pytorch model

config.json

@@ -1,4 +1,5 @@
 {
+  "_name_or_path": "./",
   "architectures": [
     "RobertaForMaskedLM"
   ],
@@ -18,6 +19,7 @@
   "num_hidden_layers": 24,
   "pad_token_id": 1,
   "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
   "transformers_version": "4.9.0.dev0",
   "type_vocab_size": 1,
   "use_cache": true,

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b301146ffbe4ae877d1ab09fbdd0684b99c18551120d3c22c60d1f1dfce31fe0
-size 711588089
+oid sha256:db38f2c9c6fbd22b34aeac8405dcd2a3cc28520045e91ff6970302d6694cffaa
+size 1421662309

flax_to_torch.py

@@ -0,0 +1,25 @@
+from transformers import FlaxRobertaForMaskedLM, RobertaForMaskedLM, AutoTokenizer
+import jax
+import jax.numpy as jnp
+
+
+def to_f32(t):
+    return jax.tree_map(
+        lambda x: x.astype(jnp.float32) if x.dtype == jnp.bfloat16 else x, t
+    )
+
+
+# load flax fp16 model
+model = FlaxRobertaForMaskedLM.from_pretrained("./")
+# convert to fp32 model
+model.params = to_f32(model.params)
+# save flax fp32 model
+model.save_pretrained("./")
+
+# convert flax fp32 model to pytorch
+model_pt = RobertaForMaskedLM.from_pretrained("./", from_flax=True)
+model_pt.save_pretrained("./")
+
+tokenizer = AutoTokenizer.from_pretrained("./")
+tokenizer.save_pretrained("./")
+

nohup.out

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2fd23419e898ca51d3fde0021241dba0f6a51ef70fb7a1b35e5b3a8f80f8b8f0
-size 19723143
+oid sha256:c2462fd513ba9b6a8c043fc3a7a7bed1c49514a34106409226c715fddf4b5f90
+size 19725834

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7fe5513a7308eb4a7dd8369b78ef18ee16cc98f612208827e2877a47e1696c9
+size 1421780139

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "<s>", "mask_token": {"content": "<mask>", "single_word": false, "lstrip": true, "rstrip": false, "normalized": false}}

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "add_prefix_space": false, "errors": "replace", "sep_token": "</s>", "cls_token": "<s>", "pad_token": "<pad>", "mask_token": "<mask>", "special_tokens_map_file": null, "name_or_path": "./", "tokenizer_class": "RobertaTokenizer"}