Run 4. Outer Step 84. Inner Step 0. Peers 82.

config.json

@@ -1,15 +1,15 @@
 {
-  "_name_or_path": "AussieAck/DTS25_model",
+  "_name_or_path": "distributed/optimized-gpt2-1b",
   "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",
     "103": "NON_PARTICIPATING",
-    "104": "NON_PARTICIPATING",
+    "104": "SUCCESS",
     "105": "NON_PARTICIPATING",
     "106": "NON_PARTICIPATING",
     "107": "NON_PARTICIPATING",
@@ -24,7 +24,7 @@
     "115": "NON_PARTICIPATING",
     "116": "NON_PARTICIPATING",
     "117": "NON_PARTICIPATING",
-    "118": "NON_PARTICIPATING",
+    "118": "SUCCESS",
     "119": "NON_PARTICIPATING",
     "12": "SUCCESS",
     "120": "NON_PARTICIPATING",
@@ -38,7 +38,7 @@
     "128": "NON_PARTICIPATING",
     "129": "SUCCESS",
     "13": "SUCCESS",
-    "130": "NON_PARTICIPATING",
+    "130": "SUCCESS",
     "131": "NON_PARTICIPATING",
     "132": "NON_PARTICIPATING",
     "133": "NON_PARTICIPATING",
@@ -48,7 +48,7 @@
     "137": "NON_PARTICIPATING",
     "138": "NON_PARTICIPATING",
     "139": "NON_PARTICIPATING",
-    "14": "NON_PARTICIPATING",
+    "14": "SUCCESS",
     "140": "NON_PARTICIPATING",
     "141": "NON_PARTICIPATING",
     "142": "NON_PARTICIPATING",
@@ -56,43 +56,43 @@
     "144": "NON_PARTICIPATING",
     "145": "NON_PARTICIPATING",
     "146": "NON_PARTICIPATING",
-    "147": "NON_PARTICIPATING",
-    "148": "NON_PARTICIPATING",
+    "147": "SUCCESS",
+    "148": "SUCCESS",
     "149": "NON_PARTICIPATING",
-    "15": "NON_PARTICIPATING",
+    "15": "SUCCESS",
     "150": "NON_PARTICIPATING",
     "151": "NON_PARTICIPATING",
-    "152": "NON_PARTICIPATING",
+    "152": "SUCCESS",
     "153": "NON_PARTICIPATING",
     "154": "NON_PARTICIPATING",
     "155": "NON_PARTICIPATING",
-    "156": "SUCCESS",
+    "156": "NON_PARTICIPATING",
     "157": "NON_PARTICIPATING",
     "158": "NON_PARTICIPATING",
     "159": "NON_PARTICIPATING",
-    "16": "NON_PARTICIPATING",
+    "16": "SUCCESS",
     "160": "NON_PARTICIPATING",
     "161": "NON_PARTICIPATING",
     "162": "NON_PARTICIPATING",
     "163": "NON_PARTICIPATING",
-    "164": "NON_PARTICIPATING",
+    "164": "SUCCESS",
     "165": "NON_PARTICIPATING",
     "166": "NON_PARTICIPATING",
     "167": "NON_PARTICIPATING",
     "168": "NON_PARTICIPATING",
     "169": "NON_PARTICIPATING",
-    "17": "NON_PARTICIPATING",
+    "17": "SUCCESS",
     "170": "NON_PARTICIPATING",
     "171": "NON_PARTICIPATING",
     "172": "NON_PARTICIPATING",
-    "173": "NON_PARTICIPATING",
+    "173": "SUCCESS",
     "174": "NON_PARTICIPATING",
     "175": "NON_PARTICIPATING",
     "176": "NON_PARTICIPATING",
     "177": "NON_PARTICIPATING",
     "178": "NON_PARTICIPATING",
     "179": "NON_PARTICIPATING",
-    "18": "NON_PARTICIPATING",
+    "18": "SUCCESS",
     "180": "NON_PARTICIPATING",
     "181": "NON_PARTICIPATING",
     "182": "NON_PARTICIPATING",
@@ -101,7 +101,7 @@
     "185": "NON_PARTICIPATING",
     "186": "NON_PARTICIPATING",
     "187": "NON_PARTICIPATING",
-    "188": "NON_PARTICIPATING",
+    "188": "SUCCESS",
     "189": "NON_PARTICIPATING",
     "19": "SUCCESS",
     "190": "NON_PARTICIPATING",
@@ -110,15 +110,15 @@
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
-    "196": "NON_PARTICIPATING",
+    "196": "SUCCESS",
     "197": "NON_PARTICIPATING",
     "198": "NON_PARTICIPATING",
     "199": "NON_PARTICIPATING",
-    "2": "NON_PARTICIPATING",
+    "2": "SUCCESS",
     "20": "SUCCESS",
     "200": "NON_PARTICIPATING",
     "201": "NON_PARTICIPATING",
-    "202": "NON_PARTICIPATING",
+    "202": "SUCCESS",
     "203": "NON_PARTICIPATING",
     "204": "NON_PARTICIPATING",
     "205": "NON_PARTICIPATING",
@@ -136,8 +136,8 @@
     "216": "NON_PARTICIPATING",
     "217": "SUCCESS",
     "218": "NON_PARTICIPATING",
-    "219": "NON_PARTICIPATING",
-    "22": "NON_PARTICIPATING",
+    "219": "SUCCESS",
+    "22": "SUCCESS",
     "220": "NON_PARTICIPATING",
     "221": "NON_PARTICIPATING",
     "222": "NON_PARTICIPATING",
@@ -155,11 +155,11 @@
     "233": "NON_PARTICIPATING",
     "234": "NON_PARTICIPATING",
     "235": "NON_PARTICIPATING",
-    "236": "NON_PARTICIPATING",
+    "236": "SUCCESS",
     "237": "NON_PARTICIPATING",
     "238": "NON_PARTICIPATING",
     "239": "NON_PARTICIPATING",
-    "24": "NON_PARTICIPATING",
+    "24": "SUCCESS",
     "240": "NON_PARTICIPATING",
     "241": "NON_PARTICIPATING",
     "242": "NON_PARTICIPATING",
@@ -177,59 +177,59 @@
     "253": "NON_PARTICIPATING",
     "254": "NON_PARTICIPATING",
     "255": "NON_PARTICIPATING",
-    "26": "NON_PARTICIPATING",
+    "26": "SUCCESS",
     "27": "SUCCESS",
     "28": "SUCCESS",
-    "29": "NON_PARTICIPATING",
+    "29": "SUCCESS",
     "3": "NON_PARTICIPATING",
     "30": "NON_PARTICIPATING",
-    "31": "NON_PARTICIPATING",
+    "31": "SUCCESS",
     "32": "NON_PARTICIPATING",
     "33": "NON_PARTICIPATING",
     "34": "SUCCESS",
-    "35": "NON_PARTICIPATING",
+    "35": "SUCCESS",
     "36": "NON_PARTICIPATING",
     "37": "NON_PARTICIPATING",
     "38": "SUCCESS",
-    "39": "SUCCESS",
-    "4": "NON_PARTICIPATING",
+    "39": "NON_PARTICIPATING",
+    "4": "SUCCESS",
     "40": "SUCCESS",
-    "41": "NON_PARTICIPATING",
+    "41": "SUCCESS",
     "42": "SUCCESS",
     "43": "NON_PARTICIPATING",
     "44": "SUCCESS",
-    "45": "NON_PARTICIPATING",
-    "46": "NON_PARTICIPATING",
+    "45": "SUCCESS",
+    "46": "SUCCESS",
     "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
-    "49": "NON_PARTICIPATING",
+    "49": "SUCCESS",
     "5": "SUCCESS",
     "50": "NON_PARTICIPATING",
     "51": "SUCCESS",
     "52": "SUCCESS",
     "53": "NON_PARTICIPATING",
-    "54": "NON_PARTICIPATING",
-    "55": "NON_PARTICIPATING",
-    "56": "NON_PARTICIPATING",
+    "54": "SUCCESS",
+    "55": "SUCCESS",
+    "56": "SUCCESS",
     "57": "NON_PARTICIPATING",
     "58": "NON_PARTICIPATING",
     "59": "SUCCESS",
     "6": "NON_PARTICIPATING",
     "60": "SUCCESS",
-    "61": "NON_PARTICIPATING",
-    "62": "NON_PARTICIPATING",
+    "61": "SUCCESS",
+    "62": "SUCCESS",
     "63": "SUCCESS",
     "64": "SUCCESS",
-    "65": "NON_PARTICIPATING",
+    "65": "SUCCESS",
     "66": "NON_PARTICIPATING",
-    "67": "NON_PARTICIPATING",
+    "67": "SUCCESS",
     "68": "NON_PARTICIPATING",
     "69": "NON_PARTICIPATING",
-    "7": "NON_PARTICIPATING",
+    "7": "SUCCESS",
     "70": "NON_PARTICIPATING",
     "71": "SUCCESS",
-    "72": "NON_PARTICIPATING",
-    "73": "NON_PARTICIPATING",
+    "72": "SUCCESS",
+    "73": "SUCCESS",
     "74": "NON_PARTICIPATING",
     "75": "NON_PARTICIPATING",
     "76": "NON_PARTICIPATING",
@@ -237,26 +237,26 @@
     "78": "NON_PARTICIPATING",
     "79": "NON_PARTICIPATING",
     "8": "SUCCESS",
-    "80": "NON_PARTICIPATING",
+    "80": "SUCCESS",
     "81": "NON_PARTICIPATING",
     "82": "NON_PARTICIPATING",
-    "83": "NON_PARTICIPATING",
-    "84": "NON_PARTICIPATING",
+    "83": "SUCCESS",
+    "84": "SUCCESS",
     "85": "SUCCESS",
     "86": "NON_PARTICIPATING",
-    "87": "NON_PARTICIPATING",
-    "88": "NON_PARTICIPATING",
-    "89": "NON_PARTICIPATING",
+    "87": "SUCCESS",
+    "88": "SUCCESS",
+    "89": "SUCCESS",
     "9": "SUCCESS",
-    "90": "NON_PARTICIPATING",
+    "90": "SUCCESS",
     "91": "SUCCESS",
     "92": "NON_PARTICIPATING",
     "93": "NON_PARTICIPATING",
-    "94": "NON_PARTICIPATING",
+    "94": "SUCCESS",
     "95": "SUCCESS",
     "96": "NON_PARTICIPATING",
-    "97": "NON_PARTICIPATING",
-    "98": "NON_PARTICIPATING",
+    "97": "SUCCESS",
+    "98": "SUCCESS",
     "99": "NON_PARTICIPATING"
   },
   "architectures": [
@@ -275,7 +275,7 @@
   "initializer_range": 0.02,
   "inner_step": 0,
   "inner_steps": 0,
-  "last_allreduce_block": 5672513,
+  "last_allreduce_block": 5673692,
   "layer_norm_epsilon": 1e-05,
   "model_type": "gpt_optimized",
   "n_embd": 1280,

inner_optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c25a9ae2d302ecb423a71f3936176ce0141020768b55f7345f5f650ac9165ab
+oid sha256:533d1d831ec047a0ed40be45d4a2e627a0059bf63ea273b28f24b7f2c7a557af
 size 8081782026

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4d5ac87564c3520b9e2285c664c6e6f86fc242e7eb147e567ef2babc49240c4f
+oid sha256:568529a818118b1adfcffb9e7bac688d0564b1d8763d605cad7a5fc413fc5f40
 size 4040701744

outer_optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2b805da76e18db2fa60a1d6748e241cf511b00a0f0714494eb36582634e4f4cf
+oid sha256:42b9752f78b6fc32d53d77af4bf8217689b01a94c64bc1557e07ee5a0de21a68
 size 4040805354