correct checkpoints see: https://github.com/facebookresearch/metaseq/pull/164

config.json

@@ -1,9 +1,10 @@
 {
   "_name_or_path": "facebook/opt-13b",
+  "_remove_final_layer_norm": false,
   "activation_dropout": 0.0,
   "activation_function": "relu",
   "architectures": [
-    "OPTForCausalLM"
+    "OPTModel"
   ],
   "attention_dropout": 0.0,
   "bos_token_id": 2,
@@ -22,7 +23,7 @@
   "pad_token_id": 1,
   "prefix": "</s>",
   "torch_dtype": "float16",
-  "transformers_version": "4.20.0.dev0",
+  "transformers_version": "4.21.0.dev0",
   "use_cache": true,
   "vocab_size": 50272,
   "word_embed_proj_dim": 5120

pytorch_model-00001-of-00003.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17228bac78fbe741f51a0a1b7b57f97b14c39d87a764515a927d4901d87505b3
-size 9975035273
+oid sha256:7f92dff6b6debf90135157bfe3663ea87f5dc17fef486e8981259692df3e284c
+size 9975056349

pytorch_model.bin.index.json

@@ -1,10 +1,12 @@
 {
   "metadata": {
-    "total_size": 25706926080
+    "total_size": 25706946560
   },
   "weight_map": {
     "decoder.embed_positions.weight": "pytorch_model-00001-of-00003.bin",
     "decoder.embed_tokens.weight": "pytorch_model-00001-of-00003.bin",
+    "decoder.final_layer_norm.bias": "pytorch_model-00001-of-00003.bin",
+    "decoder.final_layer_norm.weight": "pytorch_model-00001-of-00003.bin",
     "decoder.layers.0.fc1.bias": "pytorch_model-00001-of-00003.bin",
     "decoder.layers.0.fc1.weight": "pytorch_model-00001-of-00003.bin",
     "decoder.layers.0.fc2.bias": "pytorch_model-00001-of-00003.bin",