Run 8. Outer Step 96. Inner Step 0. Peers 53.
Browse files- config.json +156 -156
- inner_optimizer.pt +2 -2
- model.safetensors +1 -1
- outer_optimizer.pt +2 -2
config.json
CHANGED
|
@@ -1,261 +1,261 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "
|
| 3 |
"all_reduce_scores": {
|
| 4 |
"0": "NOT_ALIVE",
|
| 5 |
"1": "NOT_ALIVE",
|
| 6 |
"10": "NOT_ALIVE",
|
| 7 |
"100": "NOT_ALIVE",
|
| 8 |
-
"101": "
|
| 9 |
"102": "NOT_ALIVE",
|
| 10 |
-
"103": "
|
| 11 |
-
"104": "
|
| 12 |
-
"105": "
|
| 13 |
-
"106": "
|
| 14 |
"107": "NOT_ALIVE",
|
| 15 |
"108": "NOT_ALIVE",
|
| 16 |
-
"109": "
|
| 17 |
"11": "NOT_ALIVE",
|
| 18 |
-
"110": "
|
| 19 |
"111": "NOT_ALIVE",
|
| 20 |
"112": "NOT_ALIVE",
|
| 21 |
-
"113": "
|
| 22 |
-
"114": "
|
| 23 |
"115": "NOT_ALIVE",
|
| 24 |
"116": "SUCCESS",
|
| 25 |
"117": "NOT_ALIVE",
|
| 26 |
-
"118": "
|
| 27 |
-
"119": "
|
| 28 |
"12": "NOT_ALIVE",
|
| 29 |
-
"120": "
|
| 30 |
-
"121": "
|
| 31 |
-
"122": "
|
| 32 |
-
"123": "
|
| 33 |
-
"124": "
|
| 34 |
-
"125": "
|
| 35 |
"126": "SUCCESS",
|
| 36 |
"127": "NOT_ALIVE",
|
| 37 |
-
"128": "
|
| 38 |
-
"129": "
|
| 39 |
"13": "NOT_ALIVE",
|
| 40 |
"130": "SUCCESS",
|
| 41 |
-
"131": "
|
| 42 |
-
"132": "
|
| 43 |
-
"133": "
|
| 44 |
-
"134": "
|
| 45 |
"135": "NON_PARTICIPATING",
|
| 46 |
"136": "NOT_ALIVE",
|
| 47 |
"137": "SUCCESS",
|
| 48 |
"138": "NOT_ALIVE",
|
| 49 |
"139": "NOT_ALIVE",
|
| 50 |
-
"14": "
|
| 51 |
"140": "NOT_ALIVE",
|
| 52 |
-
"141": "
|
| 53 |
-
"142": "
|
| 54 |
"143": "NOT_ALIVE",
|
| 55 |
"144": "NOT_ALIVE",
|
| 56 |
-
"145": "
|
| 57 |
"146": "NOT_ALIVE",
|
| 58 |
-
"147": "
|
| 59 |
-
"148": "
|
| 60 |
"149": "SUCCESS",
|
| 61 |
-
"15": "
|
| 62 |
-
"150": "
|
| 63 |
-
"151": "
|
| 64 |
-
"152": "
|
| 65 |
-
"153": "
|
| 66 |
"154": "NOT_ALIVE",
|
| 67 |
-
"155": "
|
| 68 |
-
"156": "
|
| 69 |
-
"157": "
|
| 70 |
"158": "SUCCESS",
|
| 71 |
-
"159": "
|
| 72 |
"16": "SUCCESS",
|
| 73 |
"160": "SUCCESS",
|
| 74 |
"161": "SUCCESS",
|
| 75 |
"162": "SUCCESS",
|
| 76 |
-
"163": "
|
| 77 |
"164": "NOT_ALIVE",
|
| 78 |
"165": "SUCCESS",
|
| 79 |
"166": "NOT_ALIVE",
|
| 80 |
"167": "NON_PARTICIPATING",
|
| 81 |
-
"168": "
|
| 82 |
-
"169": "
|
| 83 |
"17": "NOT_ALIVE",
|
| 84 |
-
"170": "
|
| 85 |
"171": "NOT_ALIVE",
|
| 86 |
"172": "NOT_ALIVE",
|
| 87 |
-
"173": "
|
| 88 |
-
"174": "
|
| 89 |
"175": "SUCCESS",
|
| 90 |
"176": "NOT_ALIVE",
|
| 91 |
-
"177": "
|
| 92 |
-
"178": "
|
| 93 |
"179": "NOT_ALIVE",
|
| 94 |
"18": "NOT_ALIVE",
|
| 95 |
-
"180": "
|
| 96 |
"181": "NON_PARTICIPATING",
|
| 97 |
-
"182": "
|
| 98 |
-
"183": "
|
| 99 |
-
"184": "
|
| 100 |
-
"185": "
|
| 101 |
-
"186": "
|
| 102 |
-
"187": "
|
| 103 |
-
"188": "
|
| 104 |
-
"189": "
|
| 105 |
"19": "NOT_ALIVE",
|
| 106 |
"190": "NON_PARTICIPATING",
|
| 107 |
-
"191": "
|
| 108 |
-
"192": "
|
| 109 |
-
"193": "
|
| 110 |
-
"194": "
|
| 111 |
-
"195": "
|
| 112 |
"196": "NON_PARTICIPATING",
|
| 113 |
"197": "NOT_ALIVE",
|
| 114 |
-
"198": "
|
| 115 |
-
"199": "
|
| 116 |
"2": "NOT_ALIVE",
|
| 117 |
"20": "NOT_ALIVE",
|
| 118 |
-
"200": "
|
| 119 |
"201": "NOT_ALIVE",
|
| 120 |
-
"202": "
|
| 121 |
-
"203": "
|
| 122 |
-
"204": "
|
| 123 |
-
"205": "
|
| 124 |
-
"206": "
|
| 125 |
"207": "NOT_ALIVE",
|
| 126 |
-
"208": "
|
| 127 |
-
"209": "
|
| 128 |
-
"21": "
|
| 129 |
-
"210": "
|
| 130 |
-
"211": "
|
| 131 |
"212": "NOT_ALIVE",
|
| 132 |
"213": "SUCCESS",
|
| 133 |
-
"214": "
|
| 134 |
-
"215": "
|
| 135 |
"216": "SUCCESS",
|
| 136 |
"217": "SUCCESS",
|
| 137 |
-
"218": "
|
| 138 |
-
"219": "
|
| 139 |
-
"22": "
|
| 140 |
-
"220": "
|
| 141 |
"221": "NON_PARTICIPATING",
|
| 142 |
"222": "NOT_ALIVE",
|
| 143 |
-
"223": "
|
| 144 |
-
"224": "
|
| 145 |
-
"225": "
|
| 146 |
-
"226": "
|
| 147 |
-
"227": "
|
| 148 |
-
"228": "
|
| 149 |
-
"229": "
|
| 150 |
-
"23": "
|
| 151 |
-
"230": "
|
| 152 |
"231": "NOT_ALIVE",
|
| 153 |
-
"232": "
|
| 154 |
-
"233": "
|
| 155 |
-
"234": "
|
| 156 |
"235": "SUCCESS",
|
| 157 |
-
"236": "
|
| 158 |
"237": "SUCCESS",
|
| 159 |
-
"238": "
|
| 160 |
-
"239": "
|
| 161 |
-
"24": "
|
| 162 |
"240": "NOT_ALIVE",
|
| 163 |
"241": "NON_PARTICIPATING",
|
| 164 |
-
"242": "
|
| 165 |
-
"243": "
|
| 166 |
-
"244": "
|
| 167 |
-
"245": "
|
| 168 |
"246": "SUCCESS",
|
| 169 |
"247": "SUCCESS",
|
| 170 |
-
"248": "
|
| 171 |
-
"249": "
|
| 172 |
"25": "SUCCESS",
|
| 173 |
-
"250": "
|
| 174 |
-
"251": "
|
| 175 |
-
"252": "
|
| 176 |
"253": "SUCCESS",
|
| 177 |
"254": "NOT_ALIVE",
|
| 178 |
-
"255": "
|
| 179 |
"26": "NOT_ALIVE",
|
| 180 |
-
"27": "
|
| 181 |
-
"28": "
|
| 182 |
-
"29": "
|
| 183 |
-
"3": "
|
| 184 |
"30": "NOT_ALIVE",
|
| 185 |
-
"31": "
|
| 186 |
"32": "NOT_ALIVE",
|
| 187 |
-
"33": "
|
| 188 |
-
"34": "
|
| 189 |
"35": "SUCCESS",
|
| 190 |
"36": "SUCCESS",
|
| 191 |
"37": "NOT_ALIVE",
|
| 192 |
-
"38": "
|
| 193 |
-
"39": "
|
| 194 |
-
"4": "
|
| 195 |
-
"40": "
|
| 196 |
"41": "SUCCESS",
|
| 197 |
"42": "NOT_ALIVE",
|
| 198 |
-
"43": "
|
| 199 |
-
"44": "
|
| 200 |
-
"45": "
|
| 201 |
-
"46": "
|
| 202 |
"47": "NON_PARTICIPATING",
|
| 203 |
-
"48": "
|
| 204 |
-
"49": "
|
| 205 |
"5": "NOT_ALIVE",
|
| 206 |
"50": "NOT_ALIVE",
|
| 207 |
-
"51": "
|
| 208 |
-
"52": "
|
| 209 |
-
"53": "
|
| 210 |
-
"54": "
|
| 211 |
"55": "NOT_ALIVE",
|
| 212 |
-
"56": "
|
| 213 |
"57": "NOT_ALIVE",
|
| 214 |
"58": "NOT_ALIVE",
|
| 215 |
"59": "NOT_ALIVE",
|
| 216 |
"6": "NOT_ALIVE",
|
| 217 |
"60": "NON_PARTICIPATING",
|
| 218 |
-
"61": "
|
| 219 |
-
"62": "
|
| 220 |
-
"63": "
|
| 221 |
"64": "NOT_ALIVE",
|
| 222 |
-
"65": "
|
| 223 |
-
"66": "
|
| 224 |
-
"67": "
|
| 225 |
-
"68": "
|
| 226 |
-
"69": "
|
| 227 |
"7": "NOT_ALIVE",
|
| 228 |
-
"70": "
|
| 229 |
"71": "NOT_ALIVE",
|
| 230 |
-
"72": "
|
| 231 |
-
"73": "
|
| 232 |
-
"74": "
|
| 233 |
"75": "NOT_ALIVE",
|
| 234 |
"76": "NON_PARTICIPATING",
|
| 235 |
-
"77": "
|
| 236 |
-
"78": "
|
| 237 |
"79": "NOT_ALIVE",
|
| 238 |
"8": "NOT_ALIVE",
|
| 239 |
"80": "NOT_ALIVE",
|
| 240 |
"81": "NOT_ALIVE",
|
| 241 |
"82": "NOT_ALIVE",
|
| 242 |
"83": "NOT_ALIVE",
|
| 243 |
-
"84": "
|
| 244 |
"85": "NON_PARTICIPATING",
|
| 245 |
-
"86": "
|
| 246 |
-
"87": "
|
| 247 |
-
"88": "
|
| 248 |
-
"89": "
|
| 249 |
-
"9": "
|
| 250 |
-
"90": "
|
| 251 |
"91": "NOT_ALIVE",
|
| 252 |
"92": "NOT_ALIVE",
|
| 253 |
-
"93": "
|
| 254 |
-
"94": "
|
| 255 |
-
"95": "
|
| 256 |
-
"96": "
|
| 257 |
-
"97": "
|
| 258 |
-
"98": "
|
| 259 |
"99": "NON_PARTICIPATING"
|
| 260 |
},
|
| 261 |
"architectures": [
|
|
@@ -271,7 +271,7 @@
|
|
| 271 |
"initializer_range": 0.02,
|
| 272 |
"inner_step": 0,
|
| 273 |
"intermediate_size": 5632,
|
| 274 |
-
"last_allreduce_block":
|
| 275 |
"max_position_embeddings": 2048,
|
| 276 |
"mlp_bias": false,
|
| 277 |
"model_type": "llama",
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "g0001/d_p5",
|
| 3 |
"all_reduce_scores": {
|
| 4 |
"0": "NOT_ALIVE",
|
| 5 |
"1": "NOT_ALIVE",
|
| 6 |
"10": "NOT_ALIVE",
|
| 7 |
"100": "NOT_ALIVE",
|
| 8 |
+
"101": "NOT_ALIVE",
|
| 9 |
"102": "NOT_ALIVE",
|
| 10 |
+
"103": "NON_PARTICIPATING",
|
| 11 |
+
"104": "NON_PARTICIPATING",
|
| 12 |
+
"105": "NON_PARTICIPATING",
|
| 13 |
+
"106": "NON_PARTICIPATING",
|
| 14 |
"107": "NOT_ALIVE",
|
| 15 |
"108": "NOT_ALIVE",
|
| 16 |
+
"109": "FAIL",
|
| 17 |
"11": "NOT_ALIVE",
|
| 18 |
+
"110": "NON_PARTICIPATING",
|
| 19 |
"111": "NOT_ALIVE",
|
| 20 |
"112": "NOT_ALIVE",
|
| 21 |
+
"113": "SUCCESS",
|
| 22 |
+
"114": "NOT_ALIVE",
|
| 23 |
"115": "NOT_ALIVE",
|
| 24 |
"116": "SUCCESS",
|
| 25 |
"117": "NOT_ALIVE",
|
| 26 |
+
"118": "NOT_ALIVE",
|
| 27 |
+
"119": "NOT_ALIVE",
|
| 28 |
"12": "NOT_ALIVE",
|
| 29 |
+
"120": "FAIL",
|
| 30 |
+
"121": "NON_PARTICIPATING",
|
| 31 |
+
"122": "NOT_ALIVE",
|
| 32 |
+
"123": "NON_PARTICIPATING",
|
| 33 |
+
"124": "FAIL",
|
| 34 |
+
"125": "NON_PARTICIPATING",
|
| 35 |
"126": "SUCCESS",
|
| 36 |
"127": "NOT_ALIVE",
|
| 37 |
+
"128": "NOT_ALIVE",
|
| 38 |
+
"129": "NOT_ALIVE",
|
| 39 |
"13": "NOT_ALIVE",
|
| 40 |
"130": "SUCCESS",
|
| 41 |
+
"131": "NOT_ALIVE",
|
| 42 |
+
"132": "NON_PARTICIPATING",
|
| 43 |
+
"133": "NON_PARTICIPATING",
|
| 44 |
+
"134": "NON_PARTICIPATING",
|
| 45 |
"135": "NON_PARTICIPATING",
|
| 46 |
"136": "NOT_ALIVE",
|
| 47 |
"137": "SUCCESS",
|
| 48 |
"138": "NOT_ALIVE",
|
| 49 |
"139": "NOT_ALIVE",
|
| 50 |
+
"14": "SUCCESS",
|
| 51 |
"140": "NOT_ALIVE",
|
| 52 |
+
"141": "NOT_ALIVE",
|
| 53 |
+
"142": "SUCCESS",
|
| 54 |
"143": "NOT_ALIVE",
|
| 55 |
"144": "NOT_ALIVE",
|
| 56 |
+
"145": "NOT_ALIVE",
|
| 57 |
"146": "NOT_ALIVE",
|
| 58 |
+
"147": "SUCCESS",
|
| 59 |
+
"148": "SUCCESS",
|
| 60 |
"149": "SUCCESS",
|
| 61 |
+
"15": "NON_PARTICIPATING",
|
| 62 |
+
"150": "NOT_ALIVE",
|
| 63 |
+
"151": "NOT_ALIVE",
|
| 64 |
+
"152": "NON_PARTICIPATING",
|
| 65 |
+
"153": "NON_PARTICIPATING",
|
| 66 |
"154": "NOT_ALIVE",
|
| 67 |
+
"155": "SUCCESS",
|
| 68 |
+
"156": "NON_PARTICIPATING",
|
| 69 |
+
"157": "NOT_ALIVE",
|
| 70 |
"158": "SUCCESS",
|
| 71 |
+
"159": "SUCCESS",
|
| 72 |
"16": "SUCCESS",
|
| 73 |
"160": "SUCCESS",
|
| 74 |
"161": "SUCCESS",
|
| 75 |
"162": "SUCCESS",
|
| 76 |
+
"163": "SUCCESS",
|
| 77 |
"164": "NOT_ALIVE",
|
| 78 |
"165": "SUCCESS",
|
| 79 |
"166": "NOT_ALIVE",
|
| 80 |
"167": "NON_PARTICIPATING",
|
| 81 |
+
"168": "NON_PARTICIPATING",
|
| 82 |
+
"169": "NON_PARTICIPATING",
|
| 83 |
"17": "NOT_ALIVE",
|
| 84 |
+
"170": "FAIL",
|
| 85 |
"171": "NOT_ALIVE",
|
| 86 |
"172": "NOT_ALIVE",
|
| 87 |
+
"173": "NON_PARTICIPATING",
|
| 88 |
+
"174": "NOT_ALIVE",
|
| 89 |
"175": "SUCCESS",
|
| 90 |
"176": "NOT_ALIVE",
|
| 91 |
+
"177": "NON_PARTICIPATING",
|
| 92 |
+
"178": "NOT_ALIVE",
|
| 93 |
"179": "NOT_ALIVE",
|
| 94 |
"18": "NOT_ALIVE",
|
| 95 |
+
"180": "NOT_ALIVE",
|
| 96 |
"181": "NON_PARTICIPATING",
|
| 97 |
+
"182": "SUCCESS",
|
| 98 |
+
"183": "NOT_ALIVE",
|
| 99 |
+
"184": "NON_PARTICIPATING",
|
| 100 |
+
"185": "NON_PARTICIPATING",
|
| 101 |
+
"186": "NOT_ALIVE",
|
| 102 |
+
"187": "NOT_ALIVE",
|
| 103 |
+
"188": "NON_PARTICIPATING",
|
| 104 |
+
"189": "SUCCESS",
|
| 105 |
"19": "NOT_ALIVE",
|
| 106 |
"190": "NON_PARTICIPATING",
|
| 107 |
+
"191": "NON_PARTICIPATING",
|
| 108 |
+
"192": "NOT_ALIVE",
|
| 109 |
+
"193": "NOT_ALIVE",
|
| 110 |
+
"194": "NOT_ALIVE",
|
| 111 |
+
"195": "NOT_ALIVE",
|
| 112 |
"196": "NON_PARTICIPATING",
|
| 113 |
"197": "NOT_ALIVE",
|
| 114 |
+
"198": "NON_PARTICIPATING",
|
| 115 |
+
"199": "NON_PARTICIPATING",
|
| 116 |
"2": "NOT_ALIVE",
|
| 117 |
"20": "NOT_ALIVE",
|
| 118 |
+
"200": "NON_PARTICIPATING",
|
| 119 |
"201": "NOT_ALIVE",
|
| 120 |
+
"202": "NOT_ALIVE",
|
| 121 |
+
"203": "NON_PARTICIPATING",
|
| 122 |
+
"204": "FAIL",
|
| 123 |
+
"205": "SUCCESS",
|
| 124 |
+
"206": "NOT_ALIVE",
|
| 125 |
"207": "NOT_ALIVE",
|
| 126 |
+
"208": "SUCCESS",
|
| 127 |
+
"209": "NOT_ALIVE",
|
| 128 |
+
"21": "NON_PARTICIPATING",
|
| 129 |
+
"210": "SUCCESS",
|
| 130 |
+
"211": "NON_PARTICIPATING",
|
| 131 |
"212": "NOT_ALIVE",
|
| 132 |
"213": "SUCCESS",
|
| 133 |
+
"214": "NOT_ALIVE",
|
| 134 |
+
"215": "NOT_ALIVE",
|
| 135 |
"216": "SUCCESS",
|
| 136 |
"217": "SUCCESS",
|
| 137 |
+
"218": "NON_PARTICIPATING",
|
| 138 |
+
"219": "NON_PARTICIPATING",
|
| 139 |
+
"22": "NON_PARTICIPATING",
|
| 140 |
+
"220": "SUCCESS",
|
| 141 |
"221": "NON_PARTICIPATING",
|
| 142 |
"222": "NOT_ALIVE",
|
| 143 |
+
"223": "NOT_ALIVE",
|
| 144 |
+
"224": "NOT_ALIVE",
|
| 145 |
+
"225": "NON_PARTICIPATING",
|
| 146 |
+
"226": "NON_PARTICIPATING",
|
| 147 |
+
"227": "NON_PARTICIPATING",
|
| 148 |
+
"228": "NOT_ALIVE",
|
| 149 |
+
"229": "NOT_ALIVE",
|
| 150 |
+
"23": "SUCCESS",
|
| 151 |
+
"230": "SUCCESS",
|
| 152 |
"231": "NOT_ALIVE",
|
| 153 |
+
"232": "SUCCESS",
|
| 154 |
+
"233": "NOT_ALIVE",
|
| 155 |
+
"234": "SUCCESS",
|
| 156 |
"235": "SUCCESS",
|
| 157 |
+
"236": "NON_PARTICIPATING",
|
| 158 |
"237": "SUCCESS",
|
| 159 |
+
"238": "NON_PARTICIPATING",
|
| 160 |
+
"239": "NON_PARTICIPATING",
|
| 161 |
+
"24": "SUCCESS",
|
| 162 |
"240": "NOT_ALIVE",
|
| 163 |
"241": "NON_PARTICIPATING",
|
| 164 |
+
"242": "NOT_ALIVE",
|
| 165 |
+
"243": "NOT_ALIVE",
|
| 166 |
+
"244": "NOT_ALIVE",
|
| 167 |
+
"245": "NON_PARTICIPATING",
|
| 168 |
"246": "SUCCESS",
|
| 169 |
"247": "SUCCESS",
|
| 170 |
+
"248": "NON_PARTICIPATING",
|
| 171 |
+
"249": "NON_PARTICIPATING",
|
| 172 |
"25": "SUCCESS",
|
| 173 |
+
"250": "FAIL",
|
| 174 |
+
"251": "NON_PARTICIPATING",
|
| 175 |
+
"252": "NOT_ALIVE",
|
| 176 |
"253": "SUCCESS",
|
| 177 |
"254": "NOT_ALIVE",
|
| 178 |
+
"255": "NON_PARTICIPATING",
|
| 179 |
"26": "NOT_ALIVE",
|
| 180 |
+
"27": "NON_PARTICIPATING",
|
| 181 |
+
"28": "NOT_ALIVE",
|
| 182 |
+
"29": "NON_PARTICIPATING",
|
| 183 |
+
"3": "NON_PARTICIPATING",
|
| 184 |
"30": "NOT_ALIVE",
|
| 185 |
+
"31": "SUCCESS",
|
| 186 |
"32": "NOT_ALIVE",
|
| 187 |
+
"33": "NOT_ALIVE",
|
| 188 |
+
"34": "NOT_ALIVE",
|
| 189 |
"35": "SUCCESS",
|
| 190 |
"36": "SUCCESS",
|
| 191 |
"37": "NOT_ALIVE",
|
| 192 |
+
"38": "NON_PARTICIPATING",
|
| 193 |
+
"39": "NOT_ALIVE",
|
| 194 |
+
"4": "NON_PARTICIPATING",
|
| 195 |
+
"40": "NOT_ALIVE",
|
| 196 |
"41": "SUCCESS",
|
| 197 |
"42": "NOT_ALIVE",
|
| 198 |
+
"43": "FAIL",
|
| 199 |
+
"44": "NON_PARTICIPATING",
|
| 200 |
+
"45": "NON_PARTICIPATING",
|
| 201 |
+
"46": "NOT_ALIVE",
|
| 202 |
"47": "NON_PARTICIPATING",
|
| 203 |
+
"48": "NON_PARTICIPATING",
|
| 204 |
+
"49": "NOT_ALIVE",
|
| 205 |
"5": "NOT_ALIVE",
|
| 206 |
"50": "NOT_ALIVE",
|
| 207 |
+
"51": "NON_PARTICIPATING",
|
| 208 |
+
"52": "NOT_ALIVE",
|
| 209 |
+
"53": "SUCCESS",
|
| 210 |
+
"54": "NOT_ALIVE",
|
| 211 |
"55": "NOT_ALIVE",
|
| 212 |
+
"56": "NOT_ALIVE",
|
| 213 |
"57": "NOT_ALIVE",
|
| 214 |
"58": "NOT_ALIVE",
|
| 215 |
"59": "NOT_ALIVE",
|
| 216 |
"6": "NOT_ALIVE",
|
| 217 |
"60": "NON_PARTICIPATING",
|
| 218 |
+
"61": "NON_PARTICIPATING",
|
| 219 |
+
"62": "SUCCESS",
|
| 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": "SUCCESS",
|
| 227 |
"7": "NOT_ALIVE",
|
| 228 |
+
"70": "FAIL",
|
| 229 |
"71": "NOT_ALIVE",
|
| 230 |
+
"72": "NOT_ALIVE",
|
| 231 |
+
"73": "NOT_ALIVE",
|
| 232 |
+
"74": "FAIL",
|
| 233 |
"75": "NOT_ALIVE",
|
| 234 |
"76": "NON_PARTICIPATING",
|
| 235 |
+
"77": "NOT_ALIVE",
|
| 236 |
+
"78": "NOT_ALIVE",
|
| 237 |
"79": "NOT_ALIVE",
|
| 238 |
"8": "NOT_ALIVE",
|
| 239 |
"80": "NOT_ALIVE",
|
| 240 |
"81": "NOT_ALIVE",
|
| 241 |
"82": "NOT_ALIVE",
|
| 242 |
"83": "NOT_ALIVE",
|
| 243 |
+
"84": "NON_PARTICIPATING",
|
| 244 |
"85": "NON_PARTICIPATING",
|
| 245 |
+
"86": "NOT_ALIVE",
|
| 246 |
+
"87": "NOT_ALIVE",
|
| 247 |
+
"88": "SUCCESS",
|
| 248 |
+
"89": "NON_PARTICIPATING",
|
| 249 |
+
"9": "NON_PARTICIPATING",
|
| 250 |
+
"90": "NOT_ALIVE",
|
| 251 |
"91": "NOT_ALIVE",
|
| 252 |
"92": "NOT_ALIVE",
|
| 253 |
+
"93": "SUCCESS",
|
| 254 |
+
"94": "SUCCESS",
|
| 255 |
+
"95": "NOT_ALIVE",
|
| 256 |
+
"96": "SUCCESS",
|
| 257 |
+
"97": "SUCCESS",
|
| 258 |
+
"98": "SUCCESS",
|
| 259 |
"99": "NON_PARTICIPATING"
|
| 260 |
},
|
| 261 |
"architectures": [
|
|
|
|
| 271 |
"initializer_range": 0.02,
|
| 272 |
"inner_step": 0,
|
| 273 |
"intermediate_size": 5632,
|
| 274 |
+
"last_allreduce_block": 6419009,
|
| 275 |
"max_position_embeddings": 2048,
|
| 276 |
"mlp_bias": false,
|
| 277 |
"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 sha256:5e92abf9804f677da5857f3a047d4745cf045834046b67d8ca13f50592c30776
|
| 3 |
+
size 4924926906
|
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:e1117a703a73dc64bc59f88c467bed7bd4f1b4ffd3556aabe94c324a7235490d
|
| 3 |
size 4400216536
|
outer_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:ce7db79e402e51fa879f231e8e074e86d5c67a0855c1ea33f5b1f15c2a2e4860
|
| 3 |
+
size 4400252568
|