Run 5. Outer Step 51. Inner Step 2.

config.json

@@ -11,14 +11,14 @@
     "104": "NOT_ALIVE",
     "105": "SUCCESS",
     "106": "NOT_ALIVE",
-    "107": "NON_PARTICIPATING",
+    "107": "NOT_ALIVE",
     "108": "NON_PARTICIPATING",
     "109": "NON_PARTICIPATING",
     "11": "NOT_ALIVE",
     "110": "SUCCESS",
     "111": "NON_PARTICIPATING",
     "112": "NOT_ALIVE",
-    "113": "NOT_ALIVE",
+    "113": "NON_PARTICIPATING",
     "114": "NON_PARTICIPATING",
     "115": "NOT_ALIVE",
     "116": "NON_PARTICIPATING",
@@ -29,8 +29,8 @@
     "120": "NOT_ALIVE",
     "121": "NOT_ALIVE",
     "122": "SUCCESS",
-    "123": "NON_PARTICIPATING",
-    "124": "NOT_ALIVE",
+    "123": "SUCCESS",
+    "124": "SUCCESS",
     "125": "NOT_ALIVE",
     "126": "NON_PARTICIPATING",
     "127": "NON_PARTICIPATING",
@@ -41,8 +41,8 @@
     "131": "NOT_ALIVE",
     "132": "NOT_ALIVE",
     "133": "SUCCESS",
-    "134": "SUCCESS",
-    "135": "NOT_ALIVE",
+    "134": "NOT_ALIVE",
+    "135": "NON_PARTICIPATING",
     "136": "NOT_ALIVE",
     "137": "NOT_ALIVE",
     "138": "NOT_ALIVE",
@@ -50,38 +50,38 @@
     "14": "NOT_ALIVE",
     "140": "NOT_ALIVE",
     "141": "SUCCESS",
-    "142": "SUCCESS",
+    "142": "NOT_ALIVE",
     "143": "NON_PARTICIPATING",
-    "144": "NOT_ALIVE",
+    "144": "NON_PARTICIPATING",
     "145": "NOT_ALIVE",
     "146": "SUCCESS",
     "147": "NON_PARTICIPATING",
     "148": "SUCCESS",
-    "149": "NOT_ALIVE",
+    "149": "NON_PARTICIPATING",
     "15": "NON_PARTICIPATING",
     "150": "SUCCESS",
     "151": "NOT_ALIVE",
     "152": "NON_PARTICIPATING",
-    "153": "NOT_ALIVE",
+    "153": "NON_PARTICIPATING",
     "154": "NOT_ALIVE",
     "155": "SUCCESS",
-    "156": "NOT_ALIVE",
-    "157": "NON_PARTICIPATING",
+    "156": "NON_PARTICIPATING",
+    "157": "SUCCESS",
     "158": "NOT_ALIVE",
     "159": "NOT_ALIVE",
-    "16": "NOT_ALIVE",
-    "160": "NOT_ALIVE",
+    "16": "NON_PARTICIPATING",
+    "160": "NON_PARTICIPATING",
     "161": "SUCCESS",
     "162": "NON_PARTICIPATING",
     "163": "NOT_ALIVE",
     "164": "SUCCESS",
     "165": "NON_PARTICIPATING",
     "166": "NON_PARTICIPATING",
-    "167": "SUCCESS",
-    "168": "NON_PARTICIPATING",
+    "167": "NOT_ALIVE",
+    "168": "NOT_ALIVE",
     "169": "NOT_ALIVE",
     "17": "NOT_ALIVE",
-    "170": "NON_PARTICIPATING",
+    "170": "NOT_ALIVE",
     "171": "SUCCESS",
     "172": "NOT_ALIVE",
     "173": "NOT_ALIVE",
@@ -93,17 +93,17 @@
     "179": "NOT_ALIVE",
     "18": "SUCCESS",
     "180": "NOT_ALIVE",
-    "181": "NON_PARTICIPATING",
+    "181": "SUCCESS",
     "182": "NON_PARTICIPATING",
     "183": "SUCCESS",
     "184": "NOT_ALIVE",
     "185": "NON_PARTICIPATING",
     "186": "NOT_ALIVE",
     "187": "SUCCESS",
-    "188": "SUCCESS",
+    "188": "NON_PARTICIPATING",
     "189": "NOT_ALIVE",
     "19": "NOT_ALIVE",
-    "190": "NON_PARTICIPATING",
+    "190": "NOT_ALIVE",
     "191": "NOT_ALIVE",
     "192": "NON_PARTICIPATING",
     "193": "NOT_ALIVE",
@@ -114,8 +114,8 @@
     "198": "NOT_ALIVE",
     "199": "NON_PARTICIPATING",
     "2": "NOT_ALIVE",
-    "20": "NON_PARTICIPATING",
-    "200": "SUCCESS",
+    "20": "NOT_ALIVE",
+    "200": "NOT_ALIVE",
     "201": "NON_PARTICIPATING",
     "202": "NON_PARTICIPATING",
     "203": "NON_PARTICIPATING",
@@ -125,16 +125,16 @@
     "207": "NOT_ALIVE",
     "208": "NOT_ALIVE",
     "209": "SUCCESS",
-    "21": "NOT_ALIVE",
+    "21": "SUCCESS",
     "210": "SUCCESS",
-    "211": "NON_PARTICIPATING",
+    "211": "NOT_ALIVE",
     "212": "NOT_ALIVE",
     "213": "NOT_ALIVE",
     "214": "NON_PARTICIPATING",
     "215": "NON_PARTICIPATING",
     "216": "NOT_ALIVE",
     "217": "NON_PARTICIPATING",
-    "218": "NOT_ALIVE",
+    "218": "NON_PARTICIPATING",
     "219": "NOT_ALIVE",
     "22": "SUCCESS",
     "220": "SUCCESS",
@@ -145,34 +145,34 @@
     "225": "NOT_ALIVE",
     "226": "NOT_ALIVE",
     "227": "SUCCESS",
-    "228": "NOT_ALIVE",
+    "228": "NON_PARTICIPATING",
     "229": "NOT_ALIVE",
     "23": "NOT_ALIVE",
     "230": "NOT_ALIVE",
-    "231": "NOT_ALIVE",
+    "231": "NON_PARTICIPATING",
     "232": "NOT_ALIVE",
     "233": "NOT_ALIVE",
-    "234": "NON_PARTICIPATING",
+    "234": "NOT_ALIVE",
     "235": "NOT_ALIVE",
-    "236": "NON_PARTICIPATING",
+    "236": "NOT_ALIVE",
     "237": "NOT_ALIVE",
     "238": "NON_PARTICIPATING",
     "239": "NOT_ALIVE",
     "24": "SUCCESS",
     "240": "NON_PARTICIPATING",
     "241": "NOT_ALIVE",
-    "242": "NOT_ALIVE",
+    "242": "SUCCESS",
     "243": "NON_PARTICIPATING",
     "244": "NOT_ALIVE",
     "245": "NOT_ALIVE",
-    "246": "NOT_ALIVE",
+    "246": "NON_PARTICIPATING",
     "247": "SUCCESS",
     "248": "NON_PARTICIPATING",
     "249": "NON_PARTICIPATING",
     "25": "SUCCESS",
     "250": "NON_PARTICIPATING",
     "251": "NOT_ALIVE",
-    "252": "NOT_ALIVE",
+    "252": "NON_PARTICIPATING",
     "253": "NOT_ALIVE",
     "254": "NOT_ALIVE",
     "255": "NON_PARTICIPATING",
@@ -190,9 +190,9 @@
     "36": "NON_PARTICIPATING",
     "37": "NOT_ALIVE",
     "38": "SUCCESS",
-    "39": "NON_PARTICIPATING",
+    "39": "NOT_ALIVE",
     "4": "NOT_ALIVE",
-    "40": "NOT_ALIVE",
+    "40": "SUCCESS",
     "41": "SUCCESS",
     "42": "SUCCESS",
     "43": "SUCCESS",
@@ -206,27 +206,27 @@
     "50": "NON_PARTICIPATING",
     "51": "NOT_ALIVE",
     "52": "NON_PARTICIPATING",
-    "53": "NON_PARTICIPATING",
-    "54": "NOT_ALIVE",
+    "53": "NOT_ALIVE",
+    "54": "SUCCESS",
     "55": "NOT_ALIVE",
     "56": "SUCCESS",
-    "57": "SUCCESS",
+    "57": "NOT_ALIVE",
     "58": "SUCCESS",
     "59": "NOT_ALIVE",
     "6": "NOT_ALIVE",
-    "60": "SUCCESS",
-    "61": "NON_PARTICIPATING",
+    "60": "NON_PARTICIPATING",
+    "61": "NOT_ALIVE",
     "62": "NOT_ALIVE",
-    "63": "NON_PARTICIPATING",
-    "64": "NON_PARTICIPATING",
+    "63": "NOT_ALIVE",
+    "64": "NOT_ALIVE",
     "65": "NOT_ALIVE",
     "66": "SUCCESS",
     "67": "NOT_ALIVE",
     "68": "NOT_ALIVE",
-    "69": "SUCCESS",
+    "69": "NOT_ALIVE",
     "7": "SUCCESS",
     "70": "NOT_ALIVE",
-    "71": "NOT_ALIVE",
+    "71": "SUCCESS",
     "72": "NOT_ALIVE",
     "73": "SUCCESS",
     "74": "NOT_ALIVE",
@@ -240,7 +240,7 @@
     "81": "NOT_ALIVE",
     "82": "NOT_ALIVE",
     "83": "NOT_ALIVE",
-    "84": "NOT_ALIVE",
+    "84": "SUCCESS",
     "85": "NON_PARTICIPATING",
     "86": "SUCCESS",
     "87": "SUCCESS",
@@ -249,14 +249,14 @@
     "9": "NOT_ALIVE",
     "90": "SUCCESS",
     "91": "NON_PARTICIPATING",
-    "92": "NOT_ALIVE",
+    "92": "SUCCESS",
     "93": "NON_PARTICIPATING",
     "94": "NON_PARTICIPATING",
     "95": "NOT_ALIVE",
     "96": "NON_PARTICIPATING",
     "97": "SUCCESS",
     "98": "SUCCESS",
-    "99": "NOT_ALIVE"
+    "99": "NON_PARTICIPATING"
   },
   "architectures": [
     "LlamaForCausalLM"
@@ -264,17 +264,17 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "block_list": [
-    5888008,
-    5888053
+    5888160,
+    5888226
   ],
   "bos_token_id": 1,
   "eos_token_id": 2,
   "hidden_act": "silu",
   "hidden_size": 2048,
   "initializer_range": 0.02,
-  "inner_step": 30,
+  "inner_step": 2,
   "intermediate_size": 5632,
-  "last_allreduce_block": 5886845,
+  "last_allreduce_block": 5888030,
   "max_position_embeddings": 2048,
   "mlp_bias": false,
   "model_type": "llama",

inner_optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:49396439729490b1e5a799ce7a17ef36967248a3eaacf1ef148c116fb50fb813
+oid sha256:644f5ab2aa55c38e886858bfe2f8b62905b068296751c0904177603095568584
 size 8800564076

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf581cfd846689f2aea67ae12915d67b13a8f282cb21350358090c2ba4bd09f5
+oid sha256:25988f1ae9814353c93dc4d8d667684b5f31692d8e08b2ebaf2efc92fe73104e
 size 4400216536