Run 11. Outer Step 224. Inner Step 0.

config.json

@@ -264,8 +264,8 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "block_list": [
-    6668316,
-    6668352
+    6668383,
+    6668420
   ],
   "bos_token_id": 1,
   "dtype": "float32",
@@ -274,7 +274,7 @@
   "hidden_act": "silu",
   "hidden_size": 2048,
   "initializer_range": 0.02,
-  "inner_step": 40,
+  "inner_step": 0,
   "intermediate_size": 5632,
   "last_allreduce_block": 6668039,
   "max_position_embeddings": 2048,

gradients.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a8e31fbd0c9e5f4867e222ea33b6df2ff8a353e4d2b06b15794074f543aea2f7
+oid sha256:37b6e4f3c112969c89c20ed08a44dddd797226bfde4252adda8aa559c3eb8afa
 size 88389626

inner_optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d9bd7b2b181347c3d7f35cfec0abb3dffca9449b437f3d57867cb337339a4e6
+oid sha256:b3e348e7fa25c2d30196605f4ce2d4e10e6baecbec950b4c504a0fe4daad8108
 size 4924926906

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:431e056b67c1edb87b7164e1b478c6604da7881585e1dfb4b8b07ce57631b821
+oid sha256:178f0ff0f6fe7141e5c33cdb84c4b6271289cef3a6dbf11c69825ef315e6a13f
 size 4400216536