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

config.json

@@ -1,16 +1,17 @@
 {
+  "_remove_final_layer_norm": false,
   "activation_dropout": 0.0,
   "activation_function": "relu",
   "architectures": [
-    "OPTForCausalLM"
+    "OPTModel"
   ],
   "attention_dropout": 0.0,
   "bos_token_id": 2,
-  "hidden_size": 7168,
   "do_layer_norm_before": true,
   "dropout": 0.1,
   "eos_token_id": 2,
   "ffn_dim": 28672,
+  "hidden_size": 7168,
   "init_std": 0.02,
   "layerdrop": 0.0,
   "max_position_embeddings": 2048,
@@ -19,7 +20,7 @@
   "num_hidden_layers": 48,
   "pad_token_id": 1,
   "torch_dtype": "float16",
-  "transformers_version": "4.19.0.dev0",
+  "transformers_version": "4.21.0.dev0",
   "use_cache": true,
   "vocab_size": 50272,
   "word_embed_proj_dim": 7168

pytorch_model-00001-of-00007.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b7bb10daf782ce2df3bb3afb57bc7dfa5205e9ea7e39a3987f20e395eb435106
-size 9794437345
+oid sha256:65a74e19aae22c2fed07717eb786104d938a0b0d97523ca7a9ca648915d544ff
+size 9794466629

pytorch_model.bin.index.json

@@ -1,10 +1,12 @@
 {
   "metadata": {
-    "total_size": 59949051904
+    "total_size": 59949080576
   },
   "weight_map": {
     "decoder.embed_positions.weight": "pytorch_model-00001-of-00007.bin",
     "decoder.embed_tokens.weight": "pytorch_model-00001-of-00007.bin",
+    "decoder.final_layer_norm.bias": "pytorch_model-00001-of-00007.bin",
+    "decoder.final_layer_norm.weight": "pytorch_model-00001-of-00007.bin",
     "decoder.layers.0.fc1.bias": "pytorch_model-00001-of-00007.bin",
     "decoder.layers.0.fc1.weight": "pytorch_model-00001-of-00007.bin",
     "decoder.layers.0.fc2.bias": "pytorch_model-00001-of-00007.bin",