From 4b0b73c570987963d84512d23bedc998ce73d01d Mon Sep 17 00:00:00 2001 From: wql Date: Mon, 12 Aug 2024 23:24:28 +0800 Subject: [PATCH] train: train_24_8_12_23_21 --- examples/train_lora/llama2_lora_sft.yaml | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/examples/train_lora/llama2_lora_sft.yaml b/examples/train_lora/llama2_lora_sft.yaml index 792fcb78..80467bf5 100644 --- a/examples/train_lora/llama2_lora_sft.yaml +++ b/examples/train_lora/llama2_lora_sft.yaml @@ -8,7 +8,7 @@ finetuning_type: lora lora_target: q_proj,v_proj ### dataset -dataset: alpaca_en +dataset: alpaca_zh template: llama2 cutoff_len: 1024 max_samples: 100000 @@ -16,7 +16,7 @@ overwrite_cache: true preprocessing_num_workers: 16 ### output -output_dir: ./saves/LLaMA2-7B/lora/train_24_8_12_16_46 +output_dir: ./saves/LLaMA2-7B/lora/train_24_8_12_23_21 logging_steps: 3 save_steps: 100 plot_loss: true @@ -34,7 +34,7 @@ ddp_timeout: 180000000 ### eval val_size: 0.1 -per_device_eval_batch_size: 1 +per_device_eval_batch_size: 1 # 1 最大 eval_strategy: steps eval_steps: 500