make sure we eval for openorca

src/axolotl/utils/trainer.py

@@ -137,9 +137,9 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
         eval_accumulation_steps=cfg.gradient_accumulation_steps,
         num_train_epochs=cfg.num_epochs,
         learning_rate=cfg.learning_rate,
-        evaluation_strategy="steps" if cfg.val_set_size > 0 else "no",
+        evaluation_strategy="steps",
         save_strategy="steps" if cfg.save_steps else "epoch",
-        eval_steps=cfg.eval_steps if cfg.val_set_size > 0 else None,
+        eval_steps=cfg.eval_steps,
         save_steps=cfg.save_steps,
         output_dir=cfg.output_dir,
         save_total_limit=3,