Run 5. Outer Step 57. Inner Step 2.
Browse files- config.json +32 -32
- inner_optimizer.pt +1 -1
- model.safetensors +1 -1
config.json
CHANGED
|
@@ -2,7 +2,7 @@
|
|
| 2 |
"_name_or_path": "distributed/llama-1b",
|
| 3 |
"all_reduce_scores": {
|
| 4 |
"0": "NON_PARTICIPATING",
|
| 5 |
-
"1": "
|
| 6 |
"10": "SUCCESS",
|
| 7 |
"100": "NON_PARTICIPATING",
|
| 8 |
"101": "NON_PARTICIPATING",
|
|
@@ -32,7 +32,7 @@
|
|
| 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",
|
|
@@ -76,22 +76,22 @@
|
|
| 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",
|
| 83 |
"17": "SUCCESS",
|
| 84 |
"170": "NON_PARTICIPATING",
|
| 85 |
-
"171": "
|
| 86 |
"172": "NON_PARTICIPATING",
|
| 87 |
"173": "NON_PARTICIPATING",
|
| 88 |
"174": "NON_PARTICIPATING",
|
| 89 |
"175": "NON_PARTICIPATING",
|
| 90 |
"176": "NON_PARTICIPATING",
|
| 91 |
-
"177": "
|
| 92 |
"178": "NON_PARTICIPATING",
|
| 93 |
"179": "NON_PARTICIPATING",
|
| 94 |
-
"18": "
|
| 95 |
"180": "NON_PARTICIPATING",
|
| 96 |
"181": "NON_PARTICIPATING",
|
| 97 |
"182": "NON_PARTICIPATING",
|
|
@@ -99,10 +99,10 @@
|
|
| 99 |
"184": "NON_PARTICIPATING",
|
| 100 |
"185": "NON_PARTICIPATING",
|
| 101 |
"186": "NON_PARTICIPATING",
|
| 102 |
-
"187": "
|
| 103 |
-
"188": "
|
| 104 |
"189": "NON_PARTICIPATING",
|
| 105 |
-
"19": "
|
| 106 |
"190": "NON_PARTICIPATING",
|
| 107 |
"191": "NON_PARTICIPATING",
|
| 108 |
"192": "NON_PARTICIPATING",
|
|
@@ -117,8 +117,8 @@
|
|
| 117 |
"20": "NON_PARTICIPATING",
|
| 118 |
"200": "NON_PARTICIPATING",
|
| 119 |
"201": "NON_PARTICIPATING",
|
| 120 |
-
"202": "
|
| 121 |
-
"203": "
|
| 122 |
"204": "NON_PARTICIPATING",
|
| 123 |
"205": "NON_PARTICIPATING",
|
| 124 |
"206": "NON_PARTICIPATING",
|
|
@@ -126,14 +126,14 @@
|
|
| 126 |
"208": "NON_PARTICIPATING",
|
| 127 |
"209": "NON_PARTICIPATING",
|
| 128 |
"21": "SUCCESS",
|
| 129 |
-
"210": "
|
| 130 |
"211": "NON_PARTICIPATING",
|
| 131 |
"212": "NON_PARTICIPATING",
|
| 132 |
"213": "NON_PARTICIPATING",
|
| 133 |
"214": "NON_PARTICIPATING",
|
| 134 |
"215": "NON_PARTICIPATING",
|
| 135 |
"216": "NON_PARTICIPATING",
|
| 136 |
-
"217": "
|
| 137 |
"218": "NON_PARTICIPATING",
|
| 138 |
"219": "NON_PARTICIPATING",
|
| 139 |
"22": "SUCCESS",
|
|
@@ -150,7 +150,7 @@
|
|
| 150 |
"23": "SUCCESS",
|
| 151 |
"230": "NON_PARTICIPATING",
|
| 152 |
"231": "NON_PARTICIPATING",
|
| 153 |
-
"232": "
|
| 154 |
"233": "NON_PARTICIPATING",
|
| 155 |
"234": "NON_PARTICIPATING",
|
| 156 |
"235": "NON_PARTICIPATING",
|
|
@@ -168,9 +168,9 @@
|
|
| 168 |
"246": "NON_PARTICIPATING",
|
| 169 |
"247": "NON_PARTICIPATING",
|
| 170 |
"248": "NON_PARTICIPATING",
|
| 171 |
-
"249": "
|
| 172 |
"25": "SUCCESS",
|
| 173 |
-
"250": "
|
| 174 |
"251": "NON_PARTICIPATING",
|
| 175 |
"252": "NON_PARTICIPATING",
|
| 176 |
"253": "NON_PARTICIPATING",
|
|
@@ -199,11 +199,11 @@
|
|
| 199 |
"44": "NON_PARTICIPATING",
|
| 200 |
"45": "SUCCESS",
|
| 201 |
"46": "SUCCESS",
|
| 202 |
-
"47": "
|
| 203 |
-
"48": "
|
| 204 |
-
"49": "
|
| 205 |
"5": "SUCCESS",
|
| 206 |
-
"50": "
|
| 207 |
"51": "FAIL",
|
| 208 |
"52": "NON_PARTICIPATING",
|
| 209 |
"53": "SUCCESS",
|
|
@@ -212,7 +212,7 @@
|
|
| 212 |
"56": "SUCCESS",
|
| 213 |
"57": "NON_PARTICIPATING",
|
| 214 |
"58": "SUCCESS",
|
| 215 |
-
"59": "
|
| 216 |
"6": "NON_PARTICIPATING",
|
| 217 |
"60": "SUCCESS",
|
| 218 |
"61": "NON_PARTICIPATING",
|
|
@@ -222,23 +222,23 @@
|
|
| 222 |
"65": "SUCCESS",
|
| 223 |
"66": "NON_PARTICIPATING",
|
| 224 |
"67": "SUCCESS",
|
| 225 |
-
"68": "
|
| 226 |
"69": "SUCCESS",
|
| 227 |
-
"7": "
|
| 228 |
"70": "SUCCESS",
|
| 229 |
"71": "SUCCESS",
|
| 230 |
"72": "SUCCESS",
|
| 231 |
"73": "SUCCESS",
|
| 232 |
"74": "SUCCESS",
|
| 233 |
"75": "SUCCESS",
|
| 234 |
-
"76": "
|
| 235 |
"77": "NON_PARTICIPATING",
|
| 236 |
-
"78": "
|
| 237 |
"79": "SUCCESS",
|
| 238 |
"8": "NON_PARTICIPATING",
|
| 239 |
"80": "SUCCESS",
|
| 240 |
"81": "SUCCESS",
|
| 241 |
-
"82": "
|
| 242 |
"83": "SUCCESS",
|
| 243 |
"84": "SUCCESS",
|
| 244 |
"85": "SUCCESS",
|
|
@@ -253,10 +253,10 @@
|
|
| 253 |
"93": "SUCCESS",
|
| 254 |
"94": "SUCCESS",
|
| 255 |
"95": "SUCCESS",
|
| 256 |
-
"96": "
|
| 257 |
"97": "NON_PARTICIPATING",
|
| 258 |
"98": "SUCCESS",
|
| 259 |
-
"99": "
|
| 260 |
},
|
| 261 |
"architectures": [
|
| 262 |
"LlamaForCausalLM"
|
|
@@ -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",
|
|
|
|
| 2 |
"_name_or_path": "distributed/llama-1b",
|
| 3 |
"all_reduce_scores": {
|
| 4 |
"0": "NON_PARTICIPATING",
|
| 5 |
+
"1": "SUCCESS",
|
| 6 |
"10": "SUCCESS",
|
| 7 |
"100": "NON_PARTICIPATING",
|
| 8 |
"101": "NON_PARTICIPATING",
|
|
|
|
| 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",
|
|
|
|
| 76 |
"163": "NON_PARTICIPATING",
|
| 77 |
"164": "NON_PARTICIPATING",
|
| 78 |
"165": "NON_PARTICIPATING",
|
| 79 |
+
"166": "SUCCESS",
|
| 80 |
"167": "NON_PARTICIPATING",
|
| 81 |
"168": "NON_PARTICIPATING",
|
| 82 |
"169": "NON_PARTICIPATING",
|
| 83 |
"17": "SUCCESS",
|
| 84 |
"170": "NON_PARTICIPATING",
|
| 85 |
+
"171": "NON_PARTICIPATING",
|
| 86 |
"172": "NON_PARTICIPATING",
|
| 87 |
"173": "NON_PARTICIPATING",
|
| 88 |
"174": "NON_PARTICIPATING",
|
| 89 |
"175": "NON_PARTICIPATING",
|
| 90 |
"176": "NON_PARTICIPATING",
|
| 91 |
+
"177": "NON_PARTICIPATING",
|
| 92 |
"178": "NON_PARTICIPATING",
|
| 93 |
"179": "NON_PARTICIPATING",
|
| 94 |
+
"18": "NON_PARTICIPATING",
|
| 95 |
"180": "NON_PARTICIPATING",
|
| 96 |
"181": "NON_PARTICIPATING",
|
| 97 |
"182": "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": "FAIL",
|
| 106 |
"190": "NON_PARTICIPATING",
|
| 107 |
"191": "NON_PARTICIPATING",
|
| 108 |
"192": "NON_PARTICIPATING",
|
|
|
|
| 117 |
"20": "NON_PARTICIPATING",
|
| 118 |
"200": "NON_PARTICIPATING",
|
| 119 |
"201": "NON_PARTICIPATING",
|
| 120 |
+
"202": "NON_PARTICIPATING",
|
| 121 |
+
"203": "SUCCESS",
|
| 122 |
"204": "NON_PARTICIPATING",
|
| 123 |
"205": "NON_PARTICIPATING",
|
| 124 |
"206": "NON_PARTICIPATING",
|
|
|
|
| 126 |
"208": "NON_PARTICIPATING",
|
| 127 |
"209": "NON_PARTICIPATING",
|
| 128 |
"21": "SUCCESS",
|
| 129 |
+
"210": "SUCCESS",
|
| 130 |
"211": "NON_PARTICIPATING",
|
| 131 |
"212": "NON_PARTICIPATING",
|
| 132 |
"213": "NON_PARTICIPATING",
|
| 133 |
"214": "NON_PARTICIPATING",
|
| 134 |
"215": "NON_PARTICIPATING",
|
| 135 |
"216": "NON_PARTICIPATING",
|
| 136 |
+
"217": "SUCCESS",
|
| 137 |
"218": "NON_PARTICIPATING",
|
| 138 |
"219": "NON_PARTICIPATING",
|
| 139 |
"22": "SUCCESS",
|
|
|
|
| 150 |
"23": "SUCCESS",
|
| 151 |
"230": "NON_PARTICIPATING",
|
| 152 |
"231": "NON_PARTICIPATING",
|
| 153 |
+
"232": "SUCCESS",
|
| 154 |
"233": "NON_PARTICIPATING",
|
| 155 |
"234": "NON_PARTICIPATING",
|
| 156 |
"235": "NON_PARTICIPATING",
|
|
|
|
| 168 |
"246": "NON_PARTICIPATING",
|
| 169 |
"247": "NON_PARTICIPATING",
|
| 170 |
"248": "NON_PARTICIPATING",
|
| 171 |
+
"249": "NON_PARTICIPATING",
|
| 172 |
"25": "SUCCESS",
|
| 173 |
+
"250": "SUCCESS",
|
| 174 |
"251": "NON_PARTICIPATING",
|
| 175 |
"252": "NON_PARTICIPATING",
|
| 176 |
"253": "NON_PARTICIPATING",
|
|
|
|
| 199 |
"44": "NON_PARTICIPATING",
|
| 200 |
"45": "SUCCESS",
|
| 201 |
"46": "SUCCESS",
|
| 202 |
+
"47": "NON_PARTICIPATING",
|
| 203 |
+
"48": "NON_PARTICIPATING",
|
| 204 |
+
"49": "SUCCESS",
|
| 205 |
"5": "SUCCESS",
|
| 206 |
+
"50": "SUCCESS",
|
| 207 |
"51": "FAIL",
|
| 208 |
"52": "NON_PARTICIPATING",
|
| 209 |
"53": "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",
|
|
|
|
| 222 |
"65": "SUCCESS",
|
| 223 |
"66": "NON_PARTICIPATING",
|
| 224 |
"67": "SUCCESS",
|
| 225 |
+
"68": "SUCCESS",
|
| 226 |
"69": "SUCCESS",
|
| 227 |
+
"7": "NON_PARTICIPATING",
|
| 228 |
"70": "SUCCESS",
|
| 229 |
"71": "SUCCESS",
|
| 230 |
"72": "SUCCESS",
|
| 231 |
"73": "SUCCESS",
|
| 232 |
"74": "SUCCESS",
|
| 233 |
"75": "SUCCESS",
|
| 234 |
+
"76": "SUCCESS",
|
| 235 |
"77": "NON_PARTICIPATING",
|
| 236 |
+
"78": "SUCCESS",
|
| 237 |
"79": "SUCCESS",
|
| 238 |
"8": "NON_PARTICIPATING",
|
| 239 |
"80": "SUCCESS",
|
| 240 |
"81": "SUCCESS",
|
| 241 |
+
"82": "SUCCESS",
|
| 242 |
"83": "SUCCESS",
|
| 243 |
"84": "SUCCESS",
|
| 244 |
"85": "SUCCESS",
|
|
|
|
| 253 |
"93": "SUCCESS",
|
| 254 |
"94": "SUCCESS",
|
| 255 |
"95": "SUCCESS",
|
| 256 |
+
"96": "SUCCESS",
|
| 257 |
"97": "NON_PARTICIPATING",
|
| 258 |
"98": "SUCCESS",
|
| 259 |
+
"99": "SUCCESS"
|
| 260 |
},
|
| 261 |
"architectures": [
|
| 262 |
"LlamaForCausalLM"
|
|
|
|
| 264 |
"attention_bias": false,
|
| 265 |
"attention_dropout": 0.0,
|
| 266 |
"block_list": [
|
| 267 |
+
5781546,
|
| 268 |
+
5781575
|
| 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": 5781479,
|
| 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 8800564076
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:941711c44328f521df4c5072806345d27402f301d49199d254e057ff31f1c2fe
|
| 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:a08d99a650d17185fa756d5270d203660b05633479670b2fb1b5353bcc08b85c
|
| 3 |
size 4400216536
|