Run 6. Outer Step 108. Inner Step 6.

config.json

@@ -17,7 +17,7 @@
     "11": "NOT_ALIVE",
     "110": "NOT_ALIVE",
     "111": "NOT_ALIVE",
-    "112": "NOT_ALIVE",
+    "112": "NON_PARTICIPATING",
     "113": "NOT_ALIVE",
     "114": "NOT_ALIVE",
     "115": "NOT_ALIVE",
@@ -37,21 +37,21 @@
     "128": "NOT_ALIVE",
     "129": "NOT_ALIVE",
     "13": "NOT_ALIVE",
-    "130": "SUCCESS",
+    "130": "NOT_ALIVE",
     "131": "NOT_ALIVE",
     "132": "NOT_ALIVE",
-    "133": "NON_PARTICIPATING",
+    "133": "SUCCESS",
     "134": "NOT_ALIVE",
     "135": "NOT_ALIVE",
     "136": "NON_PARTICIPATING",
     "137": "NON_PARTICIPATING",
-    "138": "SUCCESS",
+    "138": "NON_PARTICIPATING",
     "139": "NOT_ALIVE",
     "14": "NOT_ALIVE",
     "140": "NON_PARTICIPATING",
     "141": "NON_PARTICIPATING",
     "142": "NON_PARTICIPATING",
-    "143": "NOT_ALIVE",
+    "143": "SUCCESS",
     "144": "NOT_ALIVE",
     "145": "NOT_ALIVE",
     "146": "NON_PARTICIPATING",
@@ -63,7 +63,7 @@
     "151": "NON_PARTICIPATING",
     "152": "NOT_ALIVE",
     "153": "NOT_ALIVE",
-    "154": "SUCCESS",
+    "154": "NON_PARTICIPATING",
     "155": "NON_PARTICIPATING",
     "156": "NOT_ALIVE",
     "157": "NOT_ALIVE",
@@ -73,18 +73,18 @@
     "160": "NON_PARTICIPATING",
     "161": "NON_PARTICIPATING",
     "162": "NOT_ALIVE",
-    "163": "NON_PARTICIPATING",
-    "164": "NOT_ALIVE",
-    "165": "NON_PARTICIPATING",
+    "163": "NOT_ALIVE",
+    "164": "NON_PARTICIPATING",
+    "165": "NOT_ALIVE",
     "166": "NON_PARTICIPATING",
     "167": "NOT_ALIVE",
     "168": "NON_PARTICIPATING",
     "169": "NON_PARTICIPATING",
     "17": "NOT_ALIVE",
-    "170": "NOT_ALIVE",
-    "171": "NON_PARTICIPATING",
+    "170": "NON_PARTICIPATING",
+    "171": "NOT_ALIVE",
     "172": "NOT_ALIVE",
-    "173": "NON_PARTICIPATING",
+    "173": "SUCCESS",
     "174": "NOT_ALIVE",
     "175": "NON_PARTICIPATING",
     "176": "NOT_ALIVE",
@@ -98,20 +98,20 @@
     "183": "NON_PARTICIPATING",
     "184": "NOT_ALIVE",
     "185": "NON_PARTICIPATING",
-    "186": "NON_PARTICIPATING",
+    "186": "SUCCESS",
     "187": "NOT_ALIVE",
     "188": "NOT_ALIVE",
-    "189": "SUCCESS",
+    "189": "NOT_ALIVE",
     "19": "NOT_ALIVE",
     "190": "NOT_ALIVE",
     "191": "NON_PARTICIPATING",
     "192": "NON_PARTICIPATING",
     "193": "NON_PARTICIPATING",
-    "194": "NOT_ALIVE",
+    "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
     "196": "NOT_ALIVE",
     "197": "NOT_ALIVE",
-    "198": "NOT_ALIVE",
+    "198": "NON_PARTICIPATING",
     "199": "NOT_ALIVE",
     "2": "NOT_ALIVE",
     "20": "NOT_ALIVE",
@@ -119,7 +119,7 @@
     "201": "NOT_ALIVE",
     "202": "NOT_ALIVE",
     "203": "NOT_ALIVE",
-    "204": "NON_PARTICIPATING",
+    "204": "NOT_ALIVE",
     "205": "NOT_ALIVE",
     "206": "NON_PARTICIPATING",
     "207": "NOT_ALIVE",
@@ -128,26 +128,26 @@
     "21": "NOT_ALIVE",
     "210": "NOT_ALIVE",
     "211": "NON_PARTICIPATING",
-    "212": "NOT_ALIVE",
+    "212": "SUCCESS",
     "213": "NOT_ALIVE",
-    "214": "NON_PARTICIPATING",
+    "214": "SUCCESS",
     "215": "NON_PARTICIPATING",
-    "216": "NOT_ALIVE",
+    "216": "NON_PARTICIPATING",
     "217": "NON_PARTICIPATING",
     "218": "NON_PARTICIPATING",
-    "219": "SUCCESS",
+    "219": "NON_PARTICIPATING",
     "22": "NOT_ALIVE",
     "220": "NOT_ALIVE",
     "221": "NOT_ALIVE",
     "222": "NOT_ALIVE",
     "223": "NOT_ALIVE",
     "224": "NON_PARTICIPATING",
-    "225": "NOT_ALIVE",
+    "225": "NON_PARTICIPATING",
     "226": "NON_PARTICIPATING",
     "227": "NOT_ALIVE",
     "228": "NON_PARTICIPATING",
     "229": "NOT_ALIVE",
-    "23": "SUCCESS",
+    "23": "NOT_ALIVE",
     "230": "NON_PARTICIPATING",
     "231": "NOT_ALIVE",
     "232": "NOT_ALIVE",
@@ -157,11 +157,11 @@
     "236": "NON_PARTICIPATING",
     "237": "NON_PARTICIPATING",
     "238": "NOT_ALIVE",
-    "239": "NOT_ALIVE",
+    "239": "NON_PARTICIPATING",
     "24": "NOT_ALIVE",
     "240": "NOT_ALIVE",
     "241": "SUCCESS",
-    "242": "SUCCESS",
+    "242": "NON_PARTICIPATING",
     "243": "NOT_ALIVE",
     "244": "NON_PARTICIPATING",
     "245": "NOT_ALIVE",
@@ -170,14 +170,14 @@
     "248": "NOT_ALIVE",
     "249": "NOT_ALIVE",
     "25": "SUCCESS",
-    "250": "NOT_ALIVE",
-    "251": "SUCCESS",
+    "250": "NON_PARTICIPATING",
+    "251": "NOT_ALIVE",
     "252": "NOT_ALIVE",
-    "253": "NON_PARTICIPATING",
-    "254": "NON_PARTICIPATING",
+    "253": "NOT_ALIVE",
+    "254": "NOT_ALIVE",
     "255": "NON_PARTICIPATING",
-    "26": "NON_PARTICIPATING",
-    "27": "NON_PARTICIPATING",
+    "26": "NOT_ALIVE",
+    "27": "NOT_ALIVE",
     "28": "NON_PARTICIPATING",
     "29": "NOT_ALIVE",
     "3": "NOT_ALIVE",
@@ -186,21 +186,21 @@
     "32": "NOT_ALIVE",
     "33": "NOT_ALIVE",
     "34": "NOT_ALIVE",
-    "35": "NOT_ALIVE",
+    "35": "SUCCESS",
     "36": "NOT_ALIVE",
     "37": "NON_PARTICIPATING",
     "38": "NOT_ALIVE",
     "39": "NOT_ALIVE",
     "4": "NOT_ALIVE",
-    "40": "NON_PARTICIPATING",
+    "40": "SUCCESS",
     "41": "NOT_ALIVE",
     "42": "NOT_ALIVE",
     "43": "NON_PARTICIPATING",
     "44": "SUCCESS",
     "45": "NOT_ALIVE",
-    "46": "NON_PARTICIPATING",
+    "46": "NOT_ALIVE",
     "47": "NOT_ALIVE",
-    "48": "NON_PARTICIPATING",
+    "48": "NOT_ALIVE",
     "49": "NOT_ALIVE",
     "5": "NOT_ALIVE",
     "50": "NOT_ALIVE",
@@ -209,19 +209,19 @@
     "53": "NOT_ALIVE",
     "54": "NOT_ALIVE",
     "55": "NON_PARTICIPATING",
-    "56": "NON_PARTICIPATING",
+    "56": "NOT_ALIVE",
     "57": "NON_PARTICIPATING",
     "58": "NOT_ALIVE",
-    "59": "NON_PARTICIPATING",
+    "59": "NOT_ALIVE",
     "6": "NOT_ALIVE",
     "60": "NON_PARTICIPATING",
     "61": "NON_PARTICIPATING",
     "62": "NOT_ALIVE",
     "63": "NOT_ALIVE",
     "64": "NON_PARTICIPATING",
-    "65": "NOT_ALIVE",
+    "65": "NON_PARTICIPATING",
     "66": "NOT_ALIVE",
-    "67": "NOT_ALIVE",
+    "67": "NON_PARTICIPATING",
     "68": "NOT_ALIVE",
     "69": "NON_PARTICIPATING",
     "7": "NOT_ALIVE",
@@ -240,19 +240,19 @@
     "81": "NOT_ALIVE",
     "82": "NOT_ALIVE",
     "83": "NOT_ALIVE",
-    "84": "NOT_ALIVE",
+    "84": "NON_PARTICIPATING",
     "85": "NOT_ALIVE",
-    "86": "NOT_ALIVE",
+    "86": "NON_PARTICIPATING",
     "87": "NOT_ALIVE",
     "88": "NOT_ALIVE",
     "89": "NOT_ALIVE",
     "9": "NOT_ALIVE",
     "90": "NOT_ALIVE",
     "91": "NOT_ALIVE",
-    "92": "NOT_ALIVE",
+    "92": "NON_PARTICIPATING",
     "93": "NOT_ALIVE",
     "94": "NON_PARTICIPATING",
-    "95": "NOT_ALIVE",
+    "95": "NON_PARTICIPATING",
     "96": "NON_PARTICIPATING",
     "97": "NOT_ALIVE",
     "98": "NON_PARTICIPATING",
@@ -264,17 +264,17 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "block_list": [
-    6208566,
-    6208585
+    6208712,
+    6208735
   ],
   "bos_token_id": 1,
   "eos_token_id": 2,
   "hidden_act": "silu",
   "hidden_size": 2048,
   "initializer_range": 0.02,
-  "inner_step": 387,
+  "inner_step": 6,
   "intermediate_size": 5632,
-  "last_allreduce_block": 6205582,
+  "last_allreduce_block": 6208650,
   "max_position_embeddings": 2048,
   "mlp_bias": false,
   "model_type": "llama",

gradients.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:27f3eb8dcd86c9c110b9b434d0b52143995e96e5e4dce6565fd6350f14480c21
+oid sha256:048bb3d1f0ebfc89025f1ec35f30d18dd8164ed71166c6ce549c81c06881b751
 size 88389626

inner_optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1acc4397ada9439e312417f3cbd07bdad67592baabf26f4a96a2de562bf9a09b
+oid sha256:77c50696844e73c97e4b27ad94577103038f4c8b2234bf61d6f13168710291ed
 size 8800564076

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d2aac26a2c89118746cb0713bc116fc61aba3304d492483dcba7ad649349f255
+oid sha256:c1682d0b786758a9ebdda336f56f2eb4c9c951cc1472fb4c335f040ca0a98299
 size 4400216536