Run 4. Outer Step 35. Inner Step 0.

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "philip324mark/mark4",
+  "_name_or_path": "distributed/optimized-gpt2-1b",
   "activation_function": "gelu_new",
   "all_reduce_scores": {
     "0": "NON_PARTICIPATING",
@@ -12,18 +12,18 @@
     "104": "NON_PARTICIPATING",
     "105": "NON_PARTICIPATING",
     "106": "NON_PARTICIPATING",
-    "107": "SUCCESS",
+    "107": "NON_PARTICIPATING",
     "108": "NON_PARTICIPATING",
-    "109": "SUCCESS",
+    "109": "NON_PARTICIPATING",
     "11": "NON_PARTICIPATING",
     "110": "NON_PARTICIPATING",
-    "111": "NON_PARTICIPATING",
+    "111": "SUCCESS",
     "112": "NON_PARTICIPATING",
     "113": "SUCCESS",
-    "114": "NON_PARTICIPATING",
+    "114": "SUCCESS",
     "115": "NON_PARTICIPATING",
     "116": "NON_PARTICIPATING",
-    "117": "NON_PARTICIPATING",
+    "117": "SUCCESS",
     "118": "NON_PARTICIPATING",
     "119": "NON_PARTICIPATING",
     "12": "SUCCESS",
@@ -34,17 +34,17 @@
     "124": "NON_PARTICIPATING",
     "125": "NON_PARTICIPATING",
     "126": "NON_PARTICIPATING",
-    "127": "SUCCESS",
+    "127": "NON_PARTICIPATING",
     "128": "NON_PARTICIPATING",
     "129": "NON_PARTICIPATING",
     "13": "SUCCESS",
-    "130": "SUCCESS",
+    "130": "NON_PARTICIPATING",
     "131": "NON_PARTICIPATING",
     "132": "NON_PARTICIPATING",
     "133": "NON_PARTICIPATING",
     "134": "NON_PARTICIPATING",
     "135": "NON_PARTICIPATING",
-    "136": "NON_PARTICIPATING",
+    "136": "SUCCESS",
     "137": "NON_PARTICIPATING",
     "138": "NON_PARTICIPATING",
     "139": "NON_PARTICIPATING",
@@ -57,16 +57,16 @@
     "145": "NON_PARTICIPATING",
     "146": "NON_PARTICIPATING",
     "147": "NON_PARTICIPATING",
-    "148": "SUCCESS",
-    "149": "NON_PARTICIPATING",
+    "148": "NON_PARTICIPATING",
+    "149": "SUCCESS",
     "15": "SUCCESS",
-    "150": "SUCCESS",
+    "150": "NON_PARTICIPATING",
     "151": "NON_PARTICIPATING",
     "152": "NON_PARTICIPATING",
-    "153": "NON_PARTICIPATING",
+    "153": "SUCCESS",
     "154": "NON_PARTICIPATING",
-    "155": "NON_PARTICIPATING",
-    "156": "NON_PARTICIPATING",
+    "155": "SUCCESS",
+    "156": "SUCCESS",
     "157": "NON_PARTICIPATING",
     "158": "NON_PARTICIPATING",
     "159": "NON_PARTICIPATING",
@@ -76,20 +76,20 @@
     "162": "NON_PARTICIPATING",
     "163": "NON_PARTICIPATING",
     "164": "NON_PARTICIPATING",
-    "165": "SUCCESS",
+    "165": "NON_PARTICIPATING",
     "166": "NON_PARTICIPATING",
     "167": "NON_PARTICIPATING",
-    "168": "NON_PARTICIPATING",
-    "169": "SUCCESS",
+    "168": "SUCCESS",
+    "169": "NON_PARTICIPATING",
     "17": "SUCCESS",
     "170": "NON_PARTICIPATING",
-    "171": "SUCCESS",
+    "171": "NON_PARTICIPATING",
     "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",
@@ -97,19 +97,19 @@
     "181": "NON_PARTICIPATING",
     "182": "NON_PARTICIPATING",
     "183": "NON_PARTICIPATING",
-    "184": "NON_PARTICIPATING",
+    "184": "SUCCESS",
     "185": "NON_PARTICIPATING",
-    "186": "NON_PARTICIPATING",
+    "186": "SUCCESS",
     "187": "SUCCESS",
     "188": "NON_PARTICIPATING",
-    "189": "SUCCESS",
-    "19": "NON_PARTICIPATING",
-    "190": "SUCCESS",
+    "189": "NON_PARTICIPATING",
+    "19": "SUCCESS",
+    "190": "NON_PARTICIPATING",
     "191": "NON_PARTICIPATING",
-    "192": "SUCCESS",
+    "192": "NON_PARTICIPATING",
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
-    "195": "SUCCESS",
+    "195": "NON_PARTICIPATING",
     "196": "NON_PARTICIPATING",
     "197": "NON_PARTICIPATING",
     "198": "NON_PARTICIPATING",
@@ -121,69 +121,69 @@
     "202": "NON_PARTICIPATING",
     "203": "SUCCESS",
     "204": "NON_PARTICIPATING",
-    "205": "FAIL",
+    "205": "NON_PARTICIPATING",
     "206": "NON_PARTICIPATING",
     "207": "NON_PARTICIPATING",
     "208": "NON_PARTICIPATING",
     "209": "NON_PARTICIPATING",
     "21": "SUCCESS",
     "210": "NON_PARTICIPATING",
-    "211": "SUCCESS",
+    "211": "NON_PARTICIPATING",
     "212": "NON_PARTICIPATING",
     "213": "NON_PARTICIPATING",
-    "214": "NON_PARTICIPATING",
+    "214": "SUCCESS",
     "215": "NON_PARTICIPATING",
     "216": "NON_PARTICIPATING",
     "217": "NON_PARTICIPATING",
     "218": "NON_PARTICIPATING",
-    "219": "NON_PARTICIPATING",
+    "219": "SUCCESS",
     "22": "SUCCESS",
     "220": "NON_PARTICIPATING",
     "221": "NON_PARTICIPATING",
     "222": "NON_PARTICIPATING",
-    "223": "SUCCESS",
+    "223": "NON_PARTICIPATING",
     "224": "NON_PARTICIPATING",
     "225": "NON_PARTICIPATING",
     "226": "NON_PARTICIPATING",
     "227": "NON_PARTICIPATING",
     "228": "NON_PARTICIPATING",
-    "229": "SUCCESS",
-    "23": "NON_PARTICIPATING",
+    "229": "NON_PARTICIPATING",
+    "23": "SUCCESS",
     "230": "NON_PARTICIPATING",
-    "231": "NON_PARTICIPATING",
+    "231": "SUCCESS",
     "232": "NON_PARTICIPATING",
     "233": "NON_PARTICIPATING",
     "234": "NON_PARTICIPATING",
-    "235": "NON_PARTICIPATING",
+    "235": "SUCCESS",
     "236": "NON_PARTICIPATING",
-    "237": "NON_PARTICIPATING",
+    "237": "SUCCESS",
     "238": "NON_PARTICIPATING",
     "239": "NON_PARTICIPATING",
     "24": "SUCCESS",
-    "240": "NON_PARTICIPATING",
+    "240": "SUCCESS",
     "241": "NON_PARTICIPATING",
-    "242": "SUCCESS",
+    "242": "NON_PARTICIPATING",
     "243": "NON_PARTICIPATING",
     "244": "NON_PARTICIPATING",
     "245": "NON_PARTICIPATING",
     "246": "NON_PARTICIPATING",
     "247": "NON_PARTICIPATING",
     "248": "NON_PARTICIPATING",
-    "249": "SUCCESS",
+    "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
     "250": "NON_PARTICIPATING",
     "251": "NON_PARTICIPATING",
-    "252": "SUCCESS",
+    "252": "NON_PARTICIPATING",
     "253": "NON_PARTICIPATING",
     "254": "NON_PARTICIPATING",
-    "255": "SUCCESS",
+    "255": "NON_PARTICIPATING",
     "26": "SUCCESS",
     "27": "SUCCESS",
     "28": "SUCCESS",
     "29": "SUCCESS",
     "3": "SUCCESS",
-    "30": "NON_PARTICIPATING",
-    "31": "SUCCESS",
+    "30": "SUCCESS",
+    "31": "NON_PARTICIPATING",
     "32": "SUCCESS",
     "33": "NON_PARTICIPATING",
     "34": "NON_PARTICIPATING",
@@ -195,7 +195,7 @@
     "4": "SUCCESS",
     "40": "SUCCESS",
     "41": "SUCCESS",
-    "42": "NON_PARTICIPATING",
+    "42": "SUCCESS",
     "43": "SUCCESS",
     "44": "SUCCESS",
     "45": "SUCCESS",
@@ -203,7 +203,7 @@
     "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
     "49": "SUCCESS",
-    "5": "NON_PARTICIPATING",
+    "5": "SUCCESS",
     "50": "SUCCESS",
     "51": "SUCCESS",
     "52": "SUCCESS",
@@ -211,7 +211,7 @@
     "54": "SUCCESS",
     "55": "SUCCESS",
     "56": "SUCCESS",
-    "57": "FAIL",
+    "57": "SUCCESS",
     "58": "NON_PARTICIPATING",
     "59": "SUCCESS",
     "6": "NON_PARTICIPATING",
@@ -224,7 +224,7 @@
     "66": "SUCCESS",
     "67": "SUCCESS",
     "68": "NON_PARTICIPATING",
-    "69": "NON_PARTICIPATING",
+    "69": "SUCCESS",
     "7": "SUCCESS",
     "70": "NON_PARTICIPATING",
     "71": "NON_PARTICIPATING",
@@ -234,7 +234,7 @@
     "75": "SUCCESS",
     "76": "SUCCESS",
     "77": "SUCCESS",
-    "78": "NON_PARTICIPATING",
+    "78": "SUCCESS",
     "79": "NON_PARTICIPATING",
     "8": "SUCCESS",
     "80": "SUCCESS",
@@ -242,8 +242,8 @@
     "82": "SUCCESS",
     "83": "NON_PARTICIPATING",
     "84": "SUCCESS",
-    "85": "SUCCESS",
-    "86": "NON_PARTICIPATING",
+    "85": "NON_PARTICIPATING",
+    "86": "SUCCESS",
     "87": "SUCCESS",
     "88": "SUCCESS",
     "89": "SUCCESS",
@@ -268,20 +268,20 @@
     "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
   },
   "block_list": [
-    5546924,
-    5546929,
-    5546933,
-    5546938,
-    5546942
+    5549400,
+    5549406,
+    5549406,
+    5549411,
+    5549417
   ],
   "block_size": 1024,
   "bos_token_id": 50256,
   "embd_pdrop": 0.1,
   "eos_token_id": 50256,
   "initializer_range": 0.02,
-  "inner_step": 4,
+  "inner_step": 0,
   "inner_steps": 0,
-  "last_allreduce_block": 5543976,
+  "last_allreduce_block": 5548219,
   "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:a6b6aee75e382e0f69fd316a52b8483b00bbd9bff18da5640ad17ca366dc7921
+oid sha256:a8eb6198ee9f127a37d57148b80a94c2e1dee758dc8d33b930aebf0efd6d3bbb
 size 8081782026

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b71b6afd72a3158e94aa2da143484e165627fd4ac15c1bb86be09d61d4ace446
+oid sha256:353a0ee2e8a46be74f32e78f97392ff51faf4dc534a55f55500268a7022296fb
 size 4040701744