Run 4. Outer Step 50. Inner Step 71.

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "distributed/optimized-gpt2-1b",
+  "_name_or_path": "philip324mark/mark7",
   "activation_function": "gelu_new",
   "all_reduce_scores": {
     "0": "NON_PARTICIPATING",
@@ -11,15 +11,15 @@
     "103": "NON_PARTICIPATING",
     "104": "NON_PARTICIPATING",
     "105": "NON_PARTICIPATING",
-    "106": "SUCCESS",
-    "107": "SUCCESS",
+    "106": "NON_PARTICIPATING",
+    "107": "NON_PARTICIPATING",
     "108": "NON_PARTICIPATING",
     "109": "NON_PARTICIPATING",
     "11": "SUCCESS",
     "110": "NON_PARTICIPATING",
     "111": "NON_PARTICIPATING",
     "112": "NON_PARTICIPATING",
-    "113": "SUCCESS",
+    "113": "NON_PARTICIPATING",
     "114": "NON_PARTICIPATING",
     "115": "NON_PARTICIPATING",
     "116": "NON_PARTICIPATING",
@@ -30,15 +30,15 @@
     "120": "NON_PARTICIPATING",
     "121": "NON_PARTICIPATING",
     "122": "NON_PARTICIPATING",
-    "123": "NON_PARTICIPATING",
+    "123": "SUCCESS",
     "124": "NON_PARTICIPATING",
     "125": "NON_PARTICIPATING",
-    "126": "SUCCESS",
+    "126": "NON_PARTICIPATING",
     "127": "NON_PARTICIPATING",
     "128": "NON_PARTICIPATING",
-    "129": "NON_PARTICIPATING",
-    "13": "NON_PARTICIPATING",
-    "130": "SUCCESS",
+    "129": "SUCCESS",
+    "13": "SUCCESS",
+    "130": "NON_PARTICIPATING",
     "131": "NON_PARTICIPATING",
     "132": "NON_PARTICIPATING",
     "133": "NON_PARTICIPATING",
@@ -48,26 +48,26 @@
     "137": "NON_PARTICIPATING",
     "138": "NON_PARTICIPATING",
     "139": "NON_PARTICIPATING",
-    "14": "SUCCESS",
+    "14": "NON_PARTICIPATING",
     "140": "NON_PARTICIPATING",
     "141": "NON_PARTICIPATING",
     "142": "NON_PARTICIPATING",
-    "143": "SUCCESS",
+    "143": "NON_PARTICIPATING",
     "144": "NON_PARTICIPATING",
     "145": "NON_PARTICIPATING",
     "146": "NON_PARTICIPATING",
     "147": "NON_PARTICIPATING",
     "148": "NON_PARTICIPATING",
-    "149": "NON_PARTICIPATING",
+    "149": "SUCCESS",
     "15": "SUCCESS",
-    "150": "SUCCESS",
+    "150": "NON_PARTICIPATING",
     "151": "NON_PARTICIPATING",
     "152": "NON_PARTICIPATING",
     "153": "NON_PARTICIPATING",
-    "154": "SUCCESS",
+    "154": "NON_PARTICIPATING",
     "155": "NON_PARTICIPATING",
-    "156": "NON_PARTICIPATING",
-    "157": "NON_PARTICIPATING",
+    "156": "SUCCESS",
+    "157": "SUCCESS",
     "158": "NON_PARTICIPATING",
     "159": "NON_PARTICIPATING",
     "16": "SUCCESS",
@@ -76,37 +76,37 @@
     "162": "NON_PARTICIPATING",
     "163": "NON_PARTICIPATING",
     "164": "NON_PARTICIPATING",
-    "165": "SUCCESS",
+    "165": "NON_PARTICIPATING",
     "166": "NON_PARTICIPATING",
-    "167": "NON_PARTICIPATING",
+    "167": "SUCCESS",
     "168": "NON_PARTICIPATING",
     "169": "NON_PARTICIPATING",
     "17": "SUCCESS",
-    "170": "NON_PARTICIPATING",
-    "171": "NON_PARTICIPATING",
+    "170": "SUCCESS",
+    "171": "SUCCESS",
     "172": "NON_PARTICIPATING",
     "173": "NON_PARTICIPATING",
     "174": "NON_PARTICIPATING",
-    "175": "NON_PARTICIPATING",
+    "175": "SUCCESS",
     "176": "NON_PARTICIPATING",
     "177": "NON_PARTICIPATING",
     "178": "NON_PARTICIPATING",
     "179": "NON_PARTICIPATING",
     "18": "SUCCESS",
-    "180": "NON_PARTICIPATING",
-    "181": "SUCCESS",
+    "180": "SUCCESS",
+    "181": "NON_PARTICIPATING",
     "182": "NON_PARTICIPATING",
     "183": "NON_PARTICIPATING",
-    "184": "NON_PARTICIPATING",
-    "185": "NON_PARTICIPATING",
-    "186": "SUCCESS",
+    "184": "SUCCESS",
+    "185": "SUCCESS",
+    "186": "NON_PARTICIPATING",
     "187": "NON_PARTICIPATING",
     "188": "NON_PARTICIPATING",
     "189": "NON_PARTICIPATING",
     "19": "SUCCESS",
     "190": "NON_PARTICIPATING",
-    "191": "NON_PARTICIPATING",
-    "192": "SUCCESS",
+    "191": "SUCCESS",
+    "192": "NON_PARTICIPATING",
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
@@ -119,7 +119,7 @@
     "200": "NON_PARTICIPATING",
     "201": "NON_PARTICIPATING",
     "202": "NON_PARTICIPATING",
-    "203": "NON_PARTICIPATING",
+    "203": "SUCCESS",
     "204": "NON_PARTICIPATING",
     "205": "NON_PARTICIPATING",
     "206": "NON_PARTICIPATING",
@@ -135,11 +135,11 @@
     "215": "NON_PARTICIPATING",
     "216": "NON_PARTICIPATING",
     "217": "NON_PARTICIPATING",
-    "218": "SUCCESS",
-    "219": "SUCCESS",
+    "218": "NON_PARTICIPATING",
+    "219": "NON_PARTICIPATING",
     "22": "SUCCESS",
-    "220": "SUCCESS",
-    "221": "SUCCESS",
+    "220": "NON_PARTICIPATING",
+    "221": "NON_PARTICIPATING",
     "222": "NON_PARTICIPATING",
     "223": "NON_PARTICIPATING",
     "224": "NON_PARTICIPATING",
@@ -151,39 +151,39 @@
     "23": "SUCCESS",
     "230": "NON_PARTICIPATING",
     "231": "NON_PARTICIPATING",
-    "232": "SUCCESS",
+    "232": "NON_PARTICIPATING",
     "233": "NON_PARTICIPATING",
     "234": "NON_PARTICIPATING",
     "235": "NON_PARTICIPATING",
     "236": "SUCCESS",
-    "237": "NON_PARTICIPATING",
+    "237": "SUCCESS",
     "238": "NON_PARTICIPATING",
     "239": "NON_PARTICIPATING",
     "24": "SUCCESS",
     "240": "NON_PARTICIPATING",
-    "241": "SUCCESS",
+    "241": "NON_PARTICIPATING",
     "242": "NON_PARTICIPATING",
     "243": "SUCCESS",
     "244": "NON_PARTICIPATING",
     "245": "NON_PARTICIPATING",
-    "246": "SUCCESS",
-    "247": "SUCCESS",
+    "246": "NON_PARTICIPATING",
+    "247": "NON_PARTICIPATING",
     "248": "NON_PARTICIPATING",
     "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
-    "250": "NON_PARTICIPATING",
+    "250": "SUCCESS",
     "251": "NON_PARTICIPATING",
     "252": "NON_PARTICIPATING",
     "253": "NON_PARTICIPATING",
     "254": "NON_PARTICIPATING",
-    "255": "NON_PARTICIPATING",
+    "255": "SUCCESS",
     "26": "SUCCESS",
     "27": "SUCCESS",
     "28": "SUCCESS",
     "29": "SUCCESS",
-    "3": "NON_PARTICIPATING",
+    "3": "SUCCESS",
     "30": "NON_PARTICIPATING",
-    "31": "NON_PARTICIPATING",
+    "31": "SUCCESS",
     "32": "SUCCESS",
     "33": "NON_PARTICIPATING",
     "34": "SUCCESS",
@@ -195,18 +195,18 @@
     "4": "SUCCESS",
     "40": "SUCCESS",
     "41": "SUCCESS",
-    "42": "SUCCESS",
+    "42": "NON_PARTICIPATING",
     "43": "SUCCESS",
     "44": "SUCCESS",
     "45": "SUCCESS",
     "46": "NON_PARTICIPATING",
-    "47": "NON_PARTICIPATING",
+    "47": "SUCCESS",
     "48": "NON_PARTICIPATING",
     "49": "SUCCESS",
-    "5": "SUCCESS",
+    "5": "NON_PARTICIPATING",
     "50": "SUCCESS",
     "51": "SUCCESS",
-    "52": "NON_PARTICIPATING",
+    "52": "SUCCESS",
     "53": "SUCCESS",
     "54": "SUCCESS",
     "55": "SUCCESS",
@@ -219,7 +219,7 @@
     "61": "NON_PARTICIPATING",
     "62": "SUCCESS",
     "63": "SUCCESS",
-    "64": "SUCCESS",
+    "64": "NON_PARTICIPATING",
     "65": "SUCCESS",
     "66": "SUCCESS",
     "67": "SUCCESS",
@@ -232,8 +232,8 @@
     "73": "NON_PARTICIPATING",
     "74": "SUCCESS",
     "75": "SUCCESS",
-    "76": "NON_PARTICIPATING",
-    "77": "NON_PARTICIPATING",
+    "76": "SUCCESS",
+    "77": "SUCCESS",
     "78": "SUCCESS",
     "79": "NON_PARTICIPATING",
     "8": "SUCCESS",
@@ -253,8 +253,8 @@
     "92": "SUCCESS",
     "93": "NON_PARTICIPATING",
     "94": "SUCCESS",
-    "95": "NON_PARTICIPATING",
-    "96": "NON_PARTICIPATING",
+    "95": "SUCCESS",
+    "96": "SUCCESS",
     "97": "SUCCESS",
     "98": "SUCCESS",
     "99": "SUCCESS"
@@ -268,20 +268,17 @@
     "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
   },
   "block_list": [
-    5622041,
-    5622045,
-    5622049,
-    5622053,
-    5622057
+    5622191,
+    5622220
   ],
   "block_size": 1024,
   "bos_token_id": 50256,
   "embd_pdrop": 0.1,
   "eos_token_id": 50256,
   "initializer_range": 0.02,
-  "inner_step": 102,
+  "inner_step": 71,
   "inner_steps": 0,
-  "last_allreduce_block": 5611658,
+  "last_allreduce_block": 5568918,
   "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 sha256:050401fc625d26eaeaef3a8a01d6ba85896a83266c8681686466cf90659b3846
-size 8081782026
+oid sha256:21ed344ae7ddcf3910c4074eff5c2f0524375ea578e852f02be67f74d54de181
+size 8081782503

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2e0577dbe106e6c42e2046fd6501c62aaa6badd05587163f818dec0a4832c8a7
+oid sha256:21d2b8e80f38746134120719cfdff47ddd24036b3b07e520bdf6d7c2be6e4065
 size 4040701744