Run 8. Outer Step 95. Inner Step 0. Peers 62.
Browse files- config.json +86 -86
- inner_optimizer.pt +1 -1
- model.safetensors +1 -1
- outer_optimizer.pt +1 -1
config.json
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 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": "NON_PARTICIPATING",
|
| 9 |
-
"102": "
|
| 10 |
"103": "NOT_ALIVE",
|
| 11 |
"104": "NOT_ALIVE",
|
| 12 |
"105": "NOT_ALIVE",
|
|
@@ -17,68 +17,68 @@
|
|
| 17 |
"11": "NOT_ALIVE",
|
| 18 |
"110": "SUCCESS",
|
| 19 |
"111": "NOT_ALIVE",
|
| 20 |
-
"112": "
|
| 21 |
"113": "NON_PARTICIPATING",
|
| 22 |
"114": "NON_PARTICIPATING",
|
| 23 |
"115": "NOT_ALIVE",
|
| 24 |
-
"116": "
|
| 25 |
"117": "NOT_ALIVE",
|
| 26 |
-
"118": "
|
| 27 |
"119": "NON_PARTICIPATING",
|
| 28 |
"12": "NOT_ALIVE",
|
| 29 |
-
"120": "
|
| 30 |
-
"121": "
|
| 31 |
"122": "NON_PARTICIPATING",
|
| 32 |
-
"123": "
|
| 33 |
"124": "SUCCESS",
|
| 34 |
"125": "NOT_ALIVE",
|
| 35 |
"126": "SUCCESS",
|
| 36 |
"127": "NOT_ALIVE",
|
| 37 |
"128": "NON_PARTICIPATING",
|
| 38 |
-
"129": "
|
| 39 |
"13": "NOT_ALIVE",
|
| 40 |
-
"130": "
|
| 41 |
"131": "NON_PARTICIPATING",
|
| 42 |
"132": "SUCCESS",
|
| 43 |
-
"133": "
|
| 44 |
-
"134": "
|
| 45 |
-
"135": "
|
| 46 |
-
"136": "
|
| 47 |
-
"137": "
|
| 48 |
"138": "NOT_ALIVE",
|
| 49 |
-
"139": "
|
| 50 |
"14": "NOT_ALIVE",
|
| 51 |
-
"140": "
|
| 52 |
"141": "NON_PARTICIPATING",
|
| 53 |
-
"142": "
|
| 54 |
"143": "NOT_ALIVE",
|
| 55 |
"144": "NOT_ALIVE",
|
| 56 |
-
"145": "
|
| 57 |
"146": "NOT_ALIVE",
|
| 58 |
"147": "NON_PARTICIPATING",
|
| 59 |
-
"148": "
|
| 60 |
"149": "SUCCESS",
|
| 61 |
"15": "NOT_ALIVE",
|
| 62 |
-
"150": "
|
| 63 |
"151": "NON_PARTICIPATING",
|
| 64 |
"152": "NOT_ALIVE",
|
| 65 |
"153": "NOT_ALIVE",
|
| 66 |
-
"154": "
|
| 67 |
-
"155": "
|
| 68 |
-
"156": "
|
| 69 |
"157": "NON_PARTICIPATING",
|
| 70 |
-
"158": "
|
| 71 |
-
"159": "
|
| 72 |
-
"16": "
|
| 73 |
"160": "SUCCESS",
|
| 74 |
"161": "SUCCESS",
|
| 75 |
-
"162": "
|
| 76 |
"163": "NOT_ALIVE",
|
| 77 |
-
"164": "
|
| 78 |
-
"165": "
|
| 79 |
"166": "NOT_ALIVE",
|
| 80 |
-
"167": "
|
| 81 |
-
"168": "
|
| 82 |
"169": "NOT_ALIVE",
|
| 83 |
"17": "NOT_ALIVE",
|
| 84 |
"170": "SUCCESS",
|
|
@@ -90,28 +90,28 @@
|
|
| 90 |
"176": "NOT_ALIVE",
|
| 91 |
"177": "SUCCESS",
|
| 92 |
"178": "NON_PARTICIPATING",
|
| 93 |
-
"179": "
|
| 94 |
"18": "NOT_ALIVE",
|
| 95 |
"180": "NON_PARTICIPATING",
|
| 96 |
"181": "NON_PARTICIPATING",
|
| 97 |
"182": "NOT_ALIVE",
|
| 98 |
-
"183": "
|
| 99 |
-
"184": "
|
| 100 |
"185": "SUCCESS",
|
| 101 |
"186": "NON_PARTICIPATING",
|
| 102 |
"187": "NON_PARTICIPATING",
|
| 103 |
"188": "SUCCESS",
|
| 104 |
-
"189": "
|
| 105 |
"19": "NOT_ALIVE",
|
| 106 |
-
"190": "
|
| 107 |
-
"191": "
|
| 108 |
"192": "NON_PARTICIPATING",
|
| 109 |
-
"193": "
|
| 110 |
-
"194": "
|
| 111 |
"195": "NON_PARTICIPATING",
|
| 112 |
-
"196": "
|
| 113 |
"197": "NOT_ALIVE",
|
| 114 |
-
"198": "
|
| 115 |
"199": "SUCCESS",
|
| 116 |
"2": "NOT_ALIVE",
|
| 117 |
"20": "NOT_ALIVE",
|
|
@@ -123,49 +123,49 @@
|
|
| 123 |
"205": "NOT_ALIVE",
|
| 124 |
"206": "NON_PARTICIPATING",
|
| 125 |
"207": "NOT_ALIVE",
|
| 126 |
-
"208": "
|
| 127 |
-
"209": "
|
| 128 |
"21": "NOT_ALIVE",
|
| 129 |
"210": "NON_PARTICIPATING",
|
| 130 |
"211": "SUCCESS",
|
| 131 |
"212": "NOT_ALIVE",
|
| 132 |
"213": "SUCCESS",
|
| 133 |
-
"214": "
|
| 134 |
"215": "NON_PARTICIPATING",
|
| 135 |
"216": "SUCCESS",
|
| 136 |
"217": "SUCCESS",
|
| 137 |
-
"218": "
|
| 138 |
-
"219": "
|
| 139 |
"22": "SUCCESS",
|
| 140 |
-
"220": "
|
| 141 |
-
"221": "
|
| 142 |
"222": "NOT_ALIVE",
|
| 143 |
-
"223": "
|
| 144 |
"224": "NON_PARTICIPATING",
|
| 145 |
"225": "SUCCESS",
|
| 146 |
-
"226": "
|
| 147 |
-
"227": "
|
| 148 |
-
"228": "
|
| 149 |
"229": "NON_PARTICIPATING",
|
| 150 |
"23": "NOT_ALIVE",
|
| 151 |
"230": "NOT_ALIVE",
|
| 152 |
-
"231": "
|
| 153 |
"232": "NOT_ALIVE",
|
| 154 |
-
"233": "
|
| 155 |
-
"234": "
|
| 156 |
-
"235": "
|
| 157 |
-
"236": "
|
| 158 |
-
"237": "
|
| 159 |
"238": "SUCCESS",
|
| 160 |
"239": "NOT_ALIVE",
|
| 161 |
-
"24": "
|
| 162 |
"240": "NOT_ALIVE",
|
| 163 |
"241": "NON_PARTICIPATING",
|
| 164 |
"242": "NON_PARTICIPATING",
|
| 165 |
"243": "NON_PARTICIPATING",
|
| 166 |
-
"244": "
|
| 167 |
"245": "SUCCESS",
|
| 168 |
-
"246": "
|
| 169 |
"247": "SUCCESS",
|
| 170 |
"248": "SUCCESS",
|
| 171 |
"249": "NOT_ALIVE",
|
|
@@ -174,27 +174,27 @@
|
|
| 174 |
"251": "SUCCESS",
|
| 175 |
"252": "NON_PARTICIPATING",
|
| 176 |
"253": "SUCCESS",
|
| 177 |
-
"254": "
|
| 178 |
"255": "SUCCESS",
|
| 179 |
"26": "NOT_ALIVE",
|
| 180 |
"27": "SUCCESS",
|
| 181 |
"28": "NON_PARTICIPATING",
|
| 182 |
-
"29": "
|
| 183 |
"3": "NOT_ALIVE",
|
| 184 |
"30": "NOT_ALIVE",
|
| 185 |
-
"31": "
|
| 186 |
"32": "NOT_ALIVE",
|
| 187 |
"33": "NON_PARTICIPATING",
|
| 188 |
"34": "NON_PARTICIPATING",
|
| 189 |
-
"35": "
|
| 190 |
-
"36": "
|
| 191 |
"37": "NOT_ALIVE",
|
| 192 |
"38": "NOT_ALIVE",
|
| 193 |
-
"39": "
|
| 194 |
"4": "NOT_ALIVE",
|
| 195 |
"40": "NON_PARTICIPATING",
|
| 196 |
-
"41": "
|
| 197 |
-
"42": "
|
| 198 |
"43": "SUCCESS",
|
| 199 |
"44": "SUCCESS",
|
| 200 |
"45": "NOT_ALIVE",
|
|
@@ -204,40 +204,40 @@
|
|
| 204 |
"49": "NON_PARTICIPATING",
|
| 205 |
"5": "NOT_ALIVE",
|
| 206 |
"50": "NOT_ALIVE",
|
| 207 |
-
"51": "
|
| 208 |
"52": "NON_PARTICIPATING",
|
| 209 |
"53": "NOT_ALIVE",
|
| 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": "
|
| 218 |
"61": "SUCCESS",
|
| 219 |
-
"62": "
|
| 220 |
"63": "NON_PARTICIPATING",
|
| 221 |
"64": "NOT_ALIVE",
|
| 222 |
"65": "NOT_ALIVE",
|
| 223 |
-
"66": "
|
| 224 |
"67": "NON_PARTICIPATING",
|
| 225 |
"68": "NON_PARTICIPATING",
|
| 226 |
"69": "NON_PARTICIPATING",
|
| 227 |
"7": "NOT_ALIVE",
|
| 228 |
"70": "NOT_ALIVE",
|
| 229 |
"71": "NOT_ALIVE",
|
| 230 |
-
"72": "
|
| 231 |
"73": "NON_PARTICIPATING",
|
| 232 |
-
"74": "
|
| 233 |
"75": "NOT_ALIVE",
|
| 234 |
-
"76": "
|
| 235 |
-
"77": "
|
| 236 |
-
"78": "
|
| 237 |
-
"79": "
|
| 238 |
"8": "NOT_ALIVE",
|
| 239 |
"80": "NOT_ALIVE",
|
| 240 |
-
"81": "
|
| 241 |
"82": "NOT_ALIVE",
|
| 242 |
"83": "NOT_ALIVE",
|
| 243 |
"84": "NOT_ALIVE",
|
|
@@ -252,8 +252,8 @@
|
|
| 252 |
"92": "NOT_ALIVE",
|
| 253 |
"93": "NOT_ALIVE",
|
| 254 |
"94": "NON_PARTICIPATING",
|
| 255 |
-
"95": "
|
| 256 |
-
"96": "
|
| 257 |
"97": "NOT_ALIVE",
|
| 258 |
"98": "NOT_ALIVE",
|
| 259 |
"99": "NON_PARTICIPATING"
|
|
@@ -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": "joundi4/M04",
|
| 3 |
"all_reduce_scores": {
|
| 4 |
"0": "NOT_ALIVE",
|
| 5 |
"1": "NOT_ALIVE",
|
| 6 |
"10": "NOT_ALIVE",
|
| 7 |
"100": "NOT_ALIVE",
|
| 8 |
"101": "NON_PARTICIPATING",
|
| 9 |
+
"102": "NOT_ALIVE",
|
| 10 |
"103": "NOT_ALIVE",
|
| 11 |
"104": "NOT_ALIVE",
|
| 12 |
"105": "NOT_ALIVE",
|
|
|
|
| 17 |
"11": "NOT_ALIVE",
|
| 18 |
"110": "SUCCESS",
|
| 19 |
"111": "NOT_ALIVE",
|
| 20 |
+
"112": "NOT_ALIVE",
|
| 21 |
"113": "NON_PARTICIPATING",
|
| 22 |
"114": "NON_PARTICIPATING",
|
| 23 |
"115": "NOT_ALIVE",
|
| 24 |
+
"116": "SUCCESS",
|
| 25 |
"117": "NOT_ALIVE",
|
| 26 |
+
"118": "NON_PARTICIPATING",
|
| 27 |
"119": "NON_PARTICIPATING",
|
| 28 |
"12": "NOT_ALIVE",
|
| 29 |
+
"120": "SUCCESS",
|
| 30 |
+
"121": "NOT_ALIVE",
|
| 31 |
"122": "NON_PARTICIPATING",
|
| 32 |
+
"123": "SUCCESS",
|
| 33 |
"124": "SUCCESS",
|
| 34 |
"125": "NOT_ALIVE",
|
| 35 |
"126": "SUCCESS",
|
| 36 |
"127": "NOT_ALIVE",
|
| 37 |
"128": "NON_PARTICIPATING",
|
| 38 |
+
"129": "NON_PARTICIPATING",
|
| 39 |
"13": "NOT_ALIVE",
|
| 40 |
+
"130": "SUCCESS",
|
| 41 |
"131": "NON_PARTICIPATING",
|
| 42 |
"132": "SUCCESS",
|
| 43 |
+
"133": "NOT_ALIVE",
|
| 44 |
+
"134": "SUCCESS",
|
| 45 |
+
"135": "NON_PARTICIPATING",
|
| 46 |
+
"136": "NOT_ALIVE",
|
| 47 |
+
"137": "SUCCESS",
|
| 48 |
"138": "NOT_ALIVE",
|
| 49 |
+
"139": "NOT_ALIVE",
|
| 50 |
"14": "NOT_ALIVE",
|
| 51 |
+
"140": "NOT_ALIVE",
|
| 52 |
"141": "NON_PARTICIPATING",
|
| 53 |
+
"142": "FAIL",
|
| 54 |
"143": "NOT_ALIVE",
|
| 55 |
"144": "NOT_ALIVE",
|
| 56 |
+
"145": "NON_PARTICIPATING",
|
| 57 |
"146": "NOT_ALIVE",
|
| 58 |
"147": "NON_PARTICIPATING",
|
| 59 |
+
"148": "NON_PARTICIPATING",
|
| 60 |
"149": "SUCCESS",
|
| 61 |
"15": "NOT_ALIVE",
|
| 62 |
+
"150": "NON_PARTICIPATING",
|
| 63 |
"151": "NON_PARTICIPATING",
|
| 64 |
"152": "NOT_ALIVE",
|
| 65 |
"153": "NOT_ALIVE",
|
| 66 |
+
"154": "NOT_ALIVE",
|
| 67 |
+
"155": "NON_PARTICIPATING",
|
| 68 |
+
"156": "NOT_ALIVE",
|
| 69 |
"157": "NON_PARTICIPATING",
|
| 70 |
+
"158": "SUCCESS",
|
| 71 |
+
"159": "NON_PARTICIPATING",
|
| 72 |
+
"16": "SUCCESS",
|
| 73 |
"160": "SUCCESS",
|
| 74 |
"161": "SUCCESS",
|
| 75 |
+
"162": "SUCCESS",
|
| 76 |
"163": "NOT_ALIVE",
|
| 77 |
+
"164": "NOT_ALIVE",
|
| 78 |
+
"165": "SUCCESS",
|
| 79 |
"166": "NOT_ALIVE",
|
| 80 |
+
"167": "NON_PARTICIPATING",
|
| 81 |
+
"168": "NOT_ALIVE",
|
| 82 |
"169": "NOT_ALIVE",
|
| 83 |
"17": "NOT_ALIVE",
|
| 84 |
"170": "SUCCESS",
|
|
|
|
| 90 |
"176": "NOT_ALIVE",
|
| 91 |
"177": "SUCCESS",
|
| 92 |
"178": "NON_PARTICIPATING",
|
| 93 |
+
"179": "NOT_ALIVE",
|
| 94 |
"18": "NOT_ALIVE",
|
| 95 |
"180": "NON_PARTICIPATING",
|
| 96 |
"181": "NON_PARTICIPATING",
|
| 97 |
"182": "NOT_ALIVE",
|
| 98 |
+
"183": "NON_PARTICIPATING",
|
| 99 |
+
"184": "SUCCESS",
|
| 100 |
"185": "SUCCESS",
|
| 101 |
"186": "NON_PARTICIPATING",
|
| 102 |
"187": "NON_PARTICIPATING",
|
| 103 |
"188": "SUCCESS",
|
| 104 |
+
"189": "FAIL",
|
| 105 |
"19": "NOT_ALIVE",
|
| 106 |
+
"190": "NON_PARTICIPATING",
|
| 107 |
+
"191": "NOT_ALIVE",
|
| 108 |
"192": "NON_PARTICIPATING",
|
| 109 |
+
"193": "SUCCESS",
|
| 110 |
+
"194": "NON_PARTICIPATING",
|
| 111 |
"195": "NON_PARTICIPATING",
|
| 112 |
+
"196": "NON_PARTICIPATING",
|
| 113 |
"197": "NOT_ALIVE",
|
| 114 |
+
"198": "SUCCESS",
|
| 115 |
"199": "SUCCESS",
|
| 116 |
"2": "NOT_ALIVE",
|
| 117 |
"20": "NOT_ALIVE",
|
|
|
|
| 123 |
"205": "NOT_ALIVE",
|
| 124 |
"206": "NON_PARTICIPATING",
|
| 125 |
"207": "NOT_ALIVE",
|
| 126 |
+
"208": "NOT_ALIVE",
|
| 127 |
+
"209": "NON_PARTICIPATING",
|
| 128 |
"21": "NOT_ALIVE",
|
| 129 |
"210": "NON_PARTICIPATING",
|
| 130 |
"211": "SUCCESS",
|
| 131 |
"212": "NOT_ALIVE",
|
| 132 |
"213": "SUCCESS",
|
| 133 |
+
"214": "NON_PARTICIPATING",
|
| 134 |
"215": "NON_PARTICIPATING",
|
| 135 |
"216": "SUCCESS",
|
| 136 |
"217": "SUCCESS",
|
| 137 |
+
"218": "FAIL",
|
| 138 |
+
"219": "SUCCESS",
|
| 139 |
"22": "SUCCESS",
|
| 140 |
+
"220": "NOT_ALIVE",
|
| 141 |
+
"221": "NON_PARTICIPATING",
|
| 142 |
"222": "NOT_ALIVE",
|
| 143 |
+
"223": "NON_PARTICIPATING",
|
| 144 |
"224": "NON_PARTICIPATING",
|
| 145 |
"225": "SUCCESS",
|
| 146 |
+
"226": "SUCCESS",
|
| 147 |
+
"227": "NOT_ALIVE",
|
| 148 |
+
"228": "SUCCESS",
|
| 149 |
"229": "NON_PARTICIPATING",
|
| 150 |
"23": "NOT_ALIVE",
|
| 151 |
"230": "NOT_ALIVE",
|
| 152 |
+
"231": "NOT_ALIVE",
|
| 153 |
"232": "NOT_ALIVE",
|
| 154 |
+
"233": "NON_PARTICIPATING",
|
| 155 |
+
"234": "NOT_ALIVE",
|
| 156 |
+
"235": "SUCCESS",
|
| 157 |
+
"236": "SUCCESS",
|
| 158 |
+
"237": "SUCCESS",
|
| 159 |
"238": "SUCCESS",
|
| 160 |
"239": "NOT_ALIVE",
|
| 161 |
+
"24": "NOT_ALIVE",
|
| 162 |
"240": "NOT_ALIVE",
|
| 163 |
"241": "NON_PARTICIPATING",
|
| 164 |
"242": "NON_PARTICIPATING",
|
| 165 |
"243": "NON_PARTICIPATING",
|
| 166 |
+
"244": "NON_PARTICIPATING",
|
| 167 |
"245": "SUCCESS",
|
| 168 |
+
"246": "SUCCESS",
|
| 169 |
"247": "SUCCESS",
|
| 170 |
"248": "SUCCESS",
|
| 171 |
"249": "NOT_ALIVE",
|
|
|
|
| 174 |
"251": "SUCCESS",
|
| 175 |
"252": "NON_PARTICIPATING",
|
| 176 |
"253": "SUCCESS",
|
| 177 |
+
"254": "NOT_ALIVE",
|
| 178 |
"255": "SUCCESS",
|
| 179 |
"26": "NOT_ALIVE",
|
| 180 |
"27": "SUCCESS",
|
| 181 |
"28": "NON_PARTICIPATING",
|
| 182 |
+
"29": "SUCCESS",
|
| 183 |
"3": "NOT_ALIVE",
|
| 184 |
"30": "NOT_ALIVE",
|
| 185 |
+
"31": "NOT_ALIVE",
|
| 186 |
"32": "NOT_ALIVE",
|
| 187 |
"33": "NON_PARTICIPATING",
|
| 188 |
"34": "NON_PARTICIPATING",
|
| 189 |
+
"35": "SUCCESS",
|
| 190 |
+
"36": "SUCCESS",
|
| 191 |
"37": "NOT_ALIVE",
|
| 192 |
"38": "NOT_ALIVE",
|
| 193 |
+
"39": "NON_PARTICIPATING",
|
| 194 |
"4": "NOT_ALIVE",
|
| 195 |
"40": "NON_PARTICIPATING",
|
| 196 |
+
"41": "SUCCESS",
|
| 197 |
+
"42": "NOT_ALIVE",
|
| 198 |
"43": "SUCCESS",
|
| 199 |
"44": "SUCCESS",
|
| 200 |
"45": "NOT_ALIVE",
|
|
|
|
| 204 |
"49": "NON_PARTICIPATING",
|
| 205 |
"5": "NOT_ALIVE",
|
| 206 |
"50": "NOT_ALIVE",
|
| 207 |
+
"51": "NOT_ALIVE",
|
| 208 |
"52": "NON_PARTICIPATING",
|
| 209 |
"53": "NOT_ALIVE",
|
| 210 |
+
"54": "NON_PARTICIPATING",
|
| 211 |
"55": "NOT_ALIVE",
|
| 212 |
+
"56": "NON_PARTICIPATING",
|
| 213 |
"57": "NOT_ALIVE",
|
| 214 |
"58": "NOT_ALIVE",
|
| 215 |
"59": "NOT_ALIVE",
|
| 216 |
"6": "NOT_ALIVE",
|
| 217 |
+
"60": "NON_PARTICIPATING",
|
| 218 |
"61": "SUCCESS",
|
| 219 |
+
"62": "NOT_ALIVE",
|
| 220 |
"63": "NON_PARTICIPATING",
|
| 221 |
"64": "NOT_ALIVE",
|
| 222 |
"65": "NOT_ALIVE",
|
| 223 |
+
"66": "NON_PARTICIPATING",
|
| 224 |
"67": "NON_PARTICIPATING",
|
| 225 |
"68": "NON_PARTICIPATING",
|
| 226 |
"69": "NON_PARTICIPATING",
|
| 227 |
"7": "NOT_ALIVE",
|
| 228 |
"70": "NOT_ALIVE",
|
| 229 |
"71": "NOT_ALIVE",
|
| 230 |
+
"72": "NON_PARTICIPATING",
|
| 231 |
"73": "NON_PARTICIPATING",
|
| 232 |
+
"74": "NOT_ALIVE",
|
| 233 |
"75": "NOT_ALIVE",
|
| 234 |
+
"76": "NON_PARTICIPATING",
|
| 235 |
+
"77": "NON_PARTICIPATING",
|
| 236 |
+
"78": "NON_PARTICIPATING",
|
| 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": "NOT_ALIVE",
|
|
|
|
| 252 |
"92": "NOT_ALIVE",
|
| 253 |
"93": "NOT_ALIVE",
|
| 254 |
"94": "NON_PARTICIPATING",
|
| 255 |
+
"95": "NON_PARTICIPATING",
|
| 256 |
+
"96": "NOT_ALIVE",
|
| 257 |
"97": "NOT_ALIVE",
|
| 258 |
"98": "NOT_ALIVE",
|
| 259 |
"99": "NON_PARTICIPATING"
|
|
|
|
| 271 |
"initializer_range": 0.02,
|
| 272 |
"inner_step": 0,
|
| 273 |
"intermediate_size": 5632,
|
| 274 |
+
"last_allreduce_block": 6415211,
|
| 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 4924928111
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:54df6015c4aed411d48be64acbb47ac484ef4eda1d21dc96d84c820f9ebc0331
|
| 3 |
size 4924928111
|
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 sha256:2ecd61e969452a9bbbec250a53533a6c909a971bb214078d97bffe362f0aeb4f
|
| 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 4400252933
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:31bd322ddb5633e6e1c2b3f061990d9a9215c1967c1f913747341f2e411bf083
|
| 3 |
size 4400252933
|