zeras141a commited on
Commit
3f8091d
·
verified ·
1 Parent(s): 60fab82

Run 6. Outer Step 40. Inner Step 9.

Browse files
Files changed (4) hide show
  1. config.json +25 -25
  2. gradients.pt +1 -1
  3. inner_optimizer.pt +1 -1
  4. model.safetensors +1 -1
config.json CHANGED
@@ -12,7 +12,7 @@
12
  "105": "NOT_ALIVE",
13
  "106": "NOT_ALIVE",
14
  "107": "NOT_ALIVE",
15
- "108": "NON_PARTICIPATING",
16
  "109": "NOT_ALIVE",
17
  "11": "NOT_ALIVE",
18
  "110": "NOT_ALIVE",
@@ -30,7 +30,7 @@
30
  "121": "NOT_ALIVE",
31
  "122": "NOT_ALIVE",
32
  "123": "NOT_ALIVE",
33
- "124": "NON_PARTICIPATING",
34
  "125": "NOT_ALIVE",
35
  "126": "NOT_ALIVE",
36
  "127": "NOT_ALIVE",
@@ -40,7 +40,7 @@
40
  "130": "NOT_ALIVE",
41
  "131": "NOT_ALIVE",
42
  "132": "NOT_ALIVE",
43
- "133": "NOT_ALIVE",
44
  "134": "NOT_ALIVE",
45
  "135": "NOT_ALIVE",
46
  "136": "NOT_ALIVE",
@@ -52,7 +52,7 @@
52
  "141": "NOT_ALIVE",
53
  "142": "NOT_ALIVE",
54
  "143": "NOT_ALIVE",
55
- "144": "SUCCESS",
56
  "145": "NOT_ALIVE",
57
  "146": "NOT_ALIVE",
58
  "147": "NOT_ALIVE",
@@ -75,11 +75,11 @@
75
  "162": "NOT_ALIVE",
76
  "163": "NON_PARTICIPATING",
77
  "164": "SUCCESS",
78
- "165": "SUCCESS",
79
  "166": "SUCCESS",
80
  "167": "NOT_ALIVE",
81
  "168": "NOT_ALIVE",
82
- "169": "NON_PARTICIPATING",
83
  "17": "NOT_ALIVE",
84
  "170": "NOT_ALIVE",
85
  "171": "NOT_ALIVE",
@@ -99,12 +99,12 @@
99
  "184": "NOT_ALIVE",
100
  "185": "NON_PARTICIPATING",
101
  "186": "NOT_ALIVE",
102
- "187": "NOT_ALIVE",
103
  "188": "NOT_ALIVE",
104
  "189": "SUCCESS",
105
  "19": "NOT_ALIVE",
106
- "190": "NON_PARTICIPATING",
107
- "191": "NON_PARTICIPATING",
108
  "192": "NOT_ALIVE",
109
  "193": "SUCCESS",
110
  "194": "NOT_ALIVE",
@@ -113,7 +113,7 @@
113
  "197": "NOT_ALIVE",
114
  "198": "NOT_ALIVE",
115
  "199": "NOT_ALIVE",
116
- "2": "NOT_ALIVE",
117
  "20": "NOT_ALIVE",
118
  "200": "NOT_ALIVE",
119
  "201": "SUCCESS",
@@ -128,14 +128,14 @@
128
  "21": "NOT_ALIVE",
129
  "210": "NOT_ALIVE",
130
  "211": "NOT_ALIVE",
131
- "212": "SUCCESS",
132
  "213": "NOT_ALIVE",
133
  "214": "NON_PARTICIPATING",
134
- "215": "NOT_ALIVE",
135
- "216": "NOT_ALIVE",
136
  "217": "NOT_ALIVE",
137
  "218": "NON_PARTICIPATING",
138
- "219": "NOT_ALIVE",
139
  "22": "NOT_ALIVE",
140
  "220": "NOT_ALIVE",
141
  "221": "NOT_ALIVE",
@@ -155,7 +155,7 @@
155
  "234": "NOT_ALIVE",
156
  "235": "NOT_ALIVE",
157
  "236": "NOT_ALIVE",
158
- "237": "NON_PARTICIPATING",
159
  "238": "NOT_ALIVE",
160
  "239": "NOT_ALIVE",
161
  "24": "NOT_ALIVE",
@@ -177,12 +177,12 @@
177
  "254": "NOT_ALIVE",
178
  "255": "NOT_ALIVE",
179
  "26": "NON_PARTICIPATING",
180
- "27": "NOT_ALIVE",
181
- "28": "NON_PARTICIPATING",
182
  "29": "NOT_ALIVE",
183
- "3": "SUCCESS",
184
  "30": "NOT_ALIVE",
185
- "31": "SUCCESS",
186
  "32": "SUCCESS",
187
  "33": "NOT_ALIVE",
188
  "34": "NOT_ALIVE",
@@ -209,7 +209,7 @@
209
  "53": "SUCCESS",
210
  "54": "NOT_ALIVE",
211
  "55": "NOT_ALIVE",
212
- "56": "NON_PARTICIPATING",
213
  "57": "NON_PARTICIPATING",
214
  "58": "NOT_ALIVE",
215
  "59": "NOT_ALIVE",
@@ -256,7 +256,7 @@
256
  "96": "NON_PARTICIPATING",
257
  "97": "NOT_ALIVE",
258
  "98": "NOT_ALIVE",
259
- "99": "SUCCESS"
260
  },
261
  "architectures": [
262
  "LlamaForCausalLM"
@@ -264,17 +264,17 @@
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
- 6155548,
268
- 6155587
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": 68,
276
  "intermediate_size": 5632,
277
- "last_allreduce_block": 6154984,
278
  "max_position_embeddings": 2048,
279
  "mlp_bias": false,
280
  "model_type": "llama",
 
12
  "105": "NOT_ALIVE",
13
  "106": "NOT_ALIVE",
14
  "107": "NOT_ALIVE",
15
+ "108": "SUCCESS",
16
  "109": "NOT_ALIVE",
17
  "11": "NOT_ALIVE",
18
  "110": "NOT_ALIVE",
 
30
  "121": "NOT_ALIVE",
31
  "122": "NOT_ALIVE",
32
  "123": "NOT_ALIVE",
33
+ "124": "NOT_ALIVE",
34
  "125": "NOT_ALIVE",
35
  "126": "NOT_ALIVE",
36
  "127": "NOT_ALIVE",
 
40
  "130": "NOT_ALIVE",
41
  "131": "NOT_ALIVE",
42
  "132": "NOT_ALIVE",
43
+ "133": "NON_PARTICIPATING",
44
  "134": "NOT_ALIVE",
45
  "135": "NOT_ALIVE",
46
  "136": "NOT_ALIVE",
 
52
  "141": "NOT_ALIVE",
53
  "142": "NOT_ALIVE",
54
  "143": "NOT_ALIVE",
55
+ "144": "NOT_ALIVE",
56
  "145": "NOT_ALIVE",
57
  "146": "NOT_ALIVE",
58
  "147": "NOT_ALIVE",
 
75
  "162": "NOT_ALIVE",
76
  "163": "NON_PARTICIPATING",
77
  "164": "SUCCESS",
78
+ "165": "NON_PARTICIPATING",
79
  "166": "SUCCESS",
80
  "167": "NOT_ALIVE",
81
  "168": "NOT_ALIVE",
82
+ "169": "NOT_ALIVE",
83
  "17": "NOT_ALIVE",
84
  "170": "NOT_ALIVE",
85
  "171": "NOT_ALIVE",
 
99
  "184": "NOT_ALIVE",
100
  "185": "NON_PARTICIPATING",
101
  "186": "NOT_ALIVE",
102
+ "187": "SUCCESS",
103
  "188": "NOT_ALIVE",
104
  "189": "SUCCESS",
105
  "19": "NOT_ALIVE",
106
+ "190": "SUCCESS",
107
+ "191": "SUCCESS",
108
  "192": "NOT_ALIVE",
109
  "193": "SUCCESS",
110
  "194": "NOT_ALIVE",
 
113
  "197": "NOT_ALIVE",
114
  "198": "NOT_ALIVE",
115
  "199": "NOT_ALIVE",
116
+ "2": "SUCCESS",
117
  "20": "NOT_ALIVE",
118
  "200": "NOT_ALIVE",
119
  "201": "SUCCESS",
 
128
  "21": "NOT_ALIVE",
129
  "210": "NOT_ALIVE",
130
  "211": "NOT_ALIVE",
131
+ "212": "NON_PARTICIPATING",
132
  "213": "NOT_ALIVE",
133
  "214": "NON_PARTICIPATING",
134
+ "215": "NON_PARTICIPATING",
135
+ "216": "NON_PARTICIPATING",
136
  "217": "NOT_ALIVE",
137
  "218": "NON_PARTICIPATING",
138
+ "219": "SUCCESS",
139
  "22": "NOT_ALIVE",
140
  "220": "NOT_ALIVE",
141
  "221": "NOT_ALIVE",
 
155
  "234": "NOT_ALIVE",
156
  "235": "NOT_ALIVE",
157
  "236": "NOT_ALIVE",
158
+ "237": "NOT_ALIVE",
159
  "238": "NOT_ALIVE",
160
  "239": "NOT_ALIVE",
161
  "24": "NOT_ALIVE",
 
177
  "254": "NOT_ALIVE",
178
  "255": "NOT_ALIVE",
179
  "26": "NON_PARTICIPATING",
180
+ "27": "NON_PARTICIPATING",
181
+ "28": "NOT_ALIVE",
182
  "29": "NOT_ALIVE",
183
+ "3": "NOT_ALIVE",
184
  "30": "NOT_ALIVE",
185
+ "31": "NOT_ALIVE",
186
  "32": "SUCCESS",
187
  "33": "NOT_ALIVE",
188
  "34": "NOT_ALIVE",
 
209
  "53": "SUCCESS",
210
  "54": "NOT_ALIVE",
211
  "55": "NOT_ALIVE",
212
+ "56": "NOT_ALIVE",
213
  "57": "NON_PARTICIPATING",
214
  "58": "NOT_ALIVE",
215
  "59": "NOT_ALIVE",
 
256
  "96": "NON_PARTICIPATING",
257
  "97": "NOT_ALIVE",
258
  "98": "NOT_ALIVE",
259
+ "99": "NOT_ALIVE"
260
  },
261
  "architectures": [
262
  "LlamaForCausalLM"
 
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
+ 6155708,
268
+ 6155745
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": 9,
276
  "intermediate_size": 5632,
277
+ "last_allreduce_block": 6155595,
278
  "max_position_embeddings": 2048,
279
  "mlp_bias": false,
280
  "model_type": "llama",
gradients.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:26933ca8fb78009f45d30451751f3ca6c9f4fc7dadafa90a95bf4abaf11a61df
3
  size 88389626
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c269b426c1cfbd1d58f2adca786c687cb347eece9359f9f59ed68c16765264d6
3
  size 88389626
inner_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2e53495d48cc41a0a04da48a76b3ad3d682c3ec9da47d97c77981a8c24e7e406
3
  size 8800564076
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f2e1a7f30f9f3c1d1dfdccf9ab1d67c3208b978378615b20fe13c0a9c6549e8
3
  size 8800564076
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:04359193a187489e3b75f5f1da29d7a427a90bd90af5f501cb6804f2f7dc03eb
3
  size 4400216536
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6aeb9155cc7e5abf57753a4cb719bff30061b396ce74435740b4eed4af31291
3
  size 4400216536