Run 4. Outer Step 74. Inner Step 1.

config.json

@@ -3,7 +3,7 @@
   "activation_function": "gelu_new",
   "all_reduce_scores": {
     "0": "NON_PARTICIPATING",
-    "1": "NON_PARTICIPATING",
+    "1": "SUCCESS",
     "10": "NON_PARTICIPATING",
     "100": "NON_PARTICIPATING",
     "101": "NON_PARTICIPATING",
@@ -12,12 +12,12 @@
     "104": "NON_PARTICIPATING",
     "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",
+    "111": "SUCCESS",
     "112": "NON_PARTICIPATING",
     "113": "NON_PARTICIPATING",
     "114": "NON_PARTICIPATING",
@@ -26,23 +26,23 @@
     "117": "NON_PARTICIPATING",
     "118": "NON_PARTICIPATING",
     "119": "NON_PARTICIPATING",
-    "12": "FAIL",
+    "12": "SUCCESS",
     "120": "NON_PARTICIPATING",
     "121": "NON_PARTICIPATING",
     "122": "NON_PARTICIPATING",
     "123": "NON_PARTICIPATING",
     "124": "NON_PARTICIPATING",
-    "125": "NON_PARTICIPATING",
-    "126": "NON_PARTICIPATING",
+    "125": "SUCCESS",
+    "126": "SUCCESS",
     "127": "NON_PARTICIPATING",
     "128": "NON_PARTICIPATING",
-    "129": "NON_PARTICIPATING",
+    "129": "SUCCESS",
     "13": "SUCCESS",
     "130": "NON_PARTICIPATING",
     "131": "NON_PARTICIPATING",
     "132": "NON_PARTICIPATING",
     "133": "NON_PARTICIPATING",
-    "134": "SUCCESS",
+    "134": "NON_PARTICIPATING",
     "135": "NON_PARTICIPATING",
     "136": "NON_PARTICIPATING",
     "137": "NON_PARTICIPATING",
@@ -62,10 +62,10 @@
     "15": "SUCCESS",
     "150": "NON_PARTICIPATING",
     "151": "NON_PARTICIPATING",
-    "152": "SUCCESS",
-    "153": "SUCCESS",
+    "152": "NON_PARTICIPATING",
+    "153": "NON_PARTICIPATING",
     "154": "NON_PARTICIPATING",
-    "155": "SUCCESS",
+    "155": "NON_PARTICIPATING",
     "156": "NON_PARTICIPATING",
     "157": "NON_PARTICIPATING",
     "158": "NON_PARTICIPATING",
@@ -74,40 +74,40 @@
     "160": "NON_PARTICIPATING",
     "161": "NON_PARTICIPATING",
     "162": "NON_PARTICIPATING",
-    "163": "NON_PARTICIPATING",
+    "163": "SUCCESS",
     "164": "NON_PARTICIPATING",
     "165": "NON_PARTICIPATING",
-    "166": "FAIL",
+    "166": "NON_PARTICIPATING",
     "167": "NON_PARTICIPATING",
     "168": "NON_PARTICIPATING",
     "169": "SUCCESS",
-    "17": "NON_PARTICIPATING",
+    "17": "SUCCESS",
     "170": "NON_PARTICIPATING",
-    "171": "NON_PARTICIPATING",
+    "171": "SUCCESS",
     "172": "NON_PARTICIPATING",
     "173": "NON_PARTICIPATING",
     "174": "NON_PARTICIPATING",
     "175": "NON_PARTICIPATING",
     "176": "NON_PARTICIPATING",
-    "177": "SUCCESS",
+    "177": "NON_PARTICIPATING",
     "178": "NON_PARTICIPATING",
     "179": "NON_PARTICIPATING",
     "18": "SUCCESS",
     "180": "NON_PARTICIPATING",
     "181": "NON_PARTICIPATING",
     "182": "NON_PARTICIPATING",
-    "183": "NON_PARTICIPATING",
+    "183": "SUCCESS",
     "184": "NON_PARTICIPATING",
     "185": "NON_PARTICIPATING",
     "186": "NON_PARTICIPATING",
     "187": "NON_PARTICIPATING",
     "188": "NON_PARTICIPATING",
     "189": "NON_PARTICIPATING",
-    "19": "NON_PARTICIPATING",
+    "19": "SUCCESS",
     "190": "NON_PARTICIPATING",
     "191": "NON_PARTICIPATING",
     "192": "NON_PARTICIPATING",
-    "193": "NON_PARTICIPATING",
+    "193": "SUCCESS",
     "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
     "196": "NON_PARTICIPATING",
@@ -134,7 +134,7 @@
     "214": "NON_PARTICIPATING",
     "215": "NON_PARTICIPATING",
     "216": "NON_PARTICIPATING",
-    "217": "NON_PARTICIPATING",
+    "217": "SUCCESS",
     "218": "NON_PARTICIPATING",
     "219": "NON_PARTICIPATING",
     "22": "SUCCESS",
@@ -148,10 +148,10 @@
     "227": "NON_PARTICIPATING",
     "228": "NON_PARTICIPATING",
     "229": "NON_PARTICIPATING",
-    "23": "NON_PARTICIPATING",
+    "23": "SUCCESS",
     "230": "NON_PARTICIPATING",
     "231": "NON_PARTICIPATING",
-    "232": "SUCCESS",
+    "232": "NON_PARTICIPATING",
     "233": "NON_PARTICIPATING",
     "234": "NON_PARTICIPATING",
     "235": "NON_PARTICIPATING",
@@ -159,19 +159,19 @@
     "237": "NON_PARTICIPATING",
     "238": "NON_PARTICIPATING",
     "239": "NON_PARTICIPATING",
-    "24": "NON_PARTICIPATING",
-    "240": "FAIL",
+    "24": "SUCCESS",
+    "240": "NON_PARTICIPATING",
     "241": "NON_PARTICIPATING",
     "242": "NON_PARTICIPATING",
-    "243": "SUCCESS",
+    "243": "NON_PARTICIPATING",
     "244": "NON_PARTICIPATING",
     "245": "NON_PARTICIPATING",
     "246": "NON_PARTICIPATING",
     "247": "NON_PARTICIPATING",
-    "248": "FAIL",
+    "248": "NON_PARTICIPATING",
     "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
-    "250": "NON_PARTICIPATING",
+    "250": "SUCCESS",
     "251": "NON_PARTICIPATING",
     "252": "NON_PARTICIPATING",
     "253": "NON_PARTICIPATING",
@@ -190,45 +190,45 @@
     "35": "SUCCESS",
     "36": "SUCCESS",
     "37": "SUCCESS",
-    "38": "NON_PARTICIPATING",
+    "38": "SUCCESS",
     "39": "SUCCESS",
     "4": "SUCCESS",
     "40": "SUCCESS",
-    "41": "NON_PARTICIPATING",
+    "41": "SUCCESS",
     "42": "SUCCESS",
     "43": "SUCCESS",
     "44": "SUCCESS",
     "45": "SUCCESS",
     "46": "NON_PARTICIPATING",
-    "47": "NON_PARTICIPATING",
+    "47": "SUCCESS",
     "48": "NON_PARTICIPATING",
     "49": "SUCCESS",
     "5": "SUCCESS",
-    "50": "NON_PARTICIPATING",
+    "50": "SUCCESS",
     "51": "SUCCESS",
     "52": "SUCCESS",
     "53": "SUCCESS",
     "54": "SUCCESS",
-    "55": "NON_PARTICIPATING",
+    "55": "SUCCESS",
     "56": "SUCCESS",
     "57": "NON_PARTICIPATING",
     "58": "NON_PARTICIPATING",
     "59": "NON_PARTICIPATING",
     "6": "SUCCESS",
     "60": "SUCCESS",
-    "61": "NON_PARTICIPATING",
+    "61": "SUCCESS",
     "62": "NON_PARTICIPATING",
     "63": "SUCCESS",
     "64": "SUCCESS",
-    "65": "NON_PARTICIPATING",
+    "65": "SUCCESS",
     "66": "SUCCESS",
     "67": "SUCCESS",
     "68": "SUCCESS",
     "69": "SUCCESS",
     "7": "SUCCESS",
-    "70": "NON_PARTICIPATING",
+    "70": "SUCCESS",
     "71": "SUCCESS",
-    "72": "NON_PARTICIPATING",
+    "72": "SUCCESS",
     "73": "SUCCESS",
     "74": "SUCCESS",
     "75": "SUCCESS",
@@ -237,27 +237,27 @@
     "78": "SUCCESS",
     "79": "NON_PARTICIPATING",
     "8": "NON_PARTICIPATING",
-    "80": "NON_PARTICIPATING",
+    "80": "SUCCESS",
     "81": "SUCCESS",
     "82": "SUCCESS",
-    "83": "NON_PARTICIPATING",
+    "83": "SUCCESS",
     "84": "SUCCESS",
     "85": "NON_PARTICIPATING",
     "86": "NON_PARTICIPATING",
     "87": "SUCCESS",
     "88": "SUCCESS",
-    "89": "NON_PARTICIPATING",
+    "89": "SUCCESS",
     "9": "SUCCESS",
-    "90": "NON_PARTICIPATING",
+    "90": "SUCCESS",
     "91": "SUCCESS",
     "92": "SUCCESS",
     "93": "NON_PARTICIPATING",
     "94": "SUCCESS",
-    "95": "NON_PARTICIPATING",
+    "95": "SUCCESS",
     "96": "SUCCESS",
-    "97": "NON_PARTICIPATING",
+    "97": "SUCCESS",
     "98": "SUCCESS",
-    "99": "NON_PARTICIPATING"
+    "99": "SUCCESS"
   },
   "architectures": [
     "GPTOptim"
@@ -268,17 +268,17 @@
     "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
   },
   "block_list": [
-    5661224,
-    5661244
+    5661280,
+    5661320
   ],
   "block_size": 1024,
   "bos_token_id": 50256,
   "embd_pdrop": 0.1,
   "eos_token_id": 50256,
   "initializer_range": 0.02,
-  "inner_step": 39,
+  "inner_step": 1,
   "inner_steps": 0,
-  "last_allreduce_block": 5659907,
+  "last_allreduce_block": 5661125,
   "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 sha256:42c9f9e1d0e86cadddfb82fc73008578fc41a4bf474c1f9e57cbf52b0dee80be
+oid sha256:bb07584cac867ecf3e8b177b3965429d7a5ee51040bfa5154f80437b99ad7f92
 size 8081782026

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fe38b91a709acae0c1111352dae3858d91b00c6924734963e0c7b3e37fd4e034
+oid sha256:212e39fd34cc1abd0101a62935a727d51dd4a02640ee3eb53f498647a337bfe0
 size 4040701744