From 746ceac74a71e501da074581376c99b32892154f Mon Sep 17 00:00:00 2001
From: wql <wengqinlan18@163.com>
Date: Mon, 19 Aug 2024 09:57:13 +0800
Subject: [PATCH] train:test train

---
 .../llama2_lora_sft_1_max_step1000.yaml       | 40 +++++++++++++++++++
 .../llama2_lora_sft_1_no_ms_bf.yaml           | 39 ++++++++++++++++++
 2 files changed, 79 insertions(+)
 create mode 100644 results/lora_sft/Llama2-7B_2/llama2_lora_sft_1_max_step1000.yaml
 create mode 100644 results/lora_sft/Llama2-7B_2/llama2_lora_sft_1_no_ms_bf.yaml

diff --git a/results/lora_sft/Llama2-7B_2/llama2_lora_sft_1_max_step1000.yaml b/results/lora_sft/Llama2-7B_2/llama2_lora_sft_1_max_step1000.yaml
new file mode 100644
index 00000000..318219c8
--- /dev/null
+++ b/results/lora_sft/Llama2-7B_2/llama2_lora_sft_1_max_step1000.yaml
@@ -0,0 +1,40 @@
+### model
+model_name_or_path: modelscope/Llama-2-7b-ms
+
+### method
+stage: sft
+do_train: true
+finetuning_type: lora
+lora_target: all
+
+### dataset
+dataset: belle_1m
+template: llama2
+cutoff_len: 1024
+max_samples: 10000
+overwrite_cache: true
+preprocessing_num_workers: 16
+
+### output
+output_dir: ./results/lora_sft/Llama2-7B/llama2_lora_sft_1
+logging_steps: 3
+save_steps: 100
+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
+fp16: true
+ddp_timeout: 180000000
+max_steps: 1000
+
+### 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/lora_sft/Llama2-7B_2/llama2_lora_sft_1_no_ms_bf.yaml b/results/lora_sft/Llama2-7B_2/llama2_lora_sft_1_no_ms_bf.yaml
new file mode 100644
index 00000000..a43c5e01
--- /dev/null
+++ b/results/lora_sft/Llama2-7B_2/llama2_lora_sft_1_no_ms_bf.yaml
@@ -0,0 +1,39 @@
+### model
+model_name_or_path: modelscope/Llama-2-7b-ms
+
+### method
+stage: sft
+do_train: true
+finetuning_type: lora
+lora_target: all
+
+### dataset
+dataset: belle_1m
+template: llama2
+cutoff_len: 1024
+max_samples: 10000
+overwrite_cache: true
+preprocessing_num_workers: 16
+
+### output
+output_dir: ./results/lora_sft/Llama2-7B/llama2_lora_sft_1
+logging_steps: 3
+save_steps: 100
+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
+
+### eval
+val_size: 0.1
+per_device_eval_batch_size: 2
+eval_strategy: steps
+eval_steps: 500
\ No newline at end of file