Run 5. Outer Step 49. Inner Step 2.
Browse files- config.json +44 -44
- inner_optimizer.pt +2 -2
- model.safetensors +1 -1
config.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "
|
| 3 |
"all_reduce_scores": {
|
| 4 |
"0": "NON_PARTICIPATING",
|
| 5 |
"1": "SUCCESS",
|
|
@@ -17,12 +17,12 @@
|
|
| 17 |
"11": "SUCCESS",
|
| 18 |
"110": "NON_PARTICIPATING",
|
| 19 |
"111": "NON_PARTICIPATING",
|
| 20 |
-
"112": "
|
| 21 |
"113": "NON_PARTICIPATING",
|
| 22 |
"114": "NON_PARTICIPATING",
|
| 23 |
"115": "NON_PARTICIPATING",
|
| 24 |
"116": "NON_PARTICIPATING",
|
| 25 |
-
"117": "
|
| 26 |
"118": "NON_PARTICIPATING",
|
| 27 |
"119": "NON_PARTICIPATING",
|
| 28 |
"12": "SUCCESS",
|
|
@@ -32,11 +32,11 @@
|
|
| 32 |
"123": "NON_PARTICIPATING",
|
| 33 |
"124": "NON_PARTICIPATING",
|
| 34 |
"125": "NON_PARTICIPATING",
|
| 35 |
-
"126": "
|
| 36 |
"127": "NON_PARTICIPATING",
|
| 37 |
"128": "NON_PARTICIPATING",
|
| 38 |
"129": "NON_PARTICIPATING",
|
| 39 |
-
"13": "
|
| 40 |
"130": "NON_PARTICIPATING",
|
| 41 |
"131": "NON_PARTICIPATING",
|
| 42 |
"132": "NON_PARTICIPATING",
|
|
@@ -51,10 +51,10 @@
|
|
| 51 |
"140": "NON_PARTICIPATING",
|
| 52 |
"141": "NON_PARTICIPATING",
|
| 53 |
"142": "NON_PARTICIPATING",
|
| 54 |
-
"143": "
|
| 55 |
"144": "NON_PARTICIPATING",
|
| 56 |
"145": "NON_PARTICIPATING",
|
| 57 |
-
"146": "
|
| 58 |
"147": "NON_PARTICIPATING",
|
| 59 |
"148": "NON_PARTICIPATING",
|
| 60 |
"149": "NON_PARTICIPATING",
|
|
@@ -76,7 +76,7 @@
|
|
| 76 |
"163": "NON_PARTICIPATING",
|
| 77 |
"164": "NON_PARTICIPATING",
|
| 78 |
"165": "NON_PARTICIPATING",
|
| 79 |
-
"166": "
|
| 80 |
"167": "NON_PARTICIPATING",
|
| 81 |
"168": "NON_PARTICIPATING",
|
| 82 |
"169": "NON_PARTICIPATING",
|
|
@@ -89,24 +89,24 @@
|
|
| 89 |
"175": "NON_PARTICIPATING",
|
| 90 |
"176": "NON_PARTICIPATING",
|
| 91 |
"177": "NON_PARTICIPATING",
|
| 92 |
-
"178": "
|
| 93 |
"179": "NON_PARTICIPATING",
|
| 94 |
"18": "SUCCESS",
|
| 95 |
"180": "NON_PARTICIPATING",
|
| 96 |
"181": "NON_PARTICIPATING",
|
| 97 |
-
"182": "
|
| 98 |
"183": "NON_PARTICIPATING",
|
| 99 |
"184": "NON_PARTICIPATING",
|
| 100 |
"185": "NON_PARTICIPATING",
|
| 101 |
"186": "NON_PARTICIPATING",
|
| 102 |
"187": "NON_PARTICIPATING",
|
| 103 |
"188": "NON_PARTICIPATING",
|
| 104 |
-
"189": "
|
| 105 |
-
"19": "
|
| 106 |
"190": "NON_PARTICIPATING",
|
| 107 |
"191": "NON_PARTICIPATING",
|
| 108 |
-
"192": "
|
| 109 |
-
"193": "
|
| 110 |
"194": "NON_PARTICIPATING",
|
| 111 |
"195": "NON_PARTICIPATING",
|
| 112 |
"196": "NON_PARTICIPATING",
|
|
@@ -114,7 +114,7 @@
|
|
| 114 |
"198": "NON_PARTICIPATING",
|
| 115 |
"199": "NON_PARTICIPATING",
|
| 116 |
"2": "SUCCESS",
|
| 117 |
-
"20": "
|
| 118 |
"200": "NON_PARTICIPATING",
|
| 119 |
"201": "NON_PARTICIPATING",
|
| 120 |
"202": "NON_PARTICIPATING",
|
|
@@ -135,16 +135,16 @@
|
|
| 135 |
"216": "NON_PARTICIPATING",
|
| 136 |
"217": "NON_PARTICIPATING",
|
| 137 |
"218": "NON_PARTICIPATING",
|
| 138 |
-
"219": "
|
| 139 |
"22": "SUCCESS",
|
| 140 |
-
"220": "
|
| 141 |
"221": "NON_PARTICIPATING",
|
| 142 |
"222": "NON_PARTICIPATING",
|
| 143 |
"223": "NON_PARTICIPATING",
|
| 144 |
"224": "NON_PARTICIPATING",
|
| 145 |
"225": "NON_PARTICIPATING",
|
| 146 |
"226": "NON_PARTICIPATING",
|
| 147 |
-
"227": "
|
| 148 |
"228": "NON_PARTICIPATING",
|
| 149 |
"229": "NON_PARTICIPATING",
|
| 150 |
"23": "SUCCESS",
|
|
@@ -170,18 +170,18 @@
|
|
| 170 |
"248": "NON_PARTICIPATING",
|
| 171 |
"249": "NON_PARTICIPATING",
|
| 172 |
"25": "SUCCESS",
|
| 173 |
-
"250": "
|
| 174 |
"251": "NON_PARTICIPATING",
|
| 175 |
"252": "NON_PARTICIPATING",
|
| 176 |
"253": "NON_PARTICIPATING",
|
| 177 |
"254": "NON_PARTICIPATING",
|
| 178 |
"255": "NON_PARTICIPATING",
|
| 179 |
-
"26": "
|
| 180 |
"27": "SUCCESS",
|
| 181 |
"28": "SUCCESS",
|
| 182 |
"29": "SUCCESS",
|
| 183 |
"3": "SUCCESS",
|
| 184 |
-
"30": "
|
| 185 |
"31": "SUCCESS",
|
| 186 |
"32": "SUCCESS",
|
| 187 |
"33": "NON_PARTICIPATING",
|
|
@@ -189,45 +189,45 @@
|
|
| 189 |
"35": "SUCCESS",
|
| 190 |
"36": "SUCCESS",
|
| 191 |
"37": "SUCCESS",
|
| 192 |
-
"38": "
|
| 193 |
"39": "SUCCESS",
|
| 194 |
"4": "SUCCESS",
|
| 195 |
"40": "SUCCESS",
|
| 196 |
-
"41": "
|
| 197 |
"42": "SUCCESS",
|
| 198 |
"43": "SUCCESS",
|
| 199 |
-
"44": "
|
| 200 |
"45": "SUCCESS",
|
| 201 |
"46": "SUCCESS",
|
| 202 |
-
"47": "
|
| 203 |
"48": "SUCCESS",
|
| 204 |
"49": "SUCCESS",
|
| 205 |
"5": "SUCCESS",
|
| 206 |
-
"50": "
|
| 207 |
-
"51": "
|
| 208 |
-
"52": "
|
| 209 |
"53": "SUCCESS",
|
| 210 |
"54": "SUCCESS",
|
| 211 |
"55": "SUCCESS",
|
| 212 |
"56": "SUCCESS",
|
| 213 |
-
"57": "
|
| 214 |
"58": "SUCCESS",
|
| 215 |
"59": "SUCCESS",
|
| 216 |
-
"6": "
|
| 217 |
"60": "SUCCESS",
|
| 218 |
-
"61": "
|
| 219 |
"62": "SUCCESS",
|
| 220 |
"63": "SUCCESS",
|
| 221 |
"64": "NON_PARTICIPATING",
|
| 222 |
-
"65": "
|
| 223 |
-
"66": "
|
| 224 |
"67": "SUCCESS",
|
| 225 |
"68": "SUCCESS",
|
| 226 |
"69": "SUCCESS",
|
| 227 |
"7": "SUCCESS",
|
| 228 |
"70": "SUCCESS",
|
| 229 |
"71": "SUCCESS",
|
| 230 |
-
"72": "
|
| 231 |
"73": "SUCCESS",
|
| 232 |
"74": "SUCCESS",
|
| 233 |
"75": "SUCCESS",
|
|
@@ -235,7 +235,7 @@
|
|
| 235 |
"77": "NON_PARTICIPATING",
|
| 236 |
"78": "SUCCESS",
|
| 237 |
"79": "SUCCESS",
|
| 238 |
-
"8": "
|
| 239 |
"80": "SUCCESS",
|
| 240 |
"81": "SUCCESS",
|
| 241 |
"82": "SUCCESS",
|
|
@@ -246,15 +246,15 @@
|
|
| 246 |
"87": "SUCCESS",
|
| 247 |
"88": "SUCCESS",
|
| 248 |
"89": "SUCCESS",
|
| 249 |
-
"9": "
|
| 250 |
"90": "SUCCESS",
|
| 251 |
-
"91": "
|
| 252 |
"92": "SUCCESS",
|
| 253 |
-
"93": "
|
| 254 |
"94": "SUCCESS",
|
| 255 |
"95": "SUCCESS",
|
| 256 |
-
"96": "
|
| 257 |
-
"97": "
|
| 258 |
"98": "SUCCESS",
|
| 259 |
"99": "SUCCESS"
|
| 260 |
},
|
|
@@ -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",
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "distributed/llama-1b",
|
| 3 |
"all_reduce_scores": {
|
| 4 |
"0": "NON_PARTICIPATING",
|
| 5 |
"1": "SUCCESS",
|
|
|
|
| 17 |
"11": "SUCCESS",
|
| 18 |
"110": "NON_PARTICIPATING",
|
| 19 |
"111": "NON_PARTICIPATING",
|
| 20 |
+
"112": "SUCCESS",
|
| 21 |
"113": "NON_PARTICIPATING",
|
| 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",
|
|
|
|
| 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",
|
| 39 |
+
"13": "SUCCESS",
|
| 40 |
"130": "NON_PARTICIPATING",
|
| 41 |
"131": "NON_PARTICIPATING",
|
| 42 |
"132": "NON_PARTICIPATING",
|
|
|
|
| 51 |
"140": "NON_PARTICIPATING",
|
| 52 |
"141": "NON_PARTICIPATING",
|
| 53 |
"142": "NON_PARTICIPATING",
|
| 54 |
+
"143": "SUCCESS",
|
| 55 |
"144": "NON_PARTICIPATING",
|
| 56 |
"145": "NON_PARTICIPATING",
|
| 57 |
+
"146": "SUCCESS",
|
| 58 |
"147": "NON_PARTICIPATING",
|
| 59 |
"148": "NON_PARTICIPATING",
|
| 60 |
"149": "NON_PARTICIPATING",
|
|
|
|
| 76 |
"163": "NON_PARTICIPATING",
|
| 77 |
"164": "NON_PARTICIPATING",
|
| 78 |
"165": "NON_PARTICIPATING",
|
| 79 |
+
"166": "NON_PARTICIPATING",
|
| 80 |
"167": "NON_PARTICIPATING",
|
| 81 |
"168": "NON_PARTICIPATING",
|
| 82 |
"169": "NON_PARTICIPATING",
|
|
|
|
| 89 |
"175": "NON_PARTICIPATING",
|
| 90 |
"176": "NON_PARTICIPATING",
|
| 91 |
"177": "NON_PARTICIPATING",
|
| 92 |
+
"178": "NON_PARTICIPATING",
|
| 93 |
"179": "NON_PARTICIPATING",
|
| 94 |
"18": "SUCCESS",
|
| 95 |
"180": "NON_PARTICIPATING",
|
| 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",
|
| 103 |
"188": "NON_PARTICIPATING",
|
| 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",
|
|
|
|
| 114 |
"198": "NON_PARTICIPATING",
|
| 115 |
"199": "NON_PARTICIPATING",
|
| 116 |
"2": "SUCCESS",
|
| 117 |
+
"20": "SUCCESS",
|
| 118 |
"200": "NON_PARTICIPATING",
|
| 119 |
"201": "NON_PARTICIPATING",
|
| 120 |
"202": "NON_PARTICIPATING",
|
|
|
|
| 135 |
"216": "NON_PARTICIPATING",
|
| 136 |
"217": "NON_PARTICIPATING",
|
| 137 |
"218": "NON_PARTICIPATING",
|
| 138 |
+
"219": "SUCCESS",
|
| 139 |
"22": "SUCCESS",
|
| 140 |
+
"220": "SUCCESS",
|
| 141 |
"221": "NON_PARTICIPATING",
|
| 142 |
"222": "NON_PARTICIPATING",
|
| 143 |
"223": "NON_PARTICIPATING",
|
| 144 |
"224": "NON_PARTICIPATING",
|
| 145 |
"225": "NON_PARTICIPATING",
|
| 146 |
"226": "NON_PARTICIPATING",
|
| 147 |
+
"227": "NON_PARTICIPATING",
|
| 148 |
"228": "NON_PARTICIPATING",
|
| 149 |
"229": "NON_PARTICIPATING",
|
| 150 |
"23": "SUCCESS",
|
|
|
|
| 170 |
"248": "NON_PARTICIPATING",
|
| 171 |
"249": "NON_PARTICIPATING",
|
| 172 |
"25": "SUCCESS",
|
| 173 |
+
"250": "NON_PARTICIPATING",
|
| 174 |
"251": "NON_PARTICIPATING",
|
| 175 |
"252": "NON_PARTICIPATING",
|
| 176 |
"253": "NON_PARTICIPATING",
|
| 177 |
"254": "NON_PARTICIPATING",
|
| 178 |
"255": "NON_PARTICIPATING",
|
| 179 |
+
"26": "NON_PARTICIPATING",
|
| 180 |
"27": "SUCCESS",
|
| 181 |
"28": "SUCCESS",
|
| 182 |
"29": "SUCCESS",
|
| 183 |
"3": "SUCCESS",
|
| 184 |
+
"30": "SUCCESS",
|
| 185 |
"31": "SUCCESS",
|
| 186 |
"32": "SUCCESS",
|
| 187 |
"33": "NON_PARTICIPATING",
|
|
|
|
| 189 |
"35": "SUCCESS",
|
| 190 |
"36": "SUCCESS",
|
| 191 |
"37": "SUCCESS",
|
| 192 |
+
"38": "SUCCESS",
|
| 193 |
"39": "SUCCESS",
|
| 194 |
"4": "SUCCESS",
|
| 195 |
"40": "SUCCESS",
|
| 196 |
+
"41": "SUCCESS",
|
| 197 |
"42": "SUCCESS",
|
| 198 |
"43": "SUCCESS",
|
| 199 |
+
"44": "SUCCESS",
|
| 200 |
"45": "SUCCESS",
|
| 201 |
"46": "SUCCESS",
|
| 202 |
+
"47": "SUCCESS",
|
| 203 |
"48": "SUCCESS",
|
| 204 |
"49": "SUCCESS",
|
| 205 |
"5": "SUCCESS",
|
| 206 |
+
"50": "SUCCESS",
|
| 207 |
+
"51": "SUCCESS",
|
| 208 |
+
"52": "SUCCESS",
|
| 209 |
"53": "SUCCESS",
|
| 210 |
"54": "SUCCESS",
|
| 211 |
"55": "SUCCESS",
|
| 212 |
"56": "SUCCESS",
|
| 213 |
+
"57": "SUCCESS",
|
| 214 |
"58": "SUCCESS",
|
| 215 |
"59": "SUCCESS",
|
| 216 |
+
"6": "SUCCESS",
|
| 217 |
"60": "SUCCESS",
|
| 218 |
+
"61": "SUCCESS",
|
| 219 |
"62": "SUCCESS",
|
| 220 |
"63": "SUCCESS",
|
| 221 |
"64": "NON_PARTICIPATING",
|
| 222 |
+
"65": "NON_PARTICIPATING",
|
| 223 |
+
"66": "SUCCESS",
|
| 224 |
"67": "SUCCESS",
|
| 225 |
"68": "SUCCESS",
|
| 226 |
"69": "SUCCESS",
|
| 227 |
"7": "SUCCESS",
|
| 228 |
"70": "SUCCESS",
|
| 229 |
"71": "SUCCESS",
|
| 230 |
+
"72": "NON_PARTICIPATING",
|
| 231 |
"73": "SUCCESS",
|
| 232 |
"74": "SUCCESS",
|
| 233 |
"75": "SUCCESS",
|
|
|
|
| 235 |
"77": "NON_PARTICIPATING",
|
| 236 |
"78": "SUCCESS",
|
| 237 |
"79": "SUCCESS",
|
| 238 |
+
"8": "SUCCESS",
|
| 239 |
"80": "SUCCESS",
|
| 240 |
"81": "SUCCESS",
|
| 241 |
"82": "SUCCESS",
|
|
|
|
| 246 |
"87": "SUCCESS",
|
| 247 |
"88": "SUCCESS",
|
| 248 |
"89": "SUCCESS",
|
| 249 |
+
"9": "SUCCESS",
|
| 250 |
"90": "SUCCESS",
|
| 251 |
+
"91": "SUCCESS",
|
| 252 |
"92": "SUCCESS",
|
| 253 |
+
"93": "SUCCESS",
|
| 254 |
"94": "SUCCESS",
|
| 255 |
"95": "SUCCESS",
|
| 256 |
+
"96": "SUCCESS",
|
| 257 |
+
"97": "SUCCESS",
|
| 258 |
"98": "SUCCESS",
|
| 259 |
"99": "SUCCESS"
|
| 260 |
},
|
|
|
|
| 264 |
"attention_bias": false,
|
| 265 |
"attention_dropout": 0.0,
|
| 266 |
"block_list": [
|
| 267 |
+
5785455,
|
| 268 |
+
5785496
|
| 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": 5769340,
|
| 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
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bac9834369d7e57c9e03646e48f7e4e6b8c42c232b9d6f93906be9a887b41f85
|
| 3 |
+
size 8800564076
|
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 sha256:00ab8fb697f6059bb1fa032c43f6e0361efbac1c4567a932d35d7bf6c6be3fd6
|
| 3 |
size 4400216536
|