "69": "SUCCESS",
@@ -264,17 +264,17 @@
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
- 5722540,
268
- 5722570
269
  ],
270
  "bos_token_id": 1,
271
  "eos_token_id": 2,
272
  "hidden_act": "silu",
273
  "hidden_size": 2048,
274
  "initializer_range": 0.02,
275
- "inner_step": 50,
276
  "intermediate_size": 5632,
277
- "last_allreduce_block": 5719828,
278
  "max_position_embeddings": 2048,
279
  "mlp_bias": false,
280
  "model_type": "llama",
 
11
  "104": "NON_PARTICIPATING",
12
  "105": "NON_PARTICIPATING",
13
  "106": "NON_PARTICIPATING",
14
+ "107": "NON_PARTICIPATING",
15
  "108": "NON_PARTICIPATING",
16
  "109": "NON_PARTICIPATING",
17
  "11": "SUCCESS",
 
28
  "12": "SUCCESS",
29
  "120": "NON_PARTICIPATING",
30
  "121": "NON_PARTICIPATING",
31
+ "122": "NON_PARTICIPATING",
32
  "123": "NON_PARTICIPATING",
33
  "124": "NON_PARTICIPATING",
34
  "125": "NON_PARTICIPATING",
35
+ "126": "SUCCESS",
36
  "127": "NON_PARTICIPATING",
37
  "128": "NON_PARTICIPATING",
38
  "129": "NON_PARTICIPATING",
 
44
  "134": "NON_PARTICIPATING",
45
  "135": "NON_PARTICIPATING",
46
  "136": "NON_PARTICIPATING",
47
+ "137": "SUCCESS",
48
  "138": "NON_PARTICIPATING",
49
  "139": "NON_PARTICIPATING",
50
  "14": "SUCCESS",
51
  "140": "NON_PARTICIPATING",
52
  "141": "NON_PARTICIPATING",
53
+ "142": "NON_PARTICIPATING",
54
  "143": "NON_PARTICIPATING",
55
  "144": "NON_PARTICIPATING",
56
  "145": "NON_PARTICIPATING",
 
71
  "159": "NON_PARTICIPATING",
72
  "16": "SUCCESS",
73
  "160": "NON_PARTICIPATING",
74
+ "161": "NON_PARTICIPATING",
75
  "162": "NON_PARTICIPATING",
76
  "163": "NON_PARTICIPATING",
77
  "164": "NON_PARTICIPATING",
 
96
  "181": "NON_PARTICIPATING",
97
  "182": "NON_PARTICIPATING",
98
  "183": "NON_PARTICIPATING",
99
+ "184": "SUCCESS",
100
  "185": "NON_PARTICIPATING",
101
  "186": "NON_PARTICIPATING",
102
  "187": "NON_PARTICIPATING",
 
107
  "191": "NON_PARTICIPATING",
108
  "192": "NON_PARTICIPATING",
109
  "193": "NON_PARTICIPATING",
110
+ "194": "NON_PARTICIPATING",
111
  "195": "NON_PARTICIPATING",
112
  "196": "NON_PARTICIPATING",
113
  "197": "NON_PARTICIPATING",
 
131
  "212": "NON_PARTICIPATING",
132
  "213": "NON_PARTICIPATING",
133
  "214": "NON_PARTICIPATING",
134
+ "215": "NON_PARTICIPATING",
135
  "216": "NON_PARTICIPATING",
136
  "217": "NON_PARTICIPATING",
137
+ "218": "SUCCESS",
138
+ "219": "SUCCESS",
139
  "22": "SUCCESS",
140
  "220": "NON_PARTICIPATING",
141
+ "221": "NON_PARTICIPATING",
142
  "222": "NON_PARTICIPATING",
143
  "223": "NON_PARTICIPATING",
144
  "224": "NON_PARTICIPATING",
 
159
  "238": "NON_PARTICIPATING",
160
  "239": "NON_PARTICIPATING",
161
  "24": "SUCCESS",
162
+ "240": "SUCCESS",
163
  "241": "NON_PARTICIPATING",
164
  "242": "NON_PARTICIPATING",
165
  "243": "NON_PARTICIPATING",
 
176
  "253": "NON_PARTICIPATING",
177
  "254": "NON_PARTICIPATING",
178
  "255": "NON_PARTICIPATING",
179
+ "26": "SUCCESS",
180
  "27": "SUCCESS",
181
  "28": "SUCCESS",
182
+ "29": "SUCCESS",
183
  "3": "SUCCESS",
184
  "30": "SUCCESS",
185
  "31": "SUCCESS",
 
210
  "54": "SUCCESS",
211
  "55": "SUCCESS",
212
  "56": "SUCCESS",
213
+ "57": "FAIL",
214
  "58": "NON_PARTICIPATING",
215
  "59": "SUCCESS",
216
  "6": "SUCCESS",
 
218
  "61": "SUCCESS",
219
  "62": "SUCCESS",
220
  "63": "SUCCESS",
221
+ "64": "NON_PARTICIPATING",
222
  "65": "SUCCESS",
223
+ "66": "FAIL",
224
  "67": "SUCCESS",
225
  "68": "SUCCESS",
226
  "69": "SUCCESS",
 
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
+ 5722753,
268
+ 5722782
269
  ],
270
  "bos_token_id": 1,
271
  "eos_token_id": 2,
272
  "hidden_act": "silu",
273
  "hidden_size": 2048,
274
  "initializer_range": 0.02,
275
+ "inner_step": 2,
276
  "intermediate_size": 5632,
277
+ "last_allreduce_block": 5722667,
278
  "max_position_embeddings": 2048,
279
  "mlp_bias": false,
280
  "model_type": "llama",
inner_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:32e061d9c30506b67500b61771208bf61bbd32f817db192af034e393be07be55
3
  size 8800564561
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f913c7b3a5f798e8b4916f437cf0083237c8b64b0d1139ea1a7316df81a13080
3
  size 8800564561
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a86390f03066f390a7accdc06da04024c6f91a50d745874fb64aff622baeb758
3
  size 4400216536
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b86f1f66adff8797441229191b7462c407ff14c27999e78b2920811db8d5382d
3
  size 4400216536