update ernie3.0 model

README.md

@@ -18,18 +18,12 @@ a series of experiments have been conducted to check the accuracy of the convers
 - Pytorch Conversion repo:  https://github.com/nghuyong/ERNIE-Pytorch
 
 ## How to use
-If you want to use ernie-3.0 series models, you need to add `task_type_id` to BERT model following this [MR](https://github.com/huggingface/transformers/pull/18686/files) 
-**OR** you can re-install the transformers from my changed branch.
-```bash
-pip uninstall transformers # optional
-pip install git+https://github.com/nghuyong/transformers@add_task_type_id # reinstall
-```
 Then you can load ERNIE-3.0 model as before:
 ```Python
-from transformers import BertTokenizer, BertModel
+from transformers import BertTokenizer, ErnieForMaskedLM
 
 tokenizer = BertTokenizer.from_pretrained("nghuyong/ernie-3.0-base-zh")
-model = BertModel.from_pretrained("nghuyong/ernie-3.0-base-zh")
+model = ErnieForMaskedLM.from_pretrained("nghuyong/ernie-3.0-base-zh")
 ```
 
 ## Citation

config.json

@@ -11,11 +11,10 @@
     "type_vocab_size": 4,
     "use_task_id": true,
     "vocab_size": 40000,
-    "init_class": "ErnieModel",
     "layer_norm_eps": 1e-05,
-    "model_type": "bert",
+    "model_type": "ernie",
     "architectures": [
-        "BertForMaskedLM"
+        "ErnieForMaskedLM"
     ],
     "intermediate_size": 3072
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2a060b11ea32dca2c35f9f0fc27539a80109ea323928569841fae6824c94587e
-size 474372903
+oid sha256:70f4eead24a3263534ec6e70d692a52a1e5fbd199f154076532e6fdcf536cbfd
+size 474373095