Outer Step 9. Inner Step 0. Batch Size 0

config.json

@@ -13,7 +13,7 @@
     "105": "NON_PARTICIPATING",
     "106": "NON_PARTICIPATING",
     "107": "NON_PARTICIPATING",
-    "108": "SUCCESS",
+    "108": "NON_PARTICIPATING",
     "109": "NON_PARTICIPATING",
     "11": "NON_PARTICIPATING",
     "110": "NON_PARTICIPATING",
@@ -23,12 +23,12 @@
     "114": "NON_PARTICIPATING",
     "115": "NON_PARTICIPATING",
     "116": "NON_PARTICIPATING",
-    "117": "NON_PARTICIPATING",
+    "117": "SUCCESS",
     "118": "NON_PARTICIPATING",
     "119": "NON_PARTICIPATING",
-    "12": "SUCCESS",
+    "12": "NON_PARTICIPATING",
     "120": "NON_PARTICIPATING",
-    "121": "SUCCESS",
+    "121": "NON_PARTICIPATING",
     "122": "NON_PARTICIPATING",
     "123": "NON_PARTICIPATING",
     "124": "NON_PARTICIPATING",
@@ -42,7 +42,7 @@
     "131": "NON_PARTICIPATING",
     "132": "NON_PARTICIPATING",
     "133": "NON_PARTICIPATING",
-    "134": "SUCCESS",
+    "134": "NON_PARTICIPATING",
     "135": "NON_PARTICIPATING",
     "136": "NON_PARTICIPATING",
     "137": "NON_PARTICIPATING",
@@ -56,7 +56,7 @@
     "144": "NON_PARTICIPATING",
     "145": "NON_PARTICIPATING",
     "146": "NON_PARTICIPATING",
-    "147": "SUCCESS",
+    "147": "NON_PARTICIPATING",
     "148": "NON_PARTICIPATING",
     "149": "NON_PARTICIPATING",
     "15": "NON_PARTICIPATING",
@@ -100,23 +100,23 @@
     "184": "NON_PARTICIPATING",
     "185": "NON_PARTICIPATING",
     "186": "NON_PARTICIPATING",
-    "187": "SUCCESS",
+    "187": "NON_PARTICIPATING",
     "188": "NON_PARTICIPATING",
     "189": "NON_PARTICIPATING",
     "19": "NON_PARTICIPATING",
     "190": "NON_PARTICIPATING",
     "191": "NON_PARTICIPATING",
-    "192": "NON_PARTICIPATING",
+    "192": "SUCCESS",
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
     "195": "NON_PARTICIPATING",
-    "196": "SUCCESS",
+    "196": "NON_PARTICIPATING",
     "197": "NON_PARTICIPATING",
     "198": "NON_PARTICIPATING",
     "199": "NON_PARTICIPATING",
     "2": "NON_PARTICIPATING",
     "20": "NON_PARTICIPATING",
-    "200": "NON_PARTICIPATING",
+    "200": "SUCCESS",
     "201": "NON_PARTICIPATING",
     "202": "NON_PARTICIPATING",
     "203": "NON_PARTICIPATING",
@@ -130,14 +130,14 @@
     "210": "NON_PARTICIPATING",
     "211": "NON_PARTICIPATING",
     "212": "NON_PARTICIPATING",
-    "213": "SUCCESS",
+    "213": "NON_PARTICIPATING",
     "214": "NON_PARTICIPATING",
     "215": "NON_PARTICIPATING",
     "216": "NON_PARTICIPATING",
     "217": "NON_PARTICIPATING",
     "218": "NON_PARTICIPATING",
     "219": "NON_PARTICIPATING",
-    "22": "NON_PARTICIPATING",
+    "22": "SUCCESS",
     "220": "NON_PARTICIPATING",
     "221": "NON_PARTICIPATING",
     "222": "NON_PARTICIPATING",
@@ -148,20 +148,20 @@
     "227": "NON_PARTICIPATING",
     "228": "NON_PARTICIPATING",
     "229": "NON_PARTICIPATING",
-    "23": "NON_PARTICIPATING",
+    "23": "FAIL",
     "230": "NON_PARTICIPATING",
     "231": "NON_PARTICIPATING",
-    "232": "SUCCESS",
+    "232": "NON_PARTICIPATING",
     "233": "NON_PARTICIPATING",
-    "234": "SUCCESS",
-    "235": "SUCCESS",
+    "234": "NON_PARTICIPATING",
+    "235": "NON_PARTICIPATING",
     "236": "NON_PARTICIPATING",
-    "237": "SUCCESS",
+    "237": "NON_PARTICIPATING",
     "238": "NON_PARTICIPATING",
     "239": "NON_PARTICIPATING",
     "24": "NON_PARTICIPATING",
     "240": "NON_PARTICIPATING",
-    "241": "SUCCESS",
+    "241": "NON_PARTICIPATING",
     "242": "NON_PARTICIPATING",
     "243": "NON_PARTICIPATING",
     "244": "NON_PARTICIPATING",
@@ -169,12 +169,12 @@
     "246": "NON_PARTICIPATING",
     "247": "NON_PARTICIPATING",
     "248": "NON_PARTICIPATING",
-    "249": "NON_PARTICIPATING",
+    "249": "SUCCESS",
     "25": "SUCCESS",
     "250": "NON_PARTICIPATING",
     "251": "NON_PARTICIPATING",
     "252": "NON_PARTICIPATING",
-    "253": "SUCCESS",
+    "253": "NON_PARTICIPATING",
     "254": "NON_PARTICIPATING",
     "255": "NON_PARTICIPATING",
     "26": "NON_PARTICIPATING",
@@ -200,31 +200,31 @@
     "44": "NON_PARTICIPATING",
     "45": "NON_PARTICIPATING",
     "46": "NON_PARTICIPATING",
-    "47": "SUCCESS",
+    "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
     "49": "NON_PARTICIPATING",
     "5": "NON_PARTICIPATING",
     "50": "NON_PARTICIPATING",
-    "51": "SUCCESS",
+    "51": "NON_PARTICIPATING",
     "52": "NON_PARTICIPATING",
     "53": "NON_PARTICIPATING",
     "54": "NON_PARTICIPATING",
     "55": "NON_PARTICIPATING",
     "56": "NON_PARTICIPATING",
     "57": "NON_PARTICIPATING",
-    "58": "NON_PARTICIPATING",
+    "58": "SUCCESS",
     "59": "NON_PARTICIPATING",
     "6": "NON_PARTICIPATING",
     "60": "NON_PARTICIPATING",
     "61": "NON_PARTICIPATING",
-    "62": "NON_PARTICIPATING",
+    "62": "SUCCESS",
     "63": "NON_PARTICIPATING",
     "64": "NON_PARTICIPATING",
-    "65": "SUCCESS",
+    "65": "NON_PARTICIPATING",
     "66": "NON_PARTICIPATING",
     "67": "SUCCESS",
     "68": "NON_PARTICIPATING",
-    "69": "SUCCESS",
+    "69": "NON_PARTICIPATING",
     "7": "NON_PARTICIPATING",
     "70": "NON_PARTICIPATING",
     "71": "NON_PARTICIPATING",
@@ -248,12 +248,12 @@
     "88": "NON_PARTICIPATING",
     "89": "NON_PARTICIPATING",
     "9": "NON_PARTICIPATING",
-    "90": "SUCCESS",
+    "90": "NON_PARTICIPATING",
     "91": "NON_PARTICIPATING",
     "92": "NON_PARTICIPATING",
     "93": "NON_PARTICIPATING",
     "94": "NON_PARTICIPATING",
-    "95": "SUCCESS",
+    "95": "NON_PARTICIPATING",
     "96": "NON_PARTICIPATING",
     "97": "NON_PARTICIPATING",
     "98": "NON_PARTICIPATING",
@@ -267,26 +267,15 @@
     "AutoConfig": "distributed/optimized-gpt2-500m--configuration_gpt_optimized.GPTOptimConfig",
     "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
   },
-  "block_list": [
-    5142433,
-    5142443,
-    5142450,
-    5142457,
-    5142464,
-    5142471,
-    5142476,
-    5142482,
-    5142488,
-    5142494
-  ],
+  "block_list": [],
   "block_size": 1024,
   "bos_token_id": 50256,
   "embd_pdrop": 0.1,
   "eos_token_id": 50256,
   "initializer_range": 0.02,
-  "inner_step": 207,
+  "inner_step": 0,
   "inner_steps": 0,
-  "last_allreduce_block": 5072405,
+  "last_allreduce_block": 5297756,
   "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 sha256:544ddef402271093ee7869797c9092fb41a553cdddd5fed6494faf9af027894f
-size 8081781770
+oid sha256:d241a16916cb5a41ac6f7276ebeabdd3b67a2aff7c39511e0c4239428992fb55
+size 2944

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7efec361cfbca2df859b7278f9e933f24b1c626d58a714ab80dcce80a958c392
+oid sha256:fb2d6803d7f8fa6434db62fea3b146b1f4ff23b347ced174e318c748a5efaa80
 size 4040701744