copies from roberta xlm

flax_model.msgpack

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4c923408bb46e4dc6eadd451c0875742de42256d53079fe7fe233c14e6278f48
-size 1113187999

generate_pt_model.py

@@ -1,6 +0,0 @@
-from transformers import XLMRobertaForMaskedLM, XLMRobertaConfig
-config = XLMRobertaConfig.from_pretrained("./")
-model = XLMRobertaForMaskedLM.from_pretrained("./",config=config,from_flax=True)
-model.save_pretrained("./")
-
-

run.sh

@@ -1,21 +0,0 @@
-python run_mlm_flax_stream.py \
-    --output_dir="../roberta-base-scand-xlm" \
-    --model_name_or_path="xlm-roberta-base" \
-    --config_name="./" \
-    --tokenizer_name="./" \
-    --dataset_name="NbAiLab/scandinavian" \
-    --max_seq_length="512" \
-    --weight_decay="0.01" \
-    --per_device_train_batch_size="62" \
-    --per_device_eval_batch_size="62" \
-    --learning_rate="3e-4" \
-    --warmup_steps="50000" \
-    --overwrite_output_dir \
-    --num_train_steps="1000000" \
-    --adam_beta1="0.9" \
-    --adam_beta2="0.98" \
-    --logging_steps="5000" \
-    --save_steps="25000" \
-    --eval_steps="25000" \
-    --dtype="bfloat16" \
-    --push_to_hub

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:62c24cdc13d4c9952d63718d6c9fa4c287974249e16b7ade6d5a85e7bbb75626
-size 17082660
+oid sha256:a898ea75433890f6610f4e470b8ebeb0c21dce5c8dd61f892eb09eb5919d2e2c
+size 9096718