Run 5. Outer Step 76. Inner Step 1.

config.json

@@ -7,9 +7,9 @@
     "100": "NOT_ALIVE",
     "101": "NOT_ALIVE",
     "102": "NON_PARTICIPATING",
-    "103": "NON_PARTICIPATING",
-    "104": "NOT_ALIVE",
-    "105": "NON_PARTICIPATING",
+    "103": "SUCCESS",
+    "104": "NON_PARTICIPATING",
+    "105": "NOT_ALIVE",
     "106": "NON_PARTICIPATING",
     "107": "NON_PARTICIPATING",
     "108": "NOT_ALIVE",
@@ -19,25 +19,25 @@
     "111": "NON_PARTICIPATING",
     "112": "NON_PARTICIPATING",
     "113": "NOT_ALIVE",
-    "114": "NOT_ALIVE",
+    "114": "NON_PARTICIPATING",
     "115": "NOT_ALIVE",
     "116": "NOT_ALIVE",
-    "117": "NON_PARTICIPATING",
+    "117": "NOT_ALIVE",
     "118": "NON_PARTICIPATING",
     "119": "NOT_ALIVE",
-    "12": "NOT_ALIVE",
+    "12": "NON_PARTICIPATING",
     "120": "NON_PARTICIPATING",
     "121": "NOT_ALIVE",
     "122": "NOT_ALIVE",
     "123": "NOT_ALIVE",
-    "124": "NON_PARTICIPATING",
+    "124": "NOT_ALIVE",
     "125": "NON_PARTICIPATING",
     "126": "NON_PARTICIPATING",
     "127": "NOT_ALIVE",
     "128": "NOT_ALIVE",
     "129": "NOT_ALIVE",
     "13": "NOT_ALIVE",
-    "130": "NON_PARTICIPATING",
+    "130": "NOT_ALIVE",
     "131": "NOT_ALIVE",
     "132": "NOT_ALIVE",
     "133": "SUCCESS",
@@ -45,16 +45,16 @@
     "135": "NON_PARTICIPATING",
     "136": "NOT_ALIVE",
     "137": "NON_PARTICIPATING",
-    "138": "NON_PARTICIPATING",
-    "139": "NOT_ALIVE",
+    "138": "NOT_ALIVE",
+    "139": "NON_PARTICIPATING",
     "14": "NON_PARTICIPATING",
-    "140": "NON_PARTICIPATING",
+    "140": "NOT_ALIVE",
     "141": "NON_PARTICIPATING",
     "142": "SUCCESS",
     "143": "NON_PARTICIPATING",
     "144": "NON_PARTICIPATING",
     "145": "NOT_ALIVE",
-    "146": "NOT_ALIVE",
+    "146": "NON_PARTICIPATING",
     "147": "NON_PARTICIPATING",
     "148": "NON_PARTICIPATING",
     "149": "NON_PARTICIPATING",
@@ -64,13 +64,13 @@
     "152": "NOT_ALIVE",
     "153": "NOT_ALIVE",
     "154": "NOT_ALIVE",
-    "155": "NON_PARTICIPATING",
+    "155": "SUCCESS",
     "156": "NOT_ALIVE",
     "157": "NOT_ALIVE",
     "158": "NOT_ALIVE",
     "159": "NOT_ALIVE",
     "16": "NON_PARTICIPATING",
-    "160": "NOT_ALIVE",
+    "160": "NON_PARTICIPATING",
     "161": "NOT_ALIVE",
     "162": "NON_PARTICIPATING",
     "163": "NOT_ALIVE",
@@ -82,30 +82,30 @@
     "169": "NOT_ALIVE",
     "17": "NOT_ALIVE",
     "170": "NON_PARTICIPATING",
-    "171": "SUCCESS",
+    "171": "NOT_ALIVE",
     "172": "NOT_ALIVE",
     "173": "NON_PARTICIPATING",
     "174": "SUCCESS",
     "175": "NON_PARTICIPATING",
     "176": "NOT_ALIVE",
     "177": "NON_PARTICIPATING",
-    "178": "NOT_ALIVE",
+    "178": "NON_PARTICIPATING",
     "179": "NOT_ALIVE",
     "18": "SUCCESS",
-    "180": "NOT_ALIVE",
+    "180": "NON_PARTICIPATING",
     "181": "NON_PARTICIPATING",
-    "182": "NOT_ALIVE",
+    "182": "NON_PARTICIPATING",
     "183": "SUCCESS",
     "184": "NON_PARTICIPATING",
     "185": "NON_PARTICIPATING",
-    "186": "NOT_ALIVE",
+    "186": "NON_PARTICIPATING",
     "187": "SUCCESS",
     "188": "NON_PARTICIPATING",
     "189": "NON_PARTICIPATING",
-    "19": "NOT_ALIVE",
+    "19": "NON_PARTICIPATING",
     "190": "NOT_ALIVE",
     "191": "NOT_ALIVE",
-    "192": "NOT_ALIVE",
+    "192": "NON_PARTICIPATING",
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
     "195": "NOT_ALIVE",
@@ -115,10 +115,10 @@
     "199": "NOT_ALIVE",
     "2": "NON_PARTICIPATING",
     "20": "NOT_ALIVE",
-    "200": "NOT_ALIVE",
+    "200": "NON_PARTICIPATING",
     "201": "NOT_ALIVE",
-    "202": "NOT_ALIVE",
-    "203": "NON_PARTICIPATING",
+    "202": "NON_PARTICIPATING",
+    "203": "NOT_ALIVE",
     "204": "NON_PARTICIPATING",
     "205": "NOT_ALIVE",
     "206": "NOT_ALIVE",
@@ -126,17 +126,17 @@
     "208": "NOT_ALIVE",
     "209": "NOT_ALIVE",
     "21": "NON_PARTICIPATING",
-    "210": "NON_PARTICIPATING",
+    "210": "SUCCESS",
     "211": "NON_PARTICIPATING",
     "212": "NOT_ALIVE",
     "213": "NON_PARTICIPATING",
     "214": "NOT_ALIVE",
     "215": "NOT_ALIVE",
     "216": "NON_PARTICIPATING",
-    "217": "NON_PARTICIPATING",
+    "217": "NOT_ALIVE",
     "218": "NON_PARTICIPATING",
     "219": "NON_PARTICIPATING",
-    "22": "NOT_ALIVE",
+    "22": "NON_PARTICIPATING",
     "220": "NON_PARTICIPATING",
     "221": "NON_PARTICIPATING",
     "222": "NOT_ALIVE",
@@ -148,17 +148,17 @@
     "228": "NOT_ALIVE",
     "229": "NOT_ALIVE",
     "23": "NOT_ALIVE",
-    "230": "NON_PARTICIPATING",
+    "230": "SUCCESS",
     "231": "NOT_ALIVE",
-    "232": "NON_PARTICIPATING",
-    "233": "SUCCESS",
+    "232": "NOT_ALIVE",
+    "233": "NOT_ALIVE",
     "234": "NOT_ALIVE",
     "235": "NOT_ALIVE",
     "236": "NOT_ALIVE",
     "237": "NOT_ALIVE",
     "238": "NON_PARTICIPATING",
     "239": "NOT_ALIVE",
-    "24": "NON_PARTICIPATING",
+    "24": "NOT_ALIVE",
     "240": "NOT_ALIVE",
     "241": "NOT_ALIVE",
     "242": "NON_PARTICIPATING",
@@ -176,43 +176,43 @@
     "253": "NOT_ALIVE",
     "254": "NOT_ALIVE",
     "255": "NOT_ALIVE",
-    "26": "NON_PARTICIPATING",
+    "26": "NOT_ALIVE",
     "27": "SUCCESS",
     "28": "SUCCESS",
     "29": "NOT_ALIVE",
     "3": "NON_PARTICIPATING",
     "30": "NOT_ALIVE",
-    "31": "NOT_ALIVE",
+    "31": "NON_PARTICIPATING",
     "32": "NOT_ALIVE",
     "33": "NOT_ALIVE",
     "34": "NON_PARTICIPATING",
     "35": "NON_PARTICIPATING",
     "36": "NOT_ALIVE",
-    "37": "NOT_ALIVE",
+    "37": "NON_PARTICIPATING",
     "38": "SUCCESS",
     "39": "NOT_ALIVE",
     "4": "NON_PARTICIPATING",
     "40": "NON_PARTICIPATING",
     "41": "SUCCESS",
     "42": "SUCCESS",
-    "43": "NOT_ALIVE",
+    "43": "NON_PARTICIPATING",
     "44": "NOT_ALIVE",
     "45": "NON_PARTICIPATING",
-    "46": "SUCCESS",
+    "46": "NOT_ALIVE",
     "47": "NON_PARTICIPATING",
-    "48": "SUCCESS",
-    "49": "NON_PARTICIPATING",
+    "48": "NOT_ALIVE",
+    "49": "NOT_ALIVE",
     "5": "NOT_ALIVE",
     "50": "NOT_ALIVE",
     "51": "NON_PARTICIPATING",
     "52": "NOT_ALIVE",
     "53": "NOT_ALIVE",
     "54": "SUCCESS",
-    "55": "NOT_ALIVE",
+    "55": "NON_PARTICIPATING",
     "56": "SUCCESS",
     "57": "SUCCESS",
     "58": "NON_PARTICIPATING",
-    "59": "NOT_ALIVE",
+    "59": "SUCCESS",
     "6": "NOT_ALIVE",
     "60": "NON_PARTICIPATING",
     "61": "NOT_ALIVE",
@@ -223,25 +223,25 @@
     "66": "NON_PARTICIPATING",
     "67": "NOT_ALIVE",
     "68": "NON_PARTICIPATING",
-    "69": "NON_PARTICIPATING",
+    "69": "SUCCESS",
     "7": "SUCCESS",
-    "70": "NON_PARTICIPATING",
-    "71": "NOT_ALIVE",
+    "70": "SUCCESS",
+    "71": "NON_PARTICIPATING",
     "72": "NON_PARTICIPATING",
     "73": "SUCCESS",
     "74": "NON_PARTICIPATING",
-    "75": "SUCCESS",
+    "75": "NOT_ALIVE",
     "76": "NON_PARTICIPATING",
     "77": "NOT_ALIVE",
     "78": "NOT_ALIVE",
     "79": "SUCCESS",
     "8": "NOT_ALIVE",
     "80": "NOT_ALIVE",
-    "81": "NOT_ALIVE",
+    "81": "NON_PARTICIPATING",
     "82": "NON_PARTICIPATING",
-    "83": "NON_PARTICIPATING",
+    "83": "NOT_ALIVE",
     "84": "SUCCESS",
-    "85": "NON_PARTICIPATING",
+    "85": "NOT_ALIVE",
     "86": "SUCCESS",
     "87": "NON_PARTICIPATING",
     "88": "NON_PARTICIPATING",
@@ -249,13 +249,13 @@
     "9": "SUCCESS",
     "90": "SUCCESS",
     "91": "NOT_ALIVE",
-    "92": "SUCCESS",
+    "92": "NOT_ALIVE",
     "93": "NON_PARTICIPATING",
     "94": "NON_PARTICIPATING",
     "95": "NON_PARTICIPATING",
     "96": "NON_PARTICIPATING",
     "97": "SUCCESS",
-    "98": "SUCCESS",
+    "98": "NOT_ALIVE",
     "99": "NON_PARTICIPATING"
   },
   "architectures": [
@@ -264,19 +264,17 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "block_list": [
-    5825370,
-    5825375,
-    5825380,
-    5825385
+    5825471,
+    5825475
   ],
   "bos_token_id": 1,
   "eos_token_id": 2,
   "hidden_act": "silu",
   "hidden_size": 2048,
   "initializer_range": 0.02,
-  "inner_step": 197,
+  "inner_step": 1,
   "intermediate_size": 5632,
-  "last_allreduce_block": 5823295,
+  "last_allreduce_block": 5825419,
   "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 sha256:406bb112130a8a807380c0186421581eeeef507ed878fae13f27dffb430adc38
+oid sha256:237cbbae5b861360f7fe8cb59604dcbba179f9531ebfb9503a7ad8e83f739e0e
 size 8800564561

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:64499f6dc3ec39433e96a4b62b55e39b78a22f218637bc98f2d957f0e151c9c5
+oid sha256:e372b7d76499b665a91a623ae820cdba5cd92723582cb542c33d4d44c77b00dc
 size 4400216536