Run 5. Outer Step 90. Inner Step 1.

config.json

@@ -3,101 +3,101 @@
   "all_reduce_scores": {
     "0": "NOT_ALIVE",
     "1": "NOT_ALIVE",
-    "10": "NOT_ALIVE",
+    "10": "NON_PARTICIPATING",
     "100": "NOT_ALIVE",
     "101": "NON_PARTICIPATING",
     "102": "NON_PARTICIPATING",
-    "103": "NON_PARTICIPATING",
+    "103": "NOT_ALIVE",
     "104": "NON_PARTICIPATING",
-    "105": "NOT_ALIVE",
-    "106": "NON_PARTICIPATING",
+    "105": "NON_PARTICIPATING",
+    "106": "NOT_ALIVE",
     "107": "NOT_ALIVE",
     "108": "NON_PARTICIPATING",
-    "109": "NON_PARTICIPATING",
-    "11": "SUCCESS",
-    "110": "NON_PARTICIPATING",
-    "111": "NOT_ALIVE",
-    "112": "SUCCESS",
+    "109": "NOT_ALIVE",
+    "11": "NON_PARTICIPATING",
+    "110": "NOT_ALIVE",
+    "111": "NON_PARTICIPATING",
+    "112": "NON_PARTICIPATING",
     "113": "NON_PARTICIPATING",
     "114": "NON_PARTICIPATING",
     "115": "NON_PARTICIPATING",
     "116": "NON_PARTICIPATING",
-    "117": "SUCCESS",
+    "117": "NOT_ALIVE",
     "118": "NON_PARTICIPATING",
     "119": "NOT_ALIVE",
     "12": "NON_PARTICIPATING",
-    "120": "NON_PARTICIPATING",
+    "120": "NOT_ALIVE",
     "121": "NON_PARTICIPATING",
     "122": "NON_PARTICIPATING",
     "123": "NON_PARTICIPATING",
-    "124": "NON_PARTICIPATING",
+    "124": "NOT_ALIVE",
     "125": "NON_PARTICIPATING",
-    "126": "NON_PARTICIPATING",
+    "126": "NOT_ALIVE",
     "127": "NON_PARTICIPATING",
     "128": "NON_PARTICIPATING",
     "129": "NON_PARTICIPATING",
     "13": "NON_PARTICIPATING",
-    "130": "NOT_ALIVE",
+    "130": "NON_PARTICIPATING",
     "131": "NOT_ALIVE",
     "132": "SUCCESS",
-    "133": "NOT_ALIVE",
+    "133": "NON_PARTICIPATING",
     "134": "NON_PARTICIPATING",
     "135": "NON_PARTICIPATING",
-    "136": "NOT_ALIVE",
-    "137": "SUCCESS",
-    "138": "NOT_ALIVE",
+    "136": "SUCCESS",
+    "137": "NON_PARTICIPATING",
+    "138": "NON_PARTICIPATING",
     "139": "NON_PARTICIPATING",
     "14": "NOT_ALIVE",
     "140": "NON_PARTICIPATING",
     "141": "NON_PARTICIPATING",
-    "142": "SUCCESS",
+    "142": "NON_PARTICIPATING",
     "143": "NON_PARTICIPATING",
     "144": "NON_PARTICIPATING",
     "145": "NON_PARTICIPATING",
     "146": "NON_PARTICIPATING",
     "147": "NON_PARTICIPATING",
-    "148": "NOT_ALIVE",
+    "148": "NON_PARTICIPATING",
     "149": "NON_PARTICIPATING",
     "15": "NOT_ALIVE",
     "150": "NON_PARTICIPATING",
     "151": "NOT_ALIVE",
     "152": "NON_PARTICIPATING",
-    "153": "SUCCESS",
+    "153": "NON_PARTICIPATING",
     "154": "NON_PARTICIPATING",
     "155": "NON_PARTICIPATING",
-    "156": "SUCCESS",
+    "156": "NON_PARTICIPATING",
     "157": "NON_PARTICIPATING",
     "158": "NON_PARTICIPATING",
     "159": "SUCCESS",
     "16": "NON_PARTICIPATING",
     "160": "NON_PARTICIPATING",
-    "161": "NON_PARTICIPATING",
+    "161": "NOT_ALIVE",
     "162": "NON_PARTICIPATING",
     "163": "NON_PARTICIPATING",
     "164": "NON_PARTICIPATING",
-    "165": "NON_PARTICIPATING",
+    "165": "SUCCESS",
     "166": "NON_PARTICIPATING",
     "167": "NON_PARTICIPATING",
-    "168": "NOT_ALIVE",
+    "168": "NON_PARTICIPATING",
     "169": "NOT_ALIVE",
     "17": "NON_PARTICIPATING",
     "170": "NON_PARTICIPATING",
     "171": "NOT_ALIVE",
     "172": "NON_PARTICIPATING",
-    "173": "NOT_ALIVE",
+    "173": "SUCCESS",
     "174": "NOT_ALIVE",
     "175": "NOT_ALIVE",
     "176": "NOT_ALIVE",
-    "177": "NON_PARTICIPATING",
+    "177": "NOT_ALIVE",
     "178": "NON_PARTICIPATING",
     "179": "NON_PARTICIPATING",
     "18": "NON_PARTICIPATING",
     "180": "NOT_ALIVE",
     "181": "NON_PARTICIPATING",
     "182": "NOT_ALIVE",
-    "183": "NON_PARTICIPATING",
+    "183": "SUCCESS",
     "184": "NON_PARTICIPATING",
-    "185": "NOT_ALIVE",
+    "185": "NON_PARTICIPATING",
     "186": "NON_PARTICIPATING",
     "187": "NOT_ALIVE",
     "188": "NON_PARTICIPATING",
@@ -109,31 +109,31 @@
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
-    "196": "NON_PARTICIPATING",
+    "196": "NOT_ALIVE",
     "197": "NOT_ALIVE",
     "198": "NON_PARTICIPATING",
     "199": "NON_PARTICIPATING",
-    "2": "NOT_ALIVE",
+    "2": "NON_PARTICIPATING",
     "20": "NOT_ALIVE",
     "200": "NOT_ALIVE",
     "201": "NON_PARTICIPATING",
     "202": "NON_PARTICIPATING",
     "203": "NON_PARTICIPATING",
     "204": "NON_PARTICIPATING",
-    "205": "NON_PARTICIPATING",
+    "205": "NOT_ALIVE",
     "206": "NON_PARTICIPATING",
     "207": "NOT_ALIVE",
     "208": "NON_PARTICIPATING",
-    "209": "NON_PARTICIPATING",
-    "21": "NOT_ALIVE",
+    "209": "NOT_ALIVE",
+    "21": "NON_PARTICIPATING",
     "210": "NON_PARTICIPATING",
     "211": "NON_PARTICIPATING",
     "212": "NON_PARTICIPATING",
-    "213": "NOT_ALIVE",
+    "213": "NON_PARTICIPATING",
     "214": "NON_PARTICIPATING",
     "215": "SUCCESS",
     "216": "NON_PARTICIPATING",
-    "217": "NOT_ALIVE",
+    "217": "NON_PARTICIPATING",
     "218": "NON_PARTICIPATING",
     "219": "NON_PARTICIPATING",
     "22": "NON_PARTICIPATING",
@@ -142,38 +142,38 @@
     "222": "NON_PARTICIPATING",
     "223": "NON_PARTICIPATING",
     "224": "NOT_ALIVE",
-    "225": "SUCCESS",
+    "225": "NOT_ALIVE",
     "226": "NOT_ALIVE",
     "227": "NON_PARTICIPATING",
     "228": "NON_PARTICIPATING",
     "229": "NON_PARTICIPATING",
     "23": "NON_PARTICIPATING",
-    "230": "NOT_ALIVE",
+    "230": "NON_PARTICIPATING",
     "231": "NON_PARTICIPATING",
-    "232": "NON_PARTICIPATING",
-    "233": "NON_PARTICIPATING",
+    "232": "NOT_ALIVE",
+    "233": "NOT_ALIVE",
     "234": "NON_PARTICIPATING",
     "235": "NON_PARTICIPATING",
     "236": "NON_PARTICIPATING",
-    "237": "SUCCESS",
-    "238": "NOT_ALIVE",
+    "237": "NON_PARTICIPATING",
+    "238": "NON_PARTICIPATING",
     "239": "NON_PARTICIPATING",
     "24": "SUCCESS",
     "240": "NON_PARTICIPATING",
     "241": "NON_PARTICIPATING",
-    "242": "NOT_ALIVE",
+    "242": "NON_PARTICIPATING",
     "243": "NON_PARTICIPATING",
-    "244": "NON_PARTICIPATING",
+    "244": "NOT_ALIVE",
     "245": "NON_PARTICIPATING",
     "246": "NON_PARTICIPATING",
-    "247": "NON_PARTICIPATING",
+    "247": "NOT_ALIVE",
     "248": "NON_PARTICIPATING",
     "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
     "250": "NON_PARTICIPATING",
-    "251": "NOT_ALIVE",
-    "252": "NOT_ALIVE",
-    "253": "NON_PARTICIPATING",
+    "251": "NON_PARTICIPATING",
+    "252": "NON_PARTICIPATING",
+    "253": "NOT_ALIVE",
     "254": "NON_PARTICIPATING",
     "255": "NON_PARTICIPATING",
     "26": "NON_PARTICIPATING",
@@ -190,49 +190,49 @@
     "36": "NON_PARTICIPATING",
     "37": "NOT_ALIVE",
     "38": "SUCCESS",
-    "39": "NON_PARTICIPATING",
+    "39": "SUCCESS",
     "4": "NON_PARTICIPATING",
     "40": "NON_PARTICIPATING",
-    "41": "NON_PARTICIPATING",
+    "41": "SUCCESS",
     "42": "NON_PARTICIPATING",
     "43": "NON_PARTICIPATING",
     "44": "NON_PARTICIPATING",
-    "45": "NON_PARTICIPATING",
+    "45": "NOT_ALIVE",
     "46": "NON_PARTICIPATING",
     "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
-    "49": "NON_PARTICIPATING",
-    "5": "SUCCESS",
-    "50": "NOT_ALIVE",
-    "51": "NOT_ALIVE",
-    "52": "NOT_ALIVE",
+    "49": "NOT_ALIVE",
+    "5": "NON_PARTICIPATING",
+    "50": "NON_PARTICIPATING",
+    "51": "NON_PARTICIPATING",
+    "52": "NON_PARTICIPATING",
     "53": "NON_PARTICIPATING",
     "54": "NON_PARTICIPATING",
     "55": "NON_PARTICIPATING",
-    "56": "NON_PARTICIPATING",
-    "57": "NON_PARTICIPATING",
-    "58": "NOT_ALIVE",
+    "56": "SUCCESS",
+    "57": "NOT_ALIVE",
+    "58": "NON_PARTICIPATING",
     "59": "NON_PARTICIPATING",
     "6": "NON_PARTICIPATING",
     "60": "NON_PARTICIPATING",
-    "61": "NON_PARTICIPATING",
-    "62": "NON_PARTICIPATING",
-    "63": "NON_PARTICIPATING",
+    "61": "NOT_ALIVE",
+    "62": "SUCCESS",
+    "63": "NOT_ALIVE",
     "64": "NON_PARTICIPATING",
     "65": "NON_PARTICIPATING",
     "66": "NON_PARTICIPATING",
     "67": "NOT_ALIVE",
-    "68": "NON_PARTICIPATING",
-    "69": "NON_PARTICIPATING",
+    "68": "SUCCESS",
+    "69": "NOT_ALIVE",
     "7": "NON_PARTICIPATING",
     "70": "NON_PARTICIPATING",
     "71": "NON_PARTICIPATING",
     "72": "NON_PARTICIPATING",
     "73": "NON_PARTICIPATING",
-    "74": "NON_PARTICIPATING",
+    "74": "NOT_ALIVE",
     "75": "NON_PARTICIPATING",
     "76": "NON_PARTICIPATING",
-    "77": "NON_PARTICIPATING",
+    "77": "NOT_ALIVE",
     "78": "NON_PARTICIPATING",
     "79": "NON_PARTICIPATING",
     "8": "NON_PARTICIPATING",
@@ -242,20 +242,20 @@
     "83": "NON_PARTICIPATING",
     "84": "NON_PARTICIPATING",
     "85": "NON_PARTICIPATING",
-    "86": "NOT_ALIVE",
+    "86": "NON_PARTICIPATING",
     "87": "NOT_ALIVE",
     "88": "NON_PARTICIPATING",
     "89": "NOT_ALIVE",
-    "9": "NOT_ALIVE",
+    "9": "NON_PARTICIPATING",
     "90": "NOT_ALIVE",
     "91": "NON_PARTICIPATING",
     "92": "NON_PARTICIPATING",
     "93": "NON_PARTICIPATING",
     "94": "NON_PARTICIPATING",
     "95": "NOT_ALIVE",
-    "96": "NOT_ALIVE",
-    "97": "NOT_ALIVE",
-    "98": "NOT_ALIVE",
+    "96": "NON_PARTICIPATING",
+    "97": "NON_PARTICIPATING",
+    "98": "NON_PARTICIPATING",
     "99": "NOT_ALIVE"
   },
   "architectures": [
@@ -264,17 +264,17 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "block_list": [
-    5938622,
-    5938638
+    5938744,
+    5938756
   ],
   "bos_token_id": 1,
   "eos_token_id": 2,
   "hidden_act": "silu",
   "hidden_size": 2048,
   "initializer_range": 0.02,
-  "inner_step": 45,
+  "inner_step": 1,
   "intermediate_size": 5632,
-  "last_allreduce_block": 5937361,
+  "last_allreduce_block": 5938691,
   "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 sha256:307e396e3652a111fccf21bcd9010b31c46be73626801374197b41d8bd594dba
+oid sha256:cc5f898bb01305d69cd052bfc97205760c32f486d2685580b847d009dd22784e
 size 8800564076

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf6fdf28b973d1bb74f791fcc5def2fedbef76debaa43ef7d40ab06187ca9b2b
+oid sha256:7598caf5d31c6c148a704b2824960e57157784de37765375e467c1d571e676d0
 size 4400216536