jkjjbk commited on
Commit
82a8e34
·
verified ·
1 Parent(s): c4a1b66

Run 6. Outer Step 42. Inner Step 5.

Browse files
Files changed (4) hide show
  1. config.json +22 -23
  2. gradients.pt +1 -1
  3. inner_optimizer.pt +1 -1
  4. model.safetensors +1 -1
config.json CHANGED
@@ -74,9 +74,9 @@
74
  "161": "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",
@@ -101,10 +101,10 @@
101
  "186": "NOT_ALIVE",
102
  "187": "SUCCESS",
103
  "188": "NOT_ALIVE",
104
- "189": "SUCCESS",
105
  "19": "NOT_ALIVE",
106
  "190": "SUCCESS",
107
- "191": "NON_PARTICIPATING",
108
  "192": "NOT_ALIVE",
109
  "193": "SUCCESS",
110
  "194": "NOT_ALIVE",
@@ -112,8 +112,8 @@
112
  "196": "NOT_ALIVE",
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",
@@ -130,10 +130,10 @@
130
  "211": "NOT_ALIVE",
131
  "212": "SUCCESS",
132
  "213": "NOT_ALIVE",
133
- "214": "NON_PARTICIPATING",
134
  "215": "NON_PARTICIPATING",
135
  "216": "NOT_ALIVE",
136
- "217": "NOT_ALIVE",
137
  "218": "NON_PARTICIPATING",
138
  "219": "SUCCESS",
139
  "22": "NOT_ALIVE",
@@ -141,7 +141,7 @@
141
  "221": "NOT_ALIVE",
142
  "222": "NOT_ALIVE",
143
  "223": "NOT_ALIVE",
144
- "224": "NOT_ALIVE",
145
  "225": "NOT_ALIVE",
146
  "226": "NON_PARTICIPATING",
147
  "227": "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": "NOT_ALIVE",
186
  "32": "SUCCESS",
187
  "33": "NOT_ALIVE",
188
  "34": "NOT_ALIVE",
@@ -192,7 +192,7 @@
192
  "38": "NOT_ALIVE",
193
  "39": "NOT_ALIVE",
194
  "4": "NOT_ALIVE",
195
- "40": "SUCCESS",
196
  "41": "NOT_ALIVE",
197
  "42": "NOT_ALIVE",
198
  "43": "NOT_ALIVE",
@@ -209,7 +209,7 @@
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",
@@ -219,14 +219,14 @@
219
  "62": "NOT_ALIVE",
220
  "63": "NOT_ALIVE",
221
  "64": "NOT_ALIVE",
222
- "65": "NON_PARTICIPATING",
223
  "66": "NOT_ALIVE",
224
  "67": "NOT_ALIVE",
225
  "68": "NOT_ALIVE",
226
  "69": "NOT_ALIVE",
227
  "7": "NOT_ALIVE",
228
- "70": "NOT_ALIVE",
229
- "71": "SUCCESS",
230
  "72": "NOT_ALIVE",
231
  "73": "NOT_ALIVE",
232
  "74": "NOT_ALIVE",
@@ -252,8 +252,8 @@
252
  "92": "NOT_ALIVE",
253
  "93": "NOT_ALIVE",
254
  "94": "NOT_ALIVE",
255
- "95": "NOT_ALIVE",
256
- "96": "NON_PARTICIPATING",
257
  "97": "NOT_ALIVE",
258
  "98": "NOT_ALIVE",
259
  "99": "SUCCESS"
@@ -264,18 +264,17 @@
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
- 6156802,
268
- 6156815,
269
- 6156828
270
  ],
271
  "bos_token_id": 1,
272
  "eos_token_id": 2,
273
  "hidden_act": "silu",
274
  "hidden_size": 2048,
275
  "initializer_range": 0.02,
276
- "inner_step": 21,
277
  "intermediate_size": 5632,
278
- "last_allreduce_block": 6156228,
279
  "max_position_embeddings": 2048,
280
  "mlp_bias": false,
281
  "model_type": "llama",
 
74
  "161": "NOT_ALIVE",
75
  "162": "NOT_ALIVE",
76
  "163": "NON_PARTICIPATING",
77
+ "164": "NOT_ALIVE",
78
  "165": "NON_PARTICIPATING",
79
+ "166": "NOT_ALIVE",
80
  "167": "NOT_ALIVE",
81
  "168": "NOT_ALIVE",
82
  "169": "NOT_ALIVE",
 
101
  "186": "NOT_ALIVE",
102
  "187": "SUCCESS",
103
  "188": "NOT_ALIVE",
104
+ "189": "NOT_ALIVE",
105
  "19": "NOT_ALIVE",
106
  "190": "SUCCESS",
107
+ "191": "NOT_ALIVE",
108
  "192": "NOT_ALIVE",
109
  "193": "SUCCESS",
110
  "194": "NOT_ALIVE",
 
112
  "196": "NOT_ALIVE",
113
  "197": "NOT_ALIVE",
114
  "198": "NOT_ALIVE",
115
+ "199": "SUCCESS",
116
+ "2": "SUCCESS",
117
  "20": "NOT_ALIVE",
118
  "200": "NOT_ALIVE",
119
  "201": "SUCCESS",
 
130
  "211": "NOT_ALIVE",
131
  "212": "SUCCESS",
132
  "213": "NOT_ALIVE",
133
+ "214": "NOT_ALIVE",
134
  "215": "NON_PARTICIPATING",
135
  "216": "NOT_ALIVE",
136
+ "217": "SUCCESS",
137
  "218": "NON_PARTICIPATING",
138
  "219": "SUCCESS",
139
  "22": "NOT_ALIVE",
 
141
  "221": "NOT_ALIVE",
142
  "222": "NOT_ALIVE",
143
  "223": "NOT_ALIVE",
144
+ "224": "NON_PARTICIPATING",
145
  "225": "NOT_ALIVE",
146
  "226": "NON_PARTICIPATING",
147
  "227": "NOT_ALIVE",
 
177
  "254": "NOT_ALIVE",
178
  "255": "NOT_ALIVE",
179
  "26": "NON_PARTICIPATING",
180
+ "27": "NON_PARTICIPATING",
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",
 
192
  "38": "NOT_ALIVE",
193
  "39": "NOT_ALIVE",
194
  "4": "NOT_ALIVE",
195
+ "40": "NOT_ALIVE",
196
  "41": "NOT_ALIVE",
197
  "42": "NOT_ALIVE",
198
  "43": "NOT_ALIVE",
 
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",
 
219
  "62": "NOT_ALIVE",
220
  "63": "NOT_ALIVE",
221
  "64": "NOT_ALIVE",
222
+ "65": "NOT_ALIVE",
223
  "66": "NOT_ALIVE",
224
  "67": "NOT_ALIVE",
225
  "68": "NOT_ALIVE",
226
  "69": "NOT_ALIVE",
227
  "7": "NOT_ALIVE",
228
+ "70": "SUCCESS",
229
+ "71": "NON_PARTICIPATING",
230
  "72": "NOT_ALIVE",
231
  "73": "NOT_ALIVE",
232
  "74": "NOT_ALIVE",
 
252
  "92": "NOT_ALIVE",
253
  "93": "NOT_ALIVE",
254
  "94": "NOT_ALIVE",
255
+ "95": "NON_PARTICIPATING",
256
+ "96": "NOT_ALIVE",
257
  "97": "NOT_ALIVE",
258
  "98": "NOT_ALIVE",
259
  "99": "SUCCESS"
 
264
  "attention_bias": false,
265
  "attention_dropout": 0.0,
266
  "block_list": [
267
+ 6156987,
268
+ 6157005
 
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": 4,
276
  "intermediate_size": 5632,
277
+ "last_allreduce_block": 6156907,
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 sha256:8275098f50afce670d7ecfbfe1e2995cf826fbe0b3002d8e8cf6861c7937b042
3
  size 88389626
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7dac9ee8db569938d72ecba7a748df85d1e7e2b77a9e2c96684c5e300da8d1e5
3
  size 88389626
inner_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fcd22459672582d9d0e3b3498185abbc5a8c525f8c40182748262d033420b7a8
3
  size 8800564076
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88209a2c279abf796d7fc1c5a5cf928cb7ff791f4590329a74e25e7383e014f1
3
  size 8800564076
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ee3b87d28b1acb481c10133980b427feb05a38d3e09c8580e9fa448ac510d5f8
3
  size 4400216536
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab893817f69823d29ec066e2d6aaec5e4a3b956b5ab9cb55699c9ebc3df8847e
3
  size 4400216536