snoopsy commited on
Commit
b2f42b2
·
verified ·
1 Parent(s): 6dbb33c

Run 5. Outer Step 18. Inner Step 1.

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
@@ -22,17 +22,17 @@
22
  "114": "NON_PARTICIPATING",
23
  "115": "NON_PARTICIPATING",
24
  "116": "NON_PARTICIPATING",
25
- "117": "NON_PARTICIPATING",
26
  "118": "NON_PARTICIPATING",
27
  "119": "NON_PARTICIPATING",
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",
@@ -41,10 +41,10 @@
41
  "131": "NON_PARTICIPATING",
42
  "132": "NON_PARTICIPATING",
43
  "133": "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",
@@ -95,8 +95,8 @@
95
  "180": "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",
@@ -104,9 +104,9 @@
104
  "189": "NON_PARTICIPATING",
105
  "19": "SUCCESS",
106
  "190": "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",
@@ -117,7 +117,7 @@
117
  "20": "SUCCESS",
118
  "200": "NON_PARTICIPATING",
119
  "201": "NON_PARTICIPATING",
120
- "202": "NON_PARTICIPATING",
121
  "203": "NON_PARTICIPATING",
122
  "204": "NON_PARTICIPATING",
123
  "205": "NON_PARTICIPATING",
@@ -127,15 +127,15 @@
127
  "209": "NON_PARTICIPATING",
128
  "21": "SUCCESS",
129
  "210": "NON_PARTICIPATING",
130
- "211": "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",
@@ -159,7 +159,7 @@
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",
@@ -210,7 +210,7 @@
210
  "54": "SUCCESS",
211
  "55": "SUCCESS",
212
  "56": "SUCCESS",
213
- "57": "FAIL",
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": "NON_PARTICIPATING",
222
  "65": "SUCCESS",
223
- "66": "FAIL",
224
  "67": "SUCCESS",
225
  "68": "SUCCESS",
226
  "69": "SUCCESS",
@@ -254,7 +254,7 @@
254
  "94": "SUCCESS",
255
  "95": "SUCCESS",
256
  "96": "SUCCESS",
257
- "97": "SUCCESS",
258
  "98": "SUCCESS",
259
  "99": "SUCCESS"
260
  },
@@ -264,17 +264,17 @@
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
- 5724098,
268
- 5724114
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": 18,
276
  "intermediate_size": 5632,
277
- "last_allreduce_block": 5722667,
278
  "max_position_embeddings": 2048,
279
  "mlp_bias": false,
280
  "model_type": "llama",
 
22
  "114": "NON_PARTICIPATING",
23
  "115": "NON_PARTICIPATING",
24
  "116": "NON_PARTICIPATING",
25
+ "117": "SUCCESS",
26
  "118": "NON_PARTICIPATING",
27
  "119": "NON_PARTICIPATING",
28
  "12": "SUCCESS",
29
+ "120": "SUCCESS",
30
  "121": "NON_PARTICIPATING",
31
  "122": "NON_PARTICIPATING",
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",
 
41
  "131": "NON_PARTICIPATING",
42
  "132": "NON_PARTICIPATING",
43
  "133": "NON_PARTICIPATING",
44
+ "134": "SUCCESS",
45
  "135": "NON_PARTICIPATING",
46
  "136": "NON_PARTICIPATING",
47
+ "137": "NON_PARTICIPATING",
48
  "138": "NON_PARTICIPATING",
49
  "139": "NON_PARTICIPATING",
50
  "14": "SUCCESS",
 
95
  "180": "NON_PARTICIPATING",
96
  "181": "NON_PARTICIPATING",
97
  "182": "NON_PARTICIPATING",
98
+ "183": "SUCCESS",
99
+ "184": "NON_PARTICIPATING",
100
  "185": "NON_PARTICIPATING",
101
  "186": "NON_PARTICIPATING",
102
  "187": "NON_PARTICIPATING",
 
104
  "189": "NON_PARTICIPATING",
105
  "19": "SUCCESS",
106
  "190": "NON_PARTICIPATING",
107
+ "191": "SUCCESS",
108
  "192": "NON_PARTICIPATING",
109
+ "193": "SUCCESS",
110
  "194": "NON_PARTICIPATING",
111
  "195": "NON_PARTICIPATING",
112
  "196": "NON_PARTICIPATING",
 
117
  "20": "SUCCESS",
118
  "200": "NON_PARTICIPATING",
119
  "201": "NON_PARTICIPATING",
120
+ "202": "SUCCESS",
121
  "203": "NON_PARTICIPATING",
122
  "204": "NON_PARTICIPATING",
123
  "205": "NON_PARTICIPATING",
 
127
  "209": "NON_PARTICIPATING",
128
  "21": "SUCCESS",
129
  "210": "NON_PARTICIPATING",
130
+ "211": "SUCCESS",
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": "NON_PARTICIPATING",
138
+ "219": "NON_PARTICIPATING",
139
  "22": "SUCCESS",
140
  "220": "NON_PARTICIPATING",
141
  "221": "NON_PARTICIPATING",
 
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",
 
210
  "54": "SUCCESS",
211
  "55": "SUCCESS",
212
  "56": "SUCCESS",
213
+ "57": "SUCCESS",
214
  "58": "NON_PARTICIPATING",
215
  "59": "SUCCESS",
216
  "6": "SUCCESS",
 
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",
 
254
  "94": "SUCCESS",
255
  "95": "SUCCESS",
256
  "96": "SUCCESS",
257
+ "97": "NON_PARTICIPATING",
258
  "98": "SUCCESS",
259
  "99": "SUCCESS"
260
  },
 
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
+ 5724178,
268
+ 5724195
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": 1,
276
  "intermediate_size": 5632,
277
+ "last_allreduce_block": 5724117,
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:dd82f78aa206a122a44ef9af37bea2a9dde555f5651eeaa3be7479b5a30f7c32
3
  size 8800564076
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ef3cdcd668fa493fda546c75f754611bad76bbc6d5bddae249ececc7baa111a
3
  size 8800564076
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:42a9c0f4a858aeeeeb775fac84832c70b810db8347d6e578386bed75acb2eb96
3
  size 4400216536
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f7b89dd4a715a7a8315521419ec64068cabd267497e7ea663772e20aae6bb90
3
  size 4400216536