From fa9a9007f9d4fad0a5c404f85463bea43653c6c1 Mon Sep 17 00:00:00 2001
From: wql <wengqinlan18@163.com>
Date: Wed, 4 Sep 2024 16:52:15 +0800
Subject: [PATCH] chore: add lora sft and predict template yaml file

---
 results/lora_sft_template.yaml | 42 ++++++++++++++++++++++++++++++++++
 results/predict_template.yaml  | 23 +++++++++++++++++++
 2 files changed, 65 insertions(+)
 create mode 100644 results/lora_sft_template.yaml
 create mode 100644 results/predict_template.yaml

diff --git a/results/lora_sft_template.yaml b/results/lora_sft_template.yaml
new file mode 100644
index 00000000..a3b42642
--- /dev/null
+++ b/results/lora_sft_template.yaml
@@ -0,0 +1,42 @@
+### model
+model_name_or_path: ../../llm/baichuan
+
+### method
+stage: sft
+do_train: true
+finetuning_type: lora
+lora_target: all
+
+### dataset
+dataset: belle_1m
+template: baichuan
+cutoff_len: 1024
+max_samples: 10000
+overwrite_cache: true
+preprocessing_num_workers: 16
+
+### output
+output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single_step500
+logging_steps: 3
+save_steps: 500
+plot_loss: true
+overwrite_output_dir: true
+
+### train
+per_device_train_batch_size: 2
+gradient_accumulation_steps: 8
+learning_rate: 1.0e-4
+num_train_epochs: 10.0
+lr_scheduler_type: cosine
+warmup_ratio: 0.1
+bf16: true
+ddp_timeout: 180000000
+max_steps: 500
+include_num_input_tokens_seen: true
+include_tokens_per_second: true
+
+### eval
+val_size: 0.1
+per_device_eval_batch_size: 2
+eval_strategy: steps
+eval_steps: 500
\ No newline at end of file
diff --git a/results/predict_template.yaml b/results/predict_template.yaml
new file mode 100644
index 00000000..bafb3f5f
--- /dev/null
+++ b/results/predict_template.yaml
@@ -0,0 +1,23 @@
+### model
+model_name_or_path: ../../llm/baichuan
+
+### method
+do_predict: true
+
+### dataset
+eval_dataset: alpaca_gpt4_zh
+template: baichuan
+cutoff_len: 1024
+max_samples: 50
+overwrite_cache: true
+preprocessing_num_workers: 16
+include_tokens_per_second: true
+
+### output
+output_dir: ./results/inference/Baichuan2-7B/Baichuan2_predict_1
+overwrite_output_dir: true
+
+### eval
+per_device_eval_batch_size: 2
+predict_with_generate: true
+ddp_timeout: 180000000