Run 5. Outer Step 93. Inner Step 2.

config.json

@@ -8,7 +8,7 @@
     "101": "NON_PARTICIPATING",
     "102": "NON_PARTICIPATING",
     "103": "NON_PARTICIPATING",
-    "104": "NON_PARTICIPATING",
+    "104": "NOT_ALIVE",
     "105": "NOT_ALIVE",
     "106": "NON_PARTICIPATING",
     "107": "NOT_ALIVE",
@@ -16,17 +16,17 @@
     "109": "NOT_ALIVE",
     "11": "SUCCESS",
     "110": "NOT_ALIVE",
-    "111": "NON_PARTICIPATING",
+    "111": "NOT_ALIVE",
     "112": "NON_PARTICIPATING",
     "113": "NOT_ALIVE",
-    "114": "NOT_ALIVE",
+    "114": "NON_PARTICIPATING",
     "115": "NOT_ALIVE",
     "116": "NON_PARTICIPATING",
     "117": "NON_PARTICIPATING",
     "118": "NON_PARTICIPATING",
     "119": "NON_PARTICIPATING",
     "12": "NON_PARTICIPATING",
-    "120": "NOT_ALIVE",
+    "120": "NON_PARTICIPATING",
     "121": "NOT_ALIVE",
     "122": "NON_PARTICIPATING",
     "123": "NON_PARTICIPATING",
@@ -35,86 +35,86 @@
     "126": "NOT_ALIVE",
     "127": "NON_PARTICIPATING",
     "128": "NOT_ALIVE",
-    "129": "NON_PARTICIPATING",
+    "129": "NOT_ALIVE",
     "13": "NON_PARTICIPATING",
     "130": "NOT_ALIVE",
     "131": "NOT_ALIVE",
     "132": "NON_PARTICIPATING",
-    "133": "NON_PARTICIPATING",
+    "133": "SUCCESS",
     "134": "NON_PARTICIPATING",
     "135": "NON_PARTICIPATING",
     "136": "NON_PARTICIPATING",
     "137": "NON_PARTICIPATING",
     "138": "NON_PARTICIPATING",
     "139": "NON_PARTICIPATING",
-    "14": "NON_PARTICIPATING",
+    "14": "NOT_ALIVE",
     "140": "NOT_ALIVE",
-    "141": "NOT_ALIVE",
+    "141": "NON_PARTICIPATING",
     "142": "NON_PARTICIPATING",
     "143": "NON_PARTICIPATING",
     "144": "NON_PARTICIPATING",
     "145": "NON_PARTICIPATING",
     "146": "NON_PARTICIPATING",
-    "147": "NOT_ALIVE",
-    "148": "NON_PARTICIPATING",
-    "149": "NON_PARTICIPATING",
+    "147": "NON_PARTICIPATING",
+    "148": "SUCCESS",
+    "149": "NOT_ALIVE",
     "15": "NON_PARTICIPATING",
-    "150": "NON_PARTICIPATING",
+    "150": "NOT_ALIVE",
     "151": "NOT_ALIVE",
     "152": "NON_PARTICIPATING",
-    "153": "NON_PARTICIPATING",
+    "153": "NOT_ALIVE",
     "154": "NON_PARTICIPATING",
     "155": "NON_PARTICIPATING",
     "156": "NON_PARTICIPATING",
     "157": "NON_PARTICIPATING",
     "158": "NON_PARTICIPATING",
-    "159": "SUCCESS",
+    "159": "NON_PARTICIPATING",
     "16": "NON_PARTICIPATING",
-    "160": "NON_PARTICIPATING",
+    "160": "NOT_ALIVE",
     "161": "NON_PARTICIPATING",
     "162": "NOT_ALIVE",
     "163": "NON_PARTICIPATING",
-    "164": "NOT_ALIVE",
+    "164": "NON_PARTICIPATING",
     "165": "SUCCESS",
     "166": "NON_PARTICIPATING",
     "167": "NON_PARTICIPATING",
     "168": "NON_PARTICIPATING",
     "169": "NOT_ALIVE",
     "17": "NON_PARTICIPATING",
-    "170": "NON_PARTICIPATING",
+    "170": "NOT_ALIVE",
     "171": "NOT_ALIVE",
-    "172": "SUCCESS",
+    "172": "NON_PARTICIPATING",
     "173": "SUCCESS",
-    "174": "NOT_ALIVE",
+    "174": "NON_PARTICIPATING",
     "175": "NOT_ALIVE",
     "176": "NOT_ALIVE",
-    "177": "NON_PARTICIPATING",
-    "178": "NOT_ALIVE",
+    "177": "NOT_ALIVE",
+    "178": "NON_PARTICIPATING",
     "179": "NON_PARTICIPATING",
-    "18": "NOT_ALIVE",
+    "18": "NON_PARTICIPATING",
     "180": "NOT_ALIVE",
     "181": "NOT_ALIVE",
     "182": "NOT_ALIVE",
     "183": "SUCCESS",
     "184": "NON_PARTICIPATING",
-    "185": "NOT_ALIVE",
-    "186": "NOT_ALIVE",
+    "185": "NON_PARTICIPATING",
+    "186": "NON_PARTICIPATING",
     "187": "NON_PARTICIPATING",
-    "188": "NON_PARTICIPATING",
+    "188": "NOT_ALIVE",
     "189": "NOT_ALIVE",
     "19": "NOT_ALIVE",
     "190": "NON_PARTICIPATING",
     "191": "NOT_ALIVE",
-    "192": "NON_PARTICIPATING",
+    "192": "NOT_ALIVE",
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
-    "196": "NON_PARTICIPATING",
-    "197": "NON_PARTICIPATING",
-    "198": "NON_PARTICIPATING",
+    "196": "NOT_ALIVE",
+    "197": "NOT_ALIVE",
+    "198": "NOT_ALIVE",
     "199": "NON_PARTICIPATING",
     "2": "NON_PARTICIPATING",
-    "20": "NOT_ALIVE",
+    "20": "NON_PARTICIPATING",
     "200": "NOT_ALIVE",
     "201": "NOT_ALIVE",
     "202": "NON_PARTICIPATING",
@@ -123,19 +123,19 @@
     "205": "NON_PARTICIPATING",
     "206": "NON_PARTICIPATING",
     "207": "NOT_ALIVE",
-    "208": "NOT_ALIVE",
-    "209": "NON_PARTICIPATING",
+    "208": "NON_PARTICIPATING",
+    "209": "NOT_ALIVE",
     "21": "NOT_ALIVE",
-    "210": "NON_PARTICIPATING",
-    "211": "NOT_ALIVE",
-    "212": "NON_PARTICIPATING",
+    "210": "NOT_ALIVE",
+    "211": "NON_PARTICIPATING",
+    "212": "NOT_ALIVE",
     "213": "NON_PARTICIPATING",
-    "214": "NOT_ALIVE",
+    "214": "NON_PARTICIPATING",
     "215": "SUCCESS",
     "216": "NON_PARTICIPATING",
     "217": "NON_PARTICIPATING",
     "218": "NON_PARTICIPATING",
-    "219": "NON_PARTICIPATING",
+    "219": "NOT_ALIVE",
     "22": "NON_PARTICIPATING",
     "220": "NON_PARTICIPATING",
     "221": "NOT_ALIVE",
@@ -144,70 +144,70 @@
     "224": "NOT_ALIVE",
     "225": "SUCCESS",
     "226": "NOT_ALIVE",
-    "227": "NON_PARTICIPATING",
-    "228": "NON_PARTICIPATING",
+    "227": "NOT_ALIVE",
+    "228": "NOT_ALIVE",
     "229": "NON_PARTICIPATING",
     "23": "NON_PARTICIPATING",
     "230": "NOT_ALIVE",
-    "231": "NON_PARTICIPATING",
+    "231": "NOT_ALIVE",
     "232": "NOT_ALIVE",
     "233": "NON_PARTICIPATING",
-    "234": "NOT_ALIVE",
+    "234": "NON_PARTICIPATING",
     "235": "NON_PARTICIPATING",
     "236": "NON_PARTICIPATING",
-    "237": "NOT_ALIVE",
-    "238": "NON_PARTICIPATING",
-    "239": "NON_PARTICIPATING",
+    "237": "NON_PARTICIPATING",
+    "238": "FAIL",
+    "239": "NOT_ALIVE",
     "24": "NON_PARTICIPATING",
     "240": "NOT_ALIVE",
     "241": "NON_PARTICIPATING",
     "242": "NON_PARTICIPATING",
-    "243": "NON_PARTICIPATING",
+    "243": "NOT_ALIVE",
     "244": "NON_PARTICIPATING",
     "245": "NON_PARTICIPATING",
     "246": "NON_PARTICIPATING",
-    "247": "NON_PARTICIPATING",
-    "248": "NON_PARTICIPATING",
+    "247": "NOT_ALIVE",
+    "248": "NOT_ALIVE",
     "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
-    "250": "NOT_ALIVE",
+    "250": "NON_PARTICIPATING",
     "251": "NON_PARTICIPATING",
     "252": "NON_PARTICIPATING",
-    "253": "NOT_ALIVE",
+    "253": "NON_PARTICIPATING",
     "254": "NON_PARTICIPATING",
     "255": "NOT_ALIVE",
-    "26": "NON_PARTICIPATING",
+    "26": "NOT_ALIVE",
     "27": "NON_PARTICIPATING",
-    "28": "NON_PARTICIPATING",
+    "28": "NOT_ALIVE",
     "29": "NON_PARTICIPATING",
     "3": "NON_PARTICIPATING",
-    "30": "NON_PARTICIPATING",
+    "30": "NOT_ALIVE",
     "31": "NON_PARTICIPATING",
     "32": "SUCCESS",
     "33": "NOT_ALIVE",
     "34": "NON_PARTICIPATING",
     "35": "SUCCESS",
-    "36": "NOT_ALIVE",
+    "36": "NON_PARTICIPATING",
     "37": "NOT_ALIVE",
-    "38": "NON_PARTICIPATING",
-    "39": "SUCCESS",
-    "4": "NOT_ALIVE",
-    "40": "SUCCESS",
+    "38": "NOT_ALIVE",
+    "39": "NOT_ALIVE",
+    "4": "NON_PARTICIPATING",
+    "40": "NON_PARTICIPATING",
     "41": "NON_PARTICIPATING",
-    "42": "NON_PARTICIPATING",
+    "42": "NOT_ALIVE",
     "43": "NON_PARTICIPATING",
-    "44": "NOT_ALIVE",
+    "44": "NON_PARTICIPATING",
     "45": "NOT_ALIVE",
     "46": "NOT_ALIVE",
-    "47": "NOT_ALIVE",
+    "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
     "49": "NON_PARTICIPATING",
-    "5": "SUCCESS",
+    "5": "NOT_ALIVE",
     "50": "NON_PARTICIPATING",
-    "51": "NON_PARTICIPATING",
-    "52": "NON_PARTICIPATING",
+    "51": "NOT_ALIVE",
+    "52": "NOT_ALIVE",
     "53": "NON_PARTICIPATING",
-    "54": "NON_PARTICIPATING",
+    "54": "NOT_ALIVE",
     "55": "NON_PARTICIPATING",
     "56": "SUCCESS",
     "57": "NON_PARTICIPATING",
@@ -217,15 +217,15 @@
     "60": "NON_PARTICIPATING",
     "61": "NON_PARTICIPATING",
     "62": "NON_PARTICIPATING",
-    "63": "NON_PARTICIPATING",
-    "64": "NON_PARTICIPATING",
-    "65": "NOT_ALIVE",
+    "63": "SUCCESS",
+    "64": "NOT_ALIVE",
+    "65": "NON_PARTICIPATING",
     "66": "NON_PARTICIPATING",
     "67": "NOT_ALIVE",
     "68": "NON_PARTICIPATING",
-    "69": "NOT_ALIVE",
-    "7": "NON_PARTICIPATING",
-    "70": "NON_PARTICIPATING",
+    "69": "NON_PARTICIPATING",
+    "7": "SUCCESS",
+    "70": "NOT_ALIVE",
     "71": "NON_PARTICIPATING",
     "72": "NOT_ALIVE",
     "73": "NON_PARTICIPATING",
@@ -233,12 +233,12 @@
     "75": "NON_PARTICIPATING",
     "76": "NON_PARTICIPATING",
     "77": "NON_PARTICIPATING",
-    "78": "NOT_ALIVE",
+    "78": "NON_PARTICIPATING",
     "79": "NON_PARTICIPATING",
     "8": "NON_PARTICIPATING",
     "80": "NON_PARTICIPATING",
-    "81": "NON_PARTICIPATING",
-    "82": "NOT_ALIVE",
+    "81": "NOT_ALIVE",
+    "82": "NON_PARTICIPATING",
     "83": "NOT_ALIVE",
     "84": "NON_PARTICIPATING",
     "85": "NOT_ALIVE",
@@ -248,14 +248,14 @@
     "89": "NOT_ALIVE",
     "9": "NOT_ALIVE",
     "90": "NON_PARTICIPATING",
-    "91": "NOT_ALIVE",
+    "91": "NON_PARTICIPATING",
     "92": "NON_PARTICIPATING",
-    "93": "NOT_ALIVE",
+    "93": "NON_PARTICIPATING",
     "94": "NON_PARTICIPATING",
     "95": "NOT_ALIVE",
-    "96": "NON_PARTICIPATING",
-    "97": "NOT_ALIVE",
-    "98": "NON_PARTICIPATING",
+    "96": "NOT_ALIVE",
+    "97": "NON_PARTICIPATING",
+    "98": "NOT_ALIVE",
     "99": "NOT_ALIVE"
   },
   "architectures": [
@@ -264,17 +264,17 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "block_list": [
-    5943815,
-    5943819
+    5943885,
+    5943888
   ],
   "bos_token_id": 1,
   "eos_token_id": 2,
   "hidden_act": "silu",
   "hidden_size": 2048,
   "initializer_range": 0.02,
-  "inner_step": 520,
+  "inner_step": 1,
   "intermediate_size": 5632,
-  "last_allreduce_block": 5941144,
+  "last_allreduce_block": 5943784,
   "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 sha256:0c5e3f38c7bbeff231220ba485ef641e48aeae7c29e8a1760a5454a58a3dbdf8
+oid sha256:63abfa1e9321167e29a3903689b0a3bd8a57a527528b4ce6b4391025148a2819
 size 8800564561

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:37bd2d42f75f8182604d946abda2a7ff0cf7f9226bf392552c68df38c3c5db29
+oid sha256:fd954d374d9129bc027467a1c8b90a6b583aafd35238ee63614f1deff13ca458
 size 4400216536