Epoch 1578. Batch Size 60082. Peers 100.

config.json

@@ -3,139 +3,139 @@
   "activation_function": "gelu_new",
   "all_reduce_scores": {
     "0": "NON_PARTICIPATING",
-    "1": "NON_PARTICIPATING",
+    "1": "SUCCESS",
     "10": "NON_PARTICIPATING",
     "100": "NON_PARTICIPATING",
-    "101": "NON_PARTICIPATING",
-    "102": "NON_PARTICIPATING",
+    "101": "SUCCESS",
+    "102": "SUCCESS",
     "103": "SUCCESS",
     "104": "SUCCESS",
-    "105": "SUCCESS",
+    "105": "NON_PARTICIPATING",
     "106": "NON_PARTICIPATING",
-    "107": "NON_PARTICIPATING",
+    "107": "SUCCESS",
     "108": "NON_PARTICIPATING",
     "109": "NON_PARTICIPATING",
     "11": "SUCCESS",
     "110": "NON_PARTICIPATING",
-    "111": "NON_PARTICIPATING",
-    "112": "NON_PARTICIPATING",
-    "113": "SUCCESS",
+    "111": "SUCCESS",
+    "112": "SUCCESS",
+    "113": "NON_PARTICIPATING",
     "114": "NON_PARTICIPATING",
-    "115": "NON_PARTICIPATING",
-    "116": "SUCCESS",
-    "117": "SUCCESS",
-    "118": "SUCCESS",
-    "119": "NON_PARTICIPATING",
+    "115": "SUCCESS",
+    "116": "NON_PARTICIPATING",
+    "117": "NON_PARTICIPATING",
+    "118": "NON_PARTICIPATING",
+    "119": "SUCCESS",
     "12": "SUCCESS",
     "120": "NON_PARTICIPATING",
     "121": "NON_PARTICIPATING",
-    "122": "SUCCESS",
+    "122": "NON_PARTICIPATING",
     "123": "NON_PARTICIPATING",
     "124": "SUCCESS",
-    "125": "SUCCESS",
-    "126": "SUCCESS",
+    "125": "NON_PARTICIPATING",
+    "126": "NON_PARTICIPATING",
     "127": "NON_PARTICIPATING",
     "128": "NON_PARTICIPATING",
-    "129": "NON_PARTICIPATING",
-    "13": "SUCCESS",
-    "130": "NON_PARTICIPATING",
+    "129": "SUCCESS",
+    "13": "NON_PARTICIPATING",
+    "130": "SUCCESS",
     "131": "NON_PARTICIPATING",
     "132": "NON_PARTICIPATING",
-    "133": "NON_PARTICIPATING",
+    "133": "SUCCESS",
     "134": "SUCCESS",
     "135": "NON_PARTICIPATING",
     "136": "NON_PARTICIPATING",
     "137": "NON_PARTICIPATING",
-    "138": "SUCCESS",
-    "139": "NON_PARTICIPATING",
-    "14": "SUCCESS",
+    "138": "NON_PARTICIPATING",
+    "139": "SUCCESS",
+    "14": "NON_PARTICIPATING",
     "140": "NON_PARTICIPATING",
-    "141": "SUCCESS",
+    "141": "NON_PARTICIPATING",
     "142": "NON_PARTICIPATING",
     "143": "NON_PARTICIPATING",
     "144": "NON_PARTICIPATING",
-    "145": "NON_PARTICIPATING",
-    "146": "SUCCESS",
+    "145": "SUCCESS",
+    "146": "NON_PARTICIPATING",
     "147": "NON_PARTICIPATING",
     "148": "SUCCESS",
-    "149": "SUCCESS",
+    "149": "NON_PARTICIPATING",
     "15": "SUCCESS",
     "150": "NON_PARTICIPATING",
     "151": "NON_PARTICIPATING",
     "152": "SUCCESS",
     "153": "NON_PARTICIPATING",
     "154": "NON_PARTICIPATING",
-    "155": "NON_PARTICIPATING",
+    "155": "SUCCESS",
     "156": "NON_PARTICIPATING",
     "157": "NON_PARTICIPATING",
-    "158": "NON_PARTICIPATING",
-    "159": "SUCCESS",
-    "16": "NON_PARTICIPATING",
+    "158": "SUCCESS",
+    "159": "NON_PARTICIPATING",
+    "16": "SUCCESS",
     "160": "NON_PARTICIPATING",
-    "161": "SUCCESS",
-    "162": "NON_PARTICIPATING",
+    "161": "NON_PARTICIPATING",
+    "162": "SUCCESS",
     "163": "NON_PARTICIPATING",
-    "164": "NON_PARTICIPATING",
-    "165": "NON_PARTICIPATING",
+    "164": "SUCCESS",
+    "165": "SUCCESS",
     "166": "NON_PARTICIPATING",
     "167": "NON_PARTICIPATING",
-    "168": "NON_PARTICIPATING",
-    "169": "NON_PARTICIPATING",
-    "17": "SUCCESS",
+    "168": "SUCCESS",
+    "169": "SUCCESS",
+    "17": "NON_PARTICIPATING",
     "170": "SUCCESS",
     "171": "NON_PARTICIPATING",
     "172": "NON_PARTICIPATING",
     "173": "NON_PARTICIPATING",
     "174": "NON_PARTICIPATING",
     "175": "NON_PARTICIPATING",
-    "176": "NON_PARTICIPATING",
-    "177": "SUCCESS",
+    "176": "SUCCESS",
+    "177": "NON_PARTICIPATING",
     "178": "SUCCESS",
     "179": "SUCCESS",
-    "18": "NON_PARTICIPATING",
+    "18": "SUCCESS",
     "180": "NON_PARTICIPATING",
-    "181": "NON_PARTICIPATING",
-    "182": "SUCCESS",
+    "181": "SUCCESS",
+    "182": "NON_PARTICIPATING",
     "183": "SUCCESS",
     "184": "NON_PARTICIPATING",
     "185": "NON_PARTICIPATING",
     "186": "NON_PARTICIPATING",
-    "187": "SUCCESS",
-    "188": "SUCCESS",
+    "187": "NON_PARTICIPATING",
+    "188": "NON_PARTICIPATING",
     "189": "NON_PARTICIPATING",
     "19": "NON_PARTICIPATING",
-    "190": "SUCCESS",
+    "190": "NON_PARTICIPATING",
     "191": "NON_PARTICIPATING",
     "192": "NON_PARTICIPATING",
-    "193": "SUCCESS",
-    "194": "NON_PARTICIPATING",
-    "195": "SUCCESS",
+    "193": "NON_PARTICIPATING",
+    "194": "SUCCESS",
+    "195": "NON_PARTICIPATING",
     "196": "SUCCESS",
     "197": "SUCCESS",
     "198": "NON_PARTICIPATING",
-    "199": "SUCCESS",
+    "199": "NON_PARTICIPATING",
     "2": "NON_PARTICIPATING",
-    "20": "NON_PARTICIPATING",
-    "200": "NON_PARTICIPATING",
-    "201": "NON_PARTICIPATING",
+    "20": "SUCCESS",
+    "200": "SUCCESS",
+    "201": "SUCCESS",
     "202": "NON_PARTICIPATING",
-    "203": "NON_PARTICIPATING",
+    "203": "SUCCESS",
     "204": "NON_PARTICIPATING",
-    "205": "NON_PARTICIPATING",
-    "206": "SUCCESS",
+    "205": "SUCCESS",
+    "206": "NON_PARTICIPATING",
     "207": "NON_PARTICIPATING",
-    "208": "NON_PARTICIPATING",
-    "209": "SUCCESS",
-    "21": "SUCCESS",
-    "210": "SUCCESS",
+    "208": "SUCCESS",
+    "209": "NON_PARTICIPATING",
+    "21": "NON_PARTICIPATING",
+    "210": "NON_PARTICIPATING",
     "211": "SUCCESS",
-    "212": "NON_PARTICIPATING",
+    "212": "SUCCESS",
     "213": "SUCCESS",
-    "214": "SUCCESS",
-    "215": "NON_PARTICIPATING",
+    "214": "NON_PARTICIPATING",
+    "215": "SUCCESS",
     "216": "NON_PARTICIPATING",
-    "217": "NON_PARTICIPATING",
-    "218": "SUCCESS",
+    "217": "SUCCESS",
+    "218": "NON_PARTICIPATING",
     "219": "NON_PARTICIPATING",
     "22": "NON_PARTICIPATING",
     "220": "NON_PARTICIPATING",
@@ -143,72 +143,72 @@
     "222": "NON_PARTICIPATING",
     "223": "NON_PARTICIPATING",
     "224": "NON_PARTICIPATING",
-    "225": "NON_PARTICIPATING",
+    "225": "SUCCESS",
     "226": "NON_PARTICIPATING",
-    "227": "SUCCESS",
-    "228": "SUCCESS",
+    "227": "NON_PARTICIPATING",
+    "228": "NON_PARTICIPATING",
     "229": "SUCCESS",
-    "23": "SUCCESS",
-    "230": "SUCCESS",
+    "23": "NON_PARTICIPATING",
+    "230": "NON_PARTICIPATING",
     "231": "NON_PARTICIPATING",
-    "232": "SUCCESS",
-    "233": "NON_PARTICIPATING",
-    "234": "SUCCESS",
-    "235": "NON_PARTICIPATING",
+    "232": "NON_PARTICIPATING",
+    "233": "SUCCESS",
+    "234": "NON_PARTICIPATING",
+    "235": "SUCCESS",
     "236": "NON_PARTICIPATING",
-    "237": "NON_PARTICIPATING",
-    "238": "NON_PARTICIPATING",
+    "237": "SUCCESS",
+    "238": "SUCCESS",
     "239": "NON_PARTICIPATING",
-    "24": "NON_PARTICIPATING",
+    "24": "SUCCESS",
     "240": "NON_PARTICIPATING",
     "241": "NON_PARTICIPATING",
-    "242": "NON_PARTICIPATING",
-    "243": "NON_PARTICIPATING",
-    "244": "SUCCESS",
+    "242": "SUCCESS",
+    "243": "SUCCESS",
+    "244": "NON_PARTICIPATING",
     "245": "NON_PARTICIPATING",
     "246": "SUCCESS",
-    "247": "SUCCESS",
-    "248": "NON_PARTICIPATING",
+    "247": "NON_PARTICIPATING",
+    "248": "SUCCESS",
     "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
     "250": "NON_PARTICIPATING",
-    "251": "NON_PARTICIPATING",
+    "251": "SUCCESS",
     "252": "SUCCESS",
     "253": "NON_PARTICIPATING",
-    "254": "NON_PARTICIPATING",
-    "255": "SUCCESS",
+    "254": "SUCCESS",
+    "255": "NON_PARTICIPATING",
     "26": "SUCCESS",
     "27": "SUCCESS",
     "28": "SUCCESS",
-    "29": "SUCCESS",
+    "29": "NON_PARTICIPATING",
     "3": "SUCCESS",
     "30": "NON_PARTICIPATING",
-    "31": "SUCCESS",
+    "31": "NON_PARTICIPATING",
     "32": "NON_PARTICIPATING",
     "33": "NON_PARTICIPATING",
-    "34": "SUCCESS",
-    "35": "NON_PARTICIPATING",
+    "34": "NON_PARTICIPATING",
+    "35": "SUCCESS",
     "36": "NON_PARTICIPATING",
     "37": "SUCCESS",
     "38": "NON_PARTICIPATING",
     "39": "NON_PARTICIPATING",
-    "4": "NON_PARTICIPATING",
-    "40": "NON_PARTICIPATING",
-    "41": "NON_PARTICIPATING",
+    "4": "SUCCESS",
+    "40": "SUCCESS",
+    "41": "SUCCESS",
     "42": "SUCCESS",
-    "43": "SUCCESS",
+    "43": "NON_PARTICIPATING",
     "44": "NON_PARTICIPATING",
     "45": "NON_PARTICIPATING",
     "46": "SUCCESS",
-    "47": "NON_PARTICIPATING",
-    "48": "SUCCESS",
+    "47": "SUCCESS",
+    "48": "NON_PARTICIPATING",
     "49": "NON_PARTICIPATING",
-    "5": "SUCCESS",
-    "50": "SUCCESS",
+    "5": "NON_PARTICIPATING",
+    "50": "NON_PARTICIPATING",
     "51": "NON_PARTICIPATING",
-    "52": "NON_PARTICIPATING",
+    "52": "SUCCESS",
     "53": "SUCCESS",
-    "54": "SUCCESS",
+    "54": "NON_PARTICIPATING",
     "55": "SUCCESS",
     "56": "SUCCESS",
     "57": "NON_PARTICIPATING",
@@ -217,43 +217,43 @@
     "6": "NON_PARTICIPATING",
     "60": "NON_PARTICIPATING",
     "61": "NON_PARTICIPATING",
-    "62": "SUCCESS",
-    "63": "SUCCESS",
+    "62": "NON_PARTICIPATING",
+    "63": "NON_PARTICIPATING",
     "64": "NON_PARTICIPATING",
     "65": "NON_PARTICIPATING",
-    "66": "NON_PARTICIPATING",
+    "66": "SUCCESS",
     "67": "NON_PARTICIPATING",
-    "68": "NON_PARTICIPATING",
-    "69": "SUCCESS",
-    "7": "SUCCESS",
-    "70": "SUCCESS",
-    "71": "SUCCESS",
+    "68": "SUCCESS",
+    "69": "NON_PARTICIPATING",
+    "7": "NON_PARTICIPATING",
+    "70": "NON_PARTICIPATING",
+    "71": "NON_PARTICIPATING",
     "72": "NON_PARTICIPATING",
-    "73": "SUCCESS",
+    "73": "NON_PARTICIPATING",
     "74": "NON_PARTICIPATING",
-    "75": "SUCCESS",
+    "75": "NON_PARTICIPATING",
     "76": "NON_PARTICIPATING",
-    "77": "SUCCESS",
-    "78": "NON_PARTICIPATING",
+    "77": "NON_PARTICIPATING",
+    "78": "SUCCESS",
     "79": "SUCCESS",
-    "8": "NON_PARTICIPATING",
-    "80": "NON_PARTICIPATING",
+    "8": "SUCCESS",
+    "80": "SUCCESS",
     "81": "NON_PARTICIPATING",
-    "82": "NON_PARTICIPATING",
-    "83": "SUCCESS",
-    "84": "NON_PARTICIPATING",
+    "82": "SUCCESS",
+    "83": "NON_PARTICIPATING",
+    "84": "SUCCESS",
     "85": "NON_PARTICIPATING",
     "86": "NON_PARTICIPATING",
     "87": "NON_PARTICIPATING",
     "88": "SUCCESS",
     "89": "SUCCESS",
-    "9": "NON_PARTICIPATING",
-    "90": "NON_PARTICIPATING",
+    "9": "SUCCESS",
+    "90": "SUCCESS",
     "91": "SUCCESS",
     "92": "NON_PARTICIPATING",
-    "93": "SUCCESS",
-    "94": "NON_PARTICIPATING",
-    "95": "SUCCESS",
+    "93": "NON_PARTICIPATING",
+    "94": "SUCCESS",
+    "95": "NON_PARTICIPATING",
     "96": "SUCCESS",
     "97": "NON_PARTICIPATING",
     "98": "SUCCESS",

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ffed3f8b4fdcd159c89a7128106629164d7cdf9d01163bfb9b68a9d90e9a2420
+oid sha256:18791aaf38e3e9b3eddefd50021ad6921d4753515df422dca658126efafe0a69
 size 4975530824

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ec8cc04dd10c0cd06e165d49a06d1ec95d4b99641513bedcb57dc1207da06166
+oid sha256:5addc0d1491e41abcfba64117392399832f3ff7bb10a8cc87d655f2df2c1a5c0
 size 3075942944

optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:232ebab8edfa936e5d7d2a9abe44e50d61cfaad625ba78668c9f09031f71499a
+oid sha256:1bb431d2270f905b9ac5d13f8acee3e9d88f7bde12413785db383e54fd60c192
 size 4675755870