PyTorch version 180k steps acc 0.6487

config.json

@@ -0,0 +1,25 @@
+{
+  "architectures": [
+    "RobertaForMaskedLM"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "transformers_version": "4.9.0.dev0",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 50265
+}

convert.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python
+import tempfile
+
+import jax
+from jax import numpy as jnp
+from transformers import AutoTokenizer, FlaxRobertaForMaskedLM, RobertaForMaskedLM
+
+
+def to_f32(t):
+    return jax.tree_map(lambda x: x.astype(jnp.float32) if x.dtype == jnp.bfloat16 else x, t)
+
+
+def main():
+    # Saving extra files from config.json and tokenizer.json files
+    tokenizer = AutoTokenizer.from_pretrained("./")
+    tokenizer.save_pretrained("./")
+
+    # Temporary saving bfloat16 Flax model into float32
+    tmp = tempfile.mkdtemp()
+    flax_model = FlaxRobertaForMaskedLM.from_pretrained("./")
+    flax_model.params = to_f32(flax_model.params)
+    flax_model.save_pretrained(tmp)
+    # Converting float32 Flax to PyTorch
+    model = RobertaForMaskedLM.from_pretrained(tmp, from_flax=True)
+    model.save_pretrained("./", save_config=False)
+
+
+if __name__ == "__main__":
+    main()

flax_model.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:393c37966461709fe51a3b3f84befb7fa7e5030025856d171308efd40dbbc7da
+size 249750019

outputs/events.out.tfevents.1626172316.underestimate.4022703.3.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54e7a88ae2dc3c9128df68ad99b735f3ae87946bc9753da8eb080eb7379dc4d3
-size 26964023
+oid sha256:514deb495913a9b51855279d52c8535930bc9c0cf00f7f56a0af14a233a45244
+size 27038593

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ec165e7aa9031ae2353284d56f1ebef021052449fdf013612050c2cbaa189f8
+size 498858859

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"}