初始化项目,由ModelHub XC社区提供模型
Model: flax-community/gpt-2-tamil Source: Original Platform
This commit is contained in:
25
scripts/train_gpt2-oscar-tamil.sh
Executable file
25
scripts/train_gpt2-oscar-tamil.sh
Executable file
@@ -0,0 +1,25 @@
|
||||
#!/usr/bin/env bash
|
||||
python ../src/run_clm_flax.py \
|
||||
--output_dir="${MODEL_DIR}" \
|
||||
--model_type="gpt2" \
|
||||
--config_name="${MODEL_DIR}" \
|
||||
--tokenizer_name="${MODEL_DIR}" \
|
||||
--dataset_name="oscar" \
|
||||
--dataset_config_name="unshuffled_deduplicated_ta" \
|
||||
--do_train --do_eval \
|
||||
--block_size="512" \
|
||||
--per_device_train_batch_size="64" \
|
||||
--per_device_eval_batch_size="64" \
|
||||
--learning_rate="3e-5" \
|
||||
--warmup_steps="1000" \
|
||||
--adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.01" \
|
||||
--overwrite_output_dir \
|
||||
--num_train_epochs="25" \
|
||||
--report_to wandb \
|
||||
--run_name trial \
|
||||
--logging_steps="500" \
|
||||
--save_steps="2500" \
|
||||
--eval_steps="2500" \
|
||||
--preprocessing_num_workers="90" \
|
||||
#--push_to_hub
|
||||
2>&1 | tee run.log
|
||||
Reference in New Issue
Block a user