File size: 1,390 Bytes
c287bf2 f82139d c287bf2 311feca c287bf2 233fa8e c287bf2 ebdbe9e c287bf2 311feca 0be6d02 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 |
#!/bin/bash
export HF_PROJECT="gpt2-medium-dutch"
# Variables for training the tokenizer and creating the config
export VOCAB_SIZE="50257"
export DATASET="/home/yeb/data/mc4_nl_cleaned/mc4_nl_cleaned.py" # Name of the dataset in the Huggingface Hub
export DATASET_CONFIG="full" # Config of the dataset in the Huggingface Hub
export DATASET_SPLIT="train" # Split to use for training tokenizer and model
export TEXT_FIELD="text" # Field containing the text to be used for training
export CONFIG_TYPE="gpt2-medium" # Config that our model will use
export MODEL_PATH="${HOME}/data/${HF_PROJECT}" # Path to the model, e.g. here inside the mount
python run_clm_flax.py \
--output_dir="${MODEL_PATH}" \
--model_type="gpt2" \
--config_name="${MODEL_PATH}" \
--model_name_or_path="${MODEL_PATH}" \
--tokenizer_name="${MODEL_PATH}" \
--preprocessing_num_workers="96" \
--do_train --do_eval \
--dataset_name="${DATASET}" \
--dataset_config_name="${DATASET_CONFIG}" \
--block_size="512" \
--per_device_train_batch_size="16" \
--per_device_eval_batch_size="16" \
--learning_rate="0.0024" --warmup_steps="5000" \
--adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.01" \
--overwrite_output_dir \
--num_train_epochs="1" \
--logging_steps="500" \
--save_steps="20001" \
--eval_steps="2500"
# \
# --push_to_hub
|