Run 5. Outer Step 59. Inner Step 2.
Browse files- config.json +34 -34
- inner_optimizer.pt +1 -1
- model.safetensors +1 -1
config.json
CHANGED
|
@@ -24,13 +24,13 @@
|
|
| 24 |
"116": "NON_PARTICIPATING",
|
| 25 |
"117": "SUCCESS",
|
| 26 |
"118": "SUCCESS",
|
| 27 |
-
"119": "
|
| 28 |
"12": "SUCCESS",
|
| 29 |
-
"120": "
|
| 30 |
"121": "NON_PARTICIPATING",
|
| 31 |
"122": "NON_PARTICIPATING",
|
| 32 |
"123": "SUCCESS",
|
| 33 |
-
"124": "
|
| 34 |
"125": "SUCCESS",
|
| 35 |
"126": "SUCCESS",
|
| 36 |
"127": "NON_PARTICIPATING",
|
|
@@ -48,9 +48,9 @@
|
|
| 48 |
"138": "SUCCESS",
|
| 49 |
"139": "SUCCESS",
|
| 50 |
"14": "SUCCESS",
|
| 51 |
-
"140": "
|
| 52 |
"141": "NON_PARTICIPATING",
|
| 53 |
-
"142": "
|
| 54 |
"143": "SUCCESS",
|
| 55 |
"144": "NON_PARTICIPATING",
|
| 56 |
"145": "NON_PARTICIPATING",
|
|
@@ -61,7 +61,7 @@
|
|
| 61 |
"15": "SUCCESS",
|
| 62 |
"150": "NON_PARTICIPATING",
|
| 63 |
"151": "NON_PARTICIPATING",
|
| 64 |
-
"152": "
|
| 65 |
"153": "NON_PARTICIPATING",
|
| 66 |
"154": "NON_PARTICIPATING",
|
| 67 |
"155": "SUCCESS",
|
|
@@ -79,19 +79,19 @@
|
|
| 79 |
"166": "SUCCESS",
|
| 80 |
"167": "NON_PARTICIPATING",
|
| 81 |
"168": "NON_PARTICIPATING",
|
| 82 |
-
"169": "
|
| 83 |
"17": "SUCCESS",
|
| 84 |
"170": "SUCCESS",
|
| 85 |
"171": "SUCCESS",
|
| 86 |
"172": "NON_PARTICIPATING",
|
| 87 |
"173": "SUCCESS",
|
| 88 |
-
"174": "
|
| 89 |
-
"175": "
|
| 90 |
"176": "NON_PARTICIPATING",
|
| 91 |
"177": "SUCCESS",
|
| 92 |
"178": "SUCCESS",
|
| 93 |
"179": "NON_PARTICIPATING",
|
| 94 |
-
"18": "
|
| 95 |
"180": "SUCCESS",
|
| 96 |
"181": "SUCCESS",
|
| 97 |
"182": "SUCCESS",
|
|
@@ -145,10 +145,10 @@
|
|
| 145 |
"225": "NON_PARTICIPATING",
|
| 146 |
"226": "NON_PARTICIPATING",
|
| 147 |
"227": "SUCCESS",
|
| 148 |
-
"228": "
|
| 149 |
"229": "NON_PARTICIPATING",
|
| 150 |
"23": "SUCCESS",
|
| 151 |
-
"230": "
|
| 152 |
"231": "NON_PARTICIPATING",
|
| 153 |
"232": "SUCCESS",
|
| 154 |
"233": "NON_PARTICIPATING",
|
|
@@ -169,15 +169,15 @@
|
|
| 169 |
"247": "SUCCESS",
|
| 170 |
"248": "NON_PARTICIPATING",
|
| 171 |
"249": "SUCCESS",
|
| 172 |
-
"25": "
|
| 173 |
"250": "SUCCESS",
|
| 174 |
"251": "NON_PARTICIPATING",
|
| 175 |
"252": "NON_PARTICIPATING",
|
| 176 |
"253": "NON_PARTICIPATING",
|
| 177 |
-
"254": "
|
| 178 |
"255": "NON_PARTICIPATING",
|
| 179 |
"26": "SUCCESS",
|
| 180 |
-
"27": "
|
| 181 |
"28": "SUCCESS",
|
| 182 |
"29": "SUCCESS",
|
| 183 |
"3": "SUCCESS",
|
|
@@ -187,16 +187,16 @@
|
|
| 187 |
"33": "NON_PARTICIPATING",
|
| 188 |
"34": "SUCCESS",
|
| 189 |
"35": "SUCCESS",
|
| 190 |
-
"36": "
|
| 191 |
"37": "SUCCESS",
|
| 192 |
"38": "SUCCESS",
|
| 193 |
"39": "NON_PARTICIPATING",
|
| 194 |
"4": "SUCCESS",
|
| 195 |
"40": "SUCCESS",
|
| 196 |
-
"41": "
|
| 197 |
"42": "SUCCESS",
|
| 198 |
-
"43": "
|
| 199 |
-
"44": "
|
| 200 |
"45": "SUCCESS",
|
| 201 |
"46": "SUCCESS",
|
| 202 |
"47": "SUCCESS",
|
|
@@ -205,18 +205,18 @@
|
|
| 205 |
"5": "SUCCESS",
|
| 206 |
"50": "NON_PARTICIPATING",
|
| 207 |
"51": "FAIL",
|
| 208 |
-
"52": "
|
| 209 |
"53": "SUCCESS",
|
| 210 |
-
"54": "
|
| 211 |
"55": "SUCCESS",
|
| 212 |
"56": "SUCCESS",
|
| 213 |
"57": "NON_PARTICIPATING",
|
| 214 |
-
"58": "
|
| 215 |
"59": "SUCCESS",
|
| 216 |
"6": "NON_PARTICIPATING",
|
| 217 |
"60": "SUCCESS",
|
| 218 |
"61": "NON_PARTICIPATING",
|
| 219 |
-
"62": "
|
| 220 |
"63": "NON_PARTICIPATING",
|
| 221 |
"64": "SUCCESS",
|
| 222 |
"65": "SUCCESS",
|
|
@@ -228,22 +228,22 @@
|
|
| 228 |
"70": "SUCCESS",
|
| 229 |
"71": "SUCCESS",
|
| 230 |
"72": "SUCCESS",
|
| 231 |
-
"73": "
|
| 232 |
"74": "SUCCESS",
|
| 233 |
"75": "SUCCESS",
|
| 234 |
"76": "SUCCESS",
|
| 235 |
"77": "SUCCESS",
|
| 236 |
-
"78": "
|
| 237 |
-
"79": "
|
| 238 |
"8": "NON_PARTICIPATING",
|
| 239 |
"80": "SUCCESS",
|
| 240 |
"81": "SUCCESS",
|
| 241 |
-
"82": "
|
| 242 |
"83": "SUCCESS",
|
| 243 |
"84": "SUCCESS",
|
| 244 |
"85": "SUCCESS",
|
| 245 |
"86": "SUCCESS",
|
| 246 |
-
"87": "
|
| 247 |
"88": "SUCCESS",
|
| 248 |
"89": "SUCCESS",
|
| 249 |
"9": "SUCCESS",
|
|
@@ -251,11 +251,11 @@
|
|
| 251 |
"91": "NON_PARTICIPATING",
|
| 252 |
"92": "SUCCESS",
|
| 253 |
"93": "SUCCESS",
|
| 254 |
-
"94": "
|
| 255 |
"95": "SUCCESS",
|
| 256 |
"96": "SUCCESS",
|
| 257 |
"97": "NON_PARTICIPATING",
|
| 258 |
-
"98": "
|
| 259 |
"99": "SUCCESS"
|
| 260 |
},
|
| 261 |
"architectures": [
|
|
@@ -264,17 +264,17 @@
|
|
| 264 |
"attention_bias": false,
|
| 265 |
"attention_dropout": 0.0,
|
| 266 |
"block_list": [
|
| 267 |
-
|
| 268 |
-
|
| 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":
|
| 276 |
"intermediate_size": 5632,
|
| 277 |
-
"last_allreduce_block":
|
| 278 |
"max_position_embeddings": 2048,
|
| 279 |
"mlp_bias": false,
|
| 280 |
"model_type": "llama",
|
|
|
|
| 24 |
"116": "NON_PARTICIPATING",
|
| 25 |
"117": "SUCCESS",
|
| 26 |
"118": "SUCCESS",
|
| 27 |
+
"119": "NON_PARTICIPATING",
|
| 28 |
"12": "SUCCESS",
|
| 29 |
+
"120": "SUCCESS",
|
| 30 |
"121": "NON_PARTICIPATING",
|
| 31 |
"122": "NON_PARTICIPATING",
|
| 32 |
"123": "SUCCESS",
|
| 33 |
+
"124": "SUCCESS",
|
| 34 |
"125": "SUCCESS",
|
| 35 |
"126": "SUCCESS",
|
| 36 |
"127": "NON_PARTICIPATING",
|
|
|
|
| 48 |
"138": "SUCCESS",
|
| 49 |
"139": "SUCCESS",
|
| 50 |
"14": "SUCCESS",
|
| 51 |
+
"140": "SUCCESS",
|
| 52 |
"141": "NON_PARTICIPATING",
|
| 53 |
+
"142": "NON_PARTICIPATING",
|
| 54 |
"143": "SUCCESS",
|
| 55 |
"144": "NON_PARTICIPATING",
|
| 56 |
"145": "NON_PARTICIPATING",
|
|
|
|
| 61 |
"15": "SUCCESS",
|
| 62 |
"150": "NON_PARTICIPATING",
|
| 63 |
"151": "NON_PARTICIPATING",
|
| 64 |
+
"152": "NON_PARTICIPATING",
|
| 65 |
"153": "NON_PARTICIPATING",
|
| 66 |
"154": "NON_PARTICIPATING",
|
| 67 |
"155": "SUCCESS",
|
|
|
|
| 79 |
"166": "SUCCESS",
|
| 80 |
"167": "NON_PARTICIPATING",
|
| 81 |
"168": "NON_PARTICIPATING",
|
| 82 |
+
"169": "NON_PARTICIPATING",
|
| 83 |
"17": "SUCCESS",
|
| 84 |
"170": "SUCCESS",
|
| 85 |
"171": "SUCCESS",
|
| 86 |
"172": "NON_PARTICIPATING",
|
| 87 |
"173": "SUCCESS",
|
| 88 |
+
"174": "SUCCESS",
|
| 89 |
+
"175": "NON_PARTICIPATING",
|
| 90 |
"176": "NON_PARTICIPATING",
|
| 91 |
"177": "SUCCESS",
|
| 92 |
"178": "SUCCESS",
|
| 93 |
"179": "NON_PARTICIPATING",
|
| 94 |
+
"18": "NON_PARTICIPATING",
|
| 95 |
"180": "SUCCESS",
|
| 96 |
"181": "SUCCESS",
|
| 97 |
"182": "SUCCESS",
|
|
|
|
| 145 |
"225": "NON_PARTICIPATING",
|
| 146 |
"226": "NON_PARTICIPATING",
|
| 147 |
"227": "SUCCESS",
|
| 148 |
+
"228": "NON_PARTICIPATING",
|
| 149 |
"229": "NON_PARTICIPATING",
|
| 150 |
"23": "SUCCESS",
|
| 151 |
+
"230": "FAIL",
|
| 152 |
"231": "NON_PARTICIPATING",
|
| 153 |
"232": "SUCCESS",
|
| 154 |
"233": "NON_PARTICIPATING",
|
|
|
|
| 169 |
"247": "SUCCESS",
|
| 170 |
"248": "NON_PARTICIPATING",
|
| 171 |
"249": "SUCCESS",
|
| 172 |
+
"25": "NON_PARTICIPATING",
|
| 173 |
"250": "SUCCESS",
|
| 174 |
"251": "NON_PARTICIPATING",
|
| 175 |
"252": "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",
|
|
|
|
| 187 |
"33": "NON_PARTICIPATING",
|
| 188 |
"34": "SUCCESS",
|
| 189 |
"35": "SUCCESS",
|
| 190 |
+
"36": "SUCCESS",
|
| 191 |
"37": "SUCCESS",
|
| 192 |
"38": "SUCCESS",
|
| 193 |
"39": "NON_PARTICIPATING",
|
| 194 |
"4": "SUCCESS",
|
| 195 |
"40": "SUCCESS",
|
| 196 |
+
"41": "SUCCESS",
|
| 197 |
"42": "SUCCESS",
|
| 198 |
+
"43": "SUCCESS",
|
| 199 |
+
"44": "NON_PARTICIPATING",
|
| 200 |
"45": "SUCCESS",
|
| 201 |
"46": "SUCCESS",
|
| 202 |
"47": "SUCCESS",
|
|
|
|
| 205 |
"5": "SUCCESS",
|
| 206 |
"50": "NON_PARTICIPATING",
|
| 207 |
"51": "FAIL",
|
| 208 |
+
"52": "NON_PARTICIPATING",
|
| 209 |
"53": "SUCCESS",
|
| 210 |
+
"54": "FAIL",
|
| 211 |
"55": "SUCCESS",
|
| 212 |
"56": "SUCCESS",
|
| 213 |
"57": "NON_PARTICIPATING",
|
| 214 |
+
"58": "SUCCESS",
|
| 215 |
"59": "SUCCESS",
|
| 216 |
"6": "NON_PARTICIPATING",
|
| 217 |
"60": "SUCCESS",
|
| 218 |
"61": "NON_PARTICIPATING",
|
| 219 |
+
"62": "SUCCESS",
|
| 220 |
"63": "NON_PARTICIPATING",
|
| 221 |
"64": "SUCCESS",
|
| 222 |
"65": "SUCCESS",
|
|
|
|
| 228 |
"70": "SUCCESS",
|
| 229 |
"71": "SUCCESS",
|
| 230 |
"72": "SUCCESS",
|
| 231 |
+
"73": "SUCCESS",
|
| 232 |
"74": "SUCCESS",
|
| 233 |
"75": "SUCCESS",
|
| 234 |
"76": "SUCCESS",
|
| 235 |
"77": "SUCCESS",
|
| 236 |
+
"78": "SUCCESS",
|
| 237 |
+
"79": "SUCCESS",
|
| 238 |
"8": "NON_PARTICIPATING",
|
| 239 |
"80": "SUCCESS",
|
| 240 |
"81": "SUCCESS",
|
| 241 |
+
"82": "FAIL",
|
| 242 |
"83": "SUCCESS",
|
| 243 |
"84": "SUCCESS",
|
| 244 |
"85": "SUCCESS",
|
| 245 |
"86": "SUCCESS",
|
| 246 |
+
"87": "SUCCESS",
|
| 247 |
"88": "SUCCESS",
|
| 248 |
"89": "SUCCESS",
|
| 249 |
"9": "SUCCESS",
|
|
|
|
| 251 |
"91": "NON_PARTICIPATING",
|
| 252 |
"92": "SUCCESS",
|
| 253 |
"93": "SUCCESS",
|
| 254 |
+
"94": "FAIL",
|
| 255 |
"95": "SUCCESS",
|
| 256 |
"96": "SUCCESS",
|
| 257 |
"97": "NON_PARTICIPATING",
|
| 258 |
+
"98": "SUCCESS",
|
| 259 |
"99": "SUCCESS"
|
| 260 |
},
|
| 261 |
"architectures": [
|
|
|
|
| 264 |
"attention_bias": false,
|
| 265 |
"attention_dropout": 0.0,
|
| 266 |
"block_list": [
|
| 267 |
+
5799902,
|
| 268 |
+
5799916
|
| 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": 5798733,
|
| 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:
|
| 3 |
size 8800564561
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a991a70354f91154e923ae39df4f88f45bbf0495ce44961e8320f8f2a210084c
|
| 3 |
size 8800564561
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4400216536
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5d9538af2863781dc2d13bc2f4743cff757e219d9e64a57bf584664e85cb0a1b
|
| 3 |
size 4400216536
|