Run 5. Outer Step 108. Inner Step 1.

config.json

@@ -7,8 +7,8 @@
     "100": "NOT_ALIVE",
     "101": "NON_PARTICIPATING",
     "102": "NON_PARTICIPATING",
-    "103": "NOT_ALIVE",
-    "104": "NON_PARTICIPATING",
+    "103": "NON_PARTICIPATING",
+    "104": "NOT_ALIVE",
     "105": "NON_PARTICIPATING",
     "106": "NON_PARTICIPATING",
     "107": "NOT_ALIVE",
@@ -24,9 +24,9 @@
     "116": "NON_PARTICIPATING",
     "117": "NOT_ALIVE",
     "118": "NOT_ALIVE",
-    "119": "NON_PARTICIPATING",
-    "12": "NON_PARTICIPATING",
-    "120": "NON_PARTICIPATING",
+    "119": "NOT_ALIVE",
+    "12": "NOT_ALIVE",
+    "120": "NOT_ALIVE",
     "121": "NOT_ALIVE",
     "122": "NON_PARTICIPATING",
     "123": "NOT_ALIVE",
@@ -40,12 +40,12 @@
     "130": "NON_PARTICIPATING",
     "131": "NOT_ALIVE",
     "132": "NON_PARTICIPATING",
-    "133": "SUCCESS",
+    "133": "NOT_ALIVE",
     "134": "NON_PARTICIPATING",
     "135": "NON_PARTICIPATING",
     "136": "SUCCESS",
-    "137": "NOT_ALIVE",
-    "138": "NON_PARTICIPATING",
+    "137": "NON_PARTICIPATING",
+    "138": "SUCCESS",
     "139": "NON_PARTICIPATING",
     "14": "SUCCESS",
     "140": "NOT_ALIVE",
@@ -55,7 +55,7 @@
     "144": "NON_PARTICIPATING",
     "145": "NON_PARTICIPATING",
     "146": "NON_PARTICIPATING",
-    "147": "NON_PARTICIPATING",
+    "147": "NOT_ALIVE",
     "148": "NON_PARTICIPATING",
     "149": "NON_PARTICIPATING",
     "15": "NOT_ALIVE",
@@ -64,26 +64,26 @@
     "152": "NOT_ALIVE",
     "153": "SUCCESS",
     "154": "SUCCESS",
-    "155": "SUCCESS",
+    "155": "NON_PARTICIPATING",
     "156": "NOT_ALIVE",
-    "157": "NOT_ALIVE",
-    "158": "NON_PARTICIPATING",
-    "159": "SUCCESS",
-    "16": "NOT_ALIVE",
-    "160": "NOT_ALIVE",
-    "161": "NOT_ALIVE",
+    "157": "NON_PARTICIPATING",
+    "158": "NOT_ALIVE",
+    "159": "NON_PARTICIPATING",
+    "16": "NON_PARTICIPATING",
+    "160": "NON_PARTICIPATING",
+    "161": "NON_PARTICIPATING",
     "162": "NON_PARTICIPATING",
     "163": "NOT_ALIVE",
-    "164": "SUCCESS",
+    "164": "NON_PARTICIPATING",
     "165": "SUCCESS",
     "166": "NON_PARTICIPATING",
     "167": "NOT_ALIVE",
-    "168": "NON_PARTICIPATING",
+    "168": "NOT_ALIVE",
     "169": "NOT_ALIVE",
     "17": "NOT_ALIVE",
-    "170": "NON_PARTICIPATING",
+    "170": "NOT_ALIVE",
     "171": "NOT_ALIVE",
-    "172": "NOT_ALIVE",
+    "172": "NON_PARTICIPATING",
     "173": "SUCCESS",
     "174": "NOT_ALIVE",
     "175": "NOT_ALIVE",
@@ -95,26 +95,26 @@
     "180": "NOT_ALIVE",
     "181": "NOT_ALIVE",
     "182": "NOT_ALIVE",
-    "183": "NON_PARTICIPATING",
+    "183": "SUCCESS",
     "184": "NOT_ALIVE",
     "185": "NON_PARTICIPATING",
     "186": "NON_PARTICIPATING",
     "187": "NON_PARTICIPATING",
-    "188": "NOT_ALIVE",
+    "188": "NON_PARTICIPATING",
     "189": "NOT_ALIVE",
     "19": "NOT_ALIVE",
     "190": "NON_PARTICIPATING",
     "191": "NOT_ALIVE",
-    "192": "NOT_ALIVE",
-    "193": "NOT_ALIVE",
-    "194": "NOT_ALIVE",
+    "192": "NON_PARTICIPATING",
+    "193": "NON_PARTICIPATING",
+    "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
-    "196": "NON_PARTICIPATING",
-    "197": "SUCCESS",
-    "198": "NON_PARTICIPATING",
-    "199": "NON_PARTICIPATING",
-    "2": "NOT_ALIVE",
-    "20": "NOT_ALIVE",
+    "196": "NOT_ALIVE",
+    "197": "NON_PARTICIPATING",
+    "198": "NOT_ALIVE",
+    "199": "NOT_ALIVE",
+    "2": "NON_PARTICIPATING",
+    "20": "NON_PARTICIPATING",
     "200": "NOT_ALIVE",
     "201": "NOT_ALIVE",
     "202": "NOT_ALIVE",
@@ -129,10 +129,10 @@
     "210": "NON_PARTICIPATING",
     "211": "NON_PARTICIPATING",
     "212": "NON_PARTICIPATING",
-    "213": "NOT_ALIVE",
+    "213": "NON_PARTICIPATING",
     "214": "NON_PARTICIPATING",
     "215": "NON_PARTICIPATING",
-    "216": "NON_PARTICIPATING",
+    "216": "NOT_ALIVE",
     "217": "NON_PARTICIPATING",
     "218": "NON_PARTICIPATING",
     "219": "NON_PARTICIPATING",
@@ -153,84 +153,84 @@
     "232": "NOT_ALIVE",
     "233": "NON_PARTICIPATING",
     "234": "NON_PARTICIPATING",
-    "235": "NOT_ALIVE",
+    "235": "NON_PARTICIPATING",
     "236": "NOT_ALIVE",
-    "237": "SUCCESS",
-    "238": "NOT_ALIVE",
+    "237": "FAIL",
+    "238": "NON_PARTICIPATING",
     "239": "NOT_ALIVE",
-    "24": "NOT_ALIVE",
+    "24": "NON_PARTICIPATING",
     "240": "NOT_ALIVE",
     "241": "NOT_ALIVE",
     "242": "NON_PARTICIPATING",
     "243": "NOT_ALIVE",
     "244": "NON_PARTICIPATING",
     "245": "NON_PARTICIPATING",
-    "246": "NON_PARTICIPATING",
+    "246": "NOT_ALIVE",
     "247": "NON_PARTICIPATING",
     "248": "NOT_ALIVE",
     "249": "NOT_ALIVE",
     "25": "SUCCESS",
-    "250": "NON_PARTICIPATING",
+    "250": "SUCCESS",
     "251": "NOT_ALIVE",
-    "252": "NOT_ALIVE",
+    "252": "NON_PARTICIPATING",
     "253": "NOT_ALIVE",
     "254": "NON_PARTICIPATING",
     "255": "NOT_ALIVE",
-    "26": "NON_PARTICIPATING",
+    "26": "NOT_ALIVE",
     "27": "SUCCESS",
     "28": "NOT_ALIVE",
     "29": "NON_PARTICIPATING",
     "3": "NON_PARTICIPATING",
     "30": "NON_PARTICIPATING",
     "31": "NON_PARTICIPATING",
-    "32": "SUCCESS",
+    "32": "NON_PARTICIPATING",
     "33": "NOT_ALIVE",
     "34": "NOT_ALIVE",
     "35": "SUCCESS",
     "36": "NOT_ALIVE",
     "37": "NOT_ALIVE",
-    "38": "NON_PARTICIPATING",
-    "39": "NOT_ALIVE",
+    "38": "SUCCESS",
+    "39": "NON_PARTICIPATING",
     "4": "NON_PARTICIPATING",
     "40": "NON_PARTICIPATING",
-    "41": "NOT_ALIVE",
+    "41": "SUCCESS",
     "42": "NON_PARTICIPATING",
     "43": "NON_PARTICIPATING",
-    "44": "NOT_ALIVE",
+    "44": "NON_PARTICIPATING",
     "45": "NOT_ALIVE",
     "46": "NOT_ALIVE",
     "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
     "49": "NON_PARTICIPATING",
     "5": "NON_PARTICIPATING",
-    "50": "NON_PARTICIPATING",
+    "50": "NOT_ALIVE",
     "51": "NON_PARTICIPATING",
     "52": "NOT_ALIVE",
-    "53": "NON_PARTICIPATING",
+    "53": "NOT_ALIVE",
     "54": "NON_PARTICIPATING",
     "55": "NOT_ALIVE",
-    "56": "NON_PARTICIPATING",
-    "57": "NOT_ALIVE",
-    "58": "NON_PARTICIPATING",
+    "56": "FAIL",
+    "57": "NON_PARTICIPATING",
+    "58": "NOT_ALIVE",
     "59": "SUCCESS",
     "6": "NON_PARTICIPATING",
     "60": "NON_PARTICIPATING",
     "61": "NON_PARTICIPATING",
     "62": "NON_PARTICIPATING",
-    "63": "NON_PARTICIPATING",
+    "63": "SUCCESS",
     "64": "NOT_ALIVE",
-    "65": "NOT_ALIVE",
+    "65": "NON_PARTICIPATING",
     "66": "NOT_ALIVE",
     "67": "NOT_ALIVE",
-    "68": "NON_PARTICIPATING",
+    "68": "NOT_ALIVE",
     "69": "NOT_ALIVE",
-    "7": "NOT_ALIVE",
-    "70": "NON_PARTICIPATING",
+    "7": "NON_PARTICIPATING",
+    "70": "NOT_ALIVE",
     "71": "NON_PARTICIPATING",
-    "72": "NOT_ALIVE",
+    "72": "NON_PARTICIPATING",
     "73": "SUCCESS",
     "74": "NOT_ALIVE",
-    "75": "NOT_ALIVE",
+    "75": "NON_PARTICIPATING",
     "76": "NON_PARTICIPATING",
     "77": "NON_PARTICIPATING",
     "78": "NOT_ALIVE",
@@ -244,17 +244,17 @@
     "85": "NON_PARTICIPATING",
     "86": "NOT_ALIVE",
     "87": "NON_PARTICIPATING",
-    "88": "NOT_ALIVE",
+    "88": "NON_PARTICIPATING",
     "89": "NOT_ALIVE",
     "9": "NOT_ALIVE",
     "90": "NON_PARTICIPATING",
     "91": "NOT_ALIVE",
-    "92": "SUCCESS",
+    "92": "NOT_ALIVE",
     "93": "NON_PARTICIPATING",
     "94": "SUCCESS",
     "95": "NOT_ALIVE",
     "96": "NON_PARTICIPATING",
-    "97": "NOT_ALIVE",
+    "97": "NON_PARTICIPATING",
     "98": "NON_PARTICIPATING",
     "99": "NOT_ALIVE"
   },
@@ -264,17 +264,17 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "block_list": [
-    5966037,
-    5966045
+    5966107,
+    5966114
   ],
   "bos_token_id": 1,
   "eos_token_id": 2,
   "hidden_act": "silu",
   "hidden_size": 2048,
   "initializer_range": 0.02,
-  "inner_step": 79,
+  "inner_step": 1,
   "intermediate_size": 5632,
-  "last_allreduce_block": 5964830,
+  "last_allreduce_block": 5966042,
   "max_position_embeddings": 2048,
   "mlp_bias": false,
   "model_type": "llama",

inner_optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d43e5cf4a13208191b495cb5bbe2edf0dc603776677240014c58b370aab44281
+oid sha256:4e77343df0b7aeecc9653e1569572e9dc0aeb6df2dc05bf2671a113aa4cf4b54
 size 8800564076

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:743c849b8f118a6686f84a59a703ccfcf450e3b974f078e0f2dc53159f4b53d9
+oid sha256:8ab50aba1ac91ca393486a439395a747914dd1b36334ce98e7d3573bb022e775
 size 4400216536