fix: fix bf16

2024-09-05 15:49:32 +08:00 · 2024-09-05 15:49:32 +08:00 · f15e37dfad
parent e754b62ccd
commit f15e37dfad
2 changed files with 2 additions and 2 deletions
--- a/prepare_yaml_file.py
+++ b/prepare_yaml_file.py
@ -28,7 +28,7 @@ def main():
        model_name_or_path = "../../models/llama-2-7b-ms"
        template = "llama2"
    elif model == "Qwen-7B":
-        model_name_or_path = "/root/models/Qwen-7B"
+        model_name_or_path = "../../models/Qwen-7B"
        template = "qwen"
    else:
        print("ERROR: model not supported.")
--- a/results/lora_sft_template.yaml
+++ b/results/lora_sft_template.yaml
@ -29,7 +29,7 @@ learning_rate: 1.0e-4
 num_train_epochs: 10.0
 lr_scheduler_type: cosine
 warmup_ratio: 0.1
-bf16: true
+fp16: true
 ddp_timeout: 180000000
 max_steps: 500
 include_num_input_tokens_seen: true