Run 4. Outer Step 11. Inner Step 0.

config.json

@@ -6,33 +6,33 @@
     "1": "NON_PARTICIPATING",
     "10": "SUCCESS",
     "100": "NON_PARTICIPATING",
-    "101": "SUCCESS",
-    "102": "NON_PARTICIPATING",
+    "101": "NON_PARTICIPATING",
+    "102": "SUCCESS",
     "103": "NON_PARTICIPATING",
     "104": "NON_PARTICIPATING",
     "105": "NON_PARTICIPATING",
     "106": "NON_PARTICIPATING",
     "107": "NON_PARTICIPATING",
     "108": "NON_PARTICIPATING",
-    "109": "SUCCESS",
+    "109": "NON_PARTICIPATING",
     "11": "SUCCESS",
     "110": "SUCCESS",
     "111": "NON_PARTICIPATING",
     "112": "NON_PARTICIPATING",
     "113": "NON_PARTICIPATING",
-    "114": "NON_PARTICIPATING",
+    "114": "SUCCESS",
     "115": "NON_PARTICIPATING",
     "116": "NON_PARTICIPATING",
-    "117": "SUCCESS",
-    "118": "SUCCESS",
+    "117": "NON_PARTICIPATING",
+    "118": "NON_PARTICIPATING",
     "119": "NON_PARTICIPATING",
-    "12": "FAIL",
-    "120": "NON_PARTICIPATING",
+    "12": "NON_PARTICIPATING",
+    "120": "SUCCESS",
     "121": "NON_PARTICIPATING",
     "122": "SUCCESS",
-    "123": "SUCCESS",
+    "123": "NON_PARTICIPATING",
     "124": "NON_PARTICIPATING",
-    "125": "NON_PARTICIPATING",
+    "125": "SUCCESS",
     "126": "NON_PARTICIPATING",
     "127": "NON_PARTICIPATING",
     "128": "NON_PARTICIPATING",
@@ -40,10 +40,10 @@
     "13": "NON_PARTICIPATING",
     "130": "NON_PARTICIPATING",
     "131": "NON_PARTICIPATING",
-    "132": "SUCCESS",
+    "132": "NON_PARTICIPATING",
     "133": "NON_PARTICIPATING",
     "134": "SUCCESS",
-    "135": "NON_PARTICIPATING",
+    "135": "SUCCESS",
     "136": "NON_PARTICIPATING",
     "137": "NON_PARTICIPATING",
     "138": "NON_PARTICIPATING",
@@ -51,53 +51,53 @@
     "14": "NON_PARTICIPATING",
     "140": "NON_PARTICIPATING",
     "141": "NON_PARTICIPATING",
-    "142": "NON_PARTICIPATING",
+    "142": "SUCCESS",
     "143": "NON_PARTICIPATING",
     "144": "NON_PARTICIPATING",
     "145": "SUCCESS",
     "146": "NON_PARTICIPATING",
-    "147": "NON_PARTICIPATING",
-    "148": "SUCCESS",
-    "149": "SUCCESS",
-    "15": "SUCCESS",
+    "147": "SUCCESS",
+    "148": "NON_PARTICIPATING",
+    "149": "NON_PARTICIPATING",
+    "15": "NON_PARTICIPATING",
     "150": "NON_PARTICIPATING",
     "151": "NON_PARTICIPATING",
     "152": "NON_PARTICIPATING",
-    "153": "NON_PARTICIPATING",
+    "153": "SUCCESS",
     "154": "NON_PARTICIPATING",
     "155": "NON_PARTICIPATING",
-    "156": "NON_PARTICIPATING",
+    "156": "SUCCESS",
     "157": "NON_PARTICIPATING",
     "158": "NON_PARTICIPATING",
-    "159": "SUCCESS",
+    "159": "NON_PARTICIPATING",
     "16": "NON_PARTICIPATING",
-    "160": "SUCCESS",
+    "160": "NON_PARTICIPATING",
     "161": "NON_PARTICIPATING",
     "162": "NON_PARTICIPATING",
     "163": "NON_PARTICIPATING",
     "164": "NON_PARTICIPATING",
     "165": "NON_PARTICIPATING",
     "166": "NON_PARTICIPATING",
-    "167": "SUCCESS",
+    "167": "NON_PARTICIPATING",
     "168": "NON_PARTICIPATING",
     "169": "NON_PARTICIPATING",
     "17": "SUCCESS",
-    "170": "NON_PARTICIPATING",
+    "170": "SUCCESS",
     "171": "NON_PARTICIPATING",
-    "172": "SUCCESS",
+    "172": "NON_PARTICIPATING",
     "173": "NON_PARTICIPATING",
     "174": "NON_PARTICIPATING",
-    "175": "SUCCESS",
+    "175": "NON_PARTICIPATING",
     "176": "NON_PARTICIPATING",
-    "177": "NON_PARTICIPATING",
-    "178": "NON_PARTICIPATING",
+    "177": "SUCCESS",
+    "178": "SUCCESS",
     "179": "NON_PARTICIPATING",
     "18": "NON_PARTICIPATING",
     "180": "NON_PARTICIPATING",
     "181": "NON_PARTICIPATING",
     "182": "NON_PARTICIPATING",
     "183": "NON_PARTICIPATING",
-    "184": "NON_PARTICIPATING",
+    "184": "SUCCESS",
     "185": "NON_PARTICIPATING",
     "186": "SUCCESS",
     "187": "NON_PARTICIPATING",
@@ -107,22 +107,22 @@
     "190": "NON_PARTICIPATING",
     "191": "NON_PARTICIPATING",
     "192": "NON_PARTICIPATING",
-    "193": "NON_PARTICIPATING",
+    "193": "SUCCESS",
     "194": "SUCCESS",
     "195": "SUCCESS",
     "196": "NON_PARTICIPATING",
     "197": "NON_PARTICIPATING",
     "198": "NON_PARTICIPATING",
     "199": "NON_PARTICIPATING",
-    "2": "SUCCESS",
+    "2": "NON_PARTICIPATING",
     "20": "SUCCESS",
     "200": "SUCCESS",
-    "201": "NON_PARTICIPATING",
+    "201": "SUCCESS",
     "202": "SUCCESS",
-    "203": "SUCCESS",
+    "203": "NON_PARTICIPATING",
     "204": "NON_PARTICIPATING",
     "205": "NON_PARTICIPATING",
-    "206": "SUCCESS",
+    "206": "NON_PARTICIPATING",
     "207": "NON_PARTICIPATING",
     "208": "NON_PARTICIPATING",
     "209": "NON_PARTICIPATING",
@@ -132,16 +132,16 @@
     "212": "NON_PARTICIPATING",
     "213": "NON_PARTICIPATING",
     "214": "NON_PARTICIPATING",
-    "215": "SUCCESS",
+    "215": "NON_PARTICIPATING",
     "216": "NON_PARTICIPATING",
     "217": "NON_PARTICIPATING",
-    "218": "SUCCESS",
+    "218": "NON_PARTICIPATING",
     "219": "NON_PARTICIPATING",
-    "22": "SUCCESS",
-    "220": "SUCCESS",
-    "221": "SUCCESS",
-    "222": "SUCCESS",
-    "223": "SUCCESS",
+    "22": "NON_PARTICIPATING",
+    "220": "NON_PARTICIPATING",
+    "221": "NON_PARTICIPATING",
+    "222": "NON_PARTICIPATING",
+    "223": "NON_PARTICIPATING",
     "224": "NON_PARTICIPATING",
     "225": "SUCCESS",
     "226": "SUCCESS",
@@ -150,7 +150,7 @@
     "229": "NON_PARTICIPATING",
     "23": "NON_PARTICIPATING",
     "230": "NON_PARTICIPATING",
-    "231": "SUCCESS",
+    "231": "NON_PARTICIPATING",
     "232": "NON_PARTICIPATING",
     "233": "NON_PARTICIPATING",
     "234": "NON_PARTICIPATING",
@@ -161,17 +161,17 @@
     "239": "NON_PARTICIPATING",
     "24": "NON_PARTICIPATING",
     "240": "NON_PARTICIPATING",
-    "241": "NON_PARTICIPATING",
+    "241": "SUCCESS",
     "242": "NON_PARTICIPATING",
     "243": "NON_PARTICIPATING",
     "244": "NON_PARTICIPATING",
     "245": "NON_PARTICIPATING",
     "246": "NON_PARTICIPATING",
     "247": "SUCCESS",
-    "248": "SUCCESS",
-    "249": "SUCCESS",
+    "248": "NON_PARTICIPATING",
+    "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
-    "250": "SUCCESS",
+    "250": "NON_PARTICIPATING",
     "251": "NON_PARTICIPATING",
     "252": "NON_PARTICIPATING",
     "253": "NON_PARTICIPATING",
@@ -180,84 +180,84 @@
     "26": "SUCCESS",
     "27": "SUCCESS",
     "28": "NON_PARTICIPATING",
-    "29": "FAIL",
+    "29": "NON_PARTICIPATING",
     "3": "NON_PARTICIPATING",
     "30": "SUCCESS",
     "31": "NON_PARTICIPATING",
-    "32": "SUCCESS",
+    "32": "NON_PARTICIPATING",
     "33": "NON_PARTICIPATING",
     "34": "SUCCESS",
-    "35": "SUCCESS",
+    "35": "NON_PARTICIPATING",
     "36": "NON_PARTICIPATING",
-    "37": "SUCCESS",
+    "37": "NON_PARTICIPATING",
     "38": "NON_PARTICIPATING",
-    "39": "NON_PARTICIPATING",
+    "39": "SUCCESS",
     "4": "SUCCESS",
-    "40": "SUCCESS",
-    "41": "SUCCESS",
+    "40": "NON_PARTICIPATING",
+    "41": "NON_PARTICIPATING",
     "42": "NON_PARTICIPATING",
-    "43": "SUCCESS",
+    "43": "NON_PARTICIPATING",
     "44": "SUCCESS",
-    "45": "SUCCESS",
-    "46": "SUCCESS",
+    "45": "NON_PARTICIPATING",
+    "46": "NON_PARTICIPATING",
     "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
     "49": "SUCCESS",
     "5": "NON_PARTICIPATING",
-    "50": "SUCCESS",
+    "50": "NON_PARTICIPATING",
     "51": "SUCCESS",
     "52": "NON_PARTICIPATING",
-    "53": "SUCCESS",
+    "53": "NON_PARTICIPATING",
     "54": "NON_PARTICIPATING",
-    "55": "FAIL",
-    "56": "SUCCESS",
-    "57": "NON_PARTICIPATING",
+    "55": "SUCCESS",
+    "56": "NON_PARTICIPATING",
+    "57": "SUCCESS",
     "58": "SUCCESS",
     "59": "NON_PARTICIPATING",
-    "6": "FAIL",
-    "60": "SUCCESS",
+    "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",
-    "67": "SUCCESS",
+    "67": "NON_PARTICIPATING",
     "68": "NON_PARTICIPATING",
     "69": "NON_PARTICIPATING",
     "7": "NON_PARTICIPATING",
-    "70": "SUCCESS",
+    "70": "NON_PARTICIPATING",
     "71": "NON_PARTICIPATING",
     "72": "NON_PARTICIPATING",
     "73": "NON_PARTICIPATING",
     "74": "NON_PARTICIPATING",
-    "75": "SUCCESS",
+    "75": "NON_PARTICIPATING",
     "76": "NON_PARTICIPATING",
     "77": "NON_PARTICIPATING",
     "78": "SUCCESS",
     "79": "NON_PARTICIPATING",
     "8": "NON_PARTICIPATING",
-    "80": "SUCCESS",
+    "80": "NON_PARTICIPATING",
     "81": "NON_PARTICIPATING",
-    "82": "SUCCESS",
+    "82": "NON_PARTICIPATING",
     "83": "NON_PARTICIPATING",
-    "84": "SUCCESS",
-    "85": "FAIL",
+    "84": "NON_PARTICIPATING",
+    "85": "SUCCESS",
     "86": "NON_PARTICIPATING",
-    "87": "SUCCESS",
+    "87": "NON_PARTICIPATING",
     "88": "SUCCESS",
-    "89": "SUCCESS",
+    "89": "NON_PARTICIPATING",
     "9": "NON_PARTICIPATING",
-    "90": "SUCCESS",
-    "91": "SUCCESS",
-    "92": "NON_PARTICIPATING",
+    "90": "NON_PARTICIPATING",
+    "91": "NON_PARTICIPATING",
+    "92": "SUCCESS",
     "93": "NON_PARTICIPATING",
     "94": "NON_PARTICIPATING",
-    "95": "SUCCESS",
-    "96": "SUCCESS",
+    "95": "NON_PARTICIPATING",
+    "96": "NON_PARTICIPATING",
     "97": "NON_PARTICIPATING",
     "98": "NON_PARTICIPATING",
-    "99": "SUCCESS"
+    "99": "NON_PARTICIPATING"
   },
   "architectures": [
     "GPTOptim"
@@ -268,22 +268,20 @@
     "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
   },
   "block_list": [
-    5504293,
-    5504296,
-    5504300,
-    5504304,
-    5504308,
-    5504312,
-    5504316
+    5508538,
+    5508542,
+    5508546,
+    5508551,
+    5508555
   ],
   "block_size": 1024,
   "bos_token_id": 50256,
   "embd_pdrop": 0.1,
   "eos_token_id": 50256,
   "initializer_range": 0.02,
-  "inner_step": 14,
+  "inner_step": 0,
   "inner_steps": 0,
-  "last_allreduce_block": 5501742,
+  "last_allreduce_block": 5506953,
   "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:f2dd70458ffee58a9bbafac1a42b89c5c04ec9538001116b05fdcd988536e4ff
+oid sha256:0d35779c9f5dbd0c6ddb93abd8dfd0286b96d68d9c25cc7d0fe47f3b1c2bae58
 size 8081782026

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:85199c090c371987f124c8377dedef0e341e0a8be6eebc65c0e69dd2ac8d5425
+oid sha256:ec321ec87ff24fbee5e185d8508c22a34151ace9e6888abf32c75a2e877d9e03
 size 4040701744