diff --git a/modules/training.py b/modules/training.py
index 6549b35e..7cb50068 100644
--- a/modules/training.py
+++ b/modules/training.py
@@ -90,19 +90,16 @@ def create_ui():
                     with gr.Accordion(label='Advanced Options', open=False, elem_classes='tgw-accordion'):
                         with gr.Row():
                             with gr.Column():
+                                optimizer = gr.Dropdown(label='Optimizer', value='adamw_torch', choices=['adamw_hf', 'adamw_torch', 'adamw_torch_fused', 'adamw_torch_xla', 'adamw_apex_fused', 'adafactor', 'adamw_bnb_8bit', 'adamw_anyprecision', 'sgd', 'adagrad'], info='Optimizer algorithm. adamw_torch is the standard choice. adamw_bnb_8bit uses less VRAM. adafactor is memory-efficient for large models.', elem_classes=['slim-dropdown'])
+                                warmup_steps = gr.Number(label='Warmup Steps', value=100, info='For this many steps at the start, the learning rate is gradually ramped up from 0 to the target value. This prevents unstable updates early in training.')
                                 lora_dropout = gr.Slider(label='LoRA Dropout', minimum=0.0, maximum=1.0, step=0.025, value=0.0, info='Percentage probability for dropout of LoRA layers. This can help reduce overfitting. Most users should leave at default.')
                                 stop_at_loss = gr.Slider(label='Stop at loss', minimum=0.0, maximum=3.0, step=0.1, value=0.00, info='The process will automatically stop once the desired loss value is reached. (reasonable numbers are 1.5-1.8)')
-                                with gr.Row():
-                                    optimizer = gr.Dropdown(label='Optimizer', value='adamw_torch', choices=['adamw_hf', 'adamw_torch', 'adamw_torch_fused', 'adamw_torch_xla', 'adamw_apex_fused', 'adafactor', 'adamw_bnb_8bit', 'adamw_anyprecision', 'sgd', 'adagrad'], info='Optimizer algorithm. adamw_torch is the standard choice. adamw_bnb_8bit uses less VRAM. adafactor is memory-efficient for large models.', elem_classes=['slim-dropdown'])
 
                             with gr.Column():
-                                warmup_steps = gr.Number(label='Warmup Steps', value=100, info='For this many steps at the start, the learning rate is gradually ramped up from 0 to the target value. This prevents unstable updates early in training.')
-
-                                add_eos_token = gr.Checkbox(label='Add EOS token', value=True, info="Adds EOS token for each document in text datasets.")
-                                excess_length = gr.Dropdown(label='Excess length', value='drop', choices=['drop', 'truncate'], info='What to do with conversations that exceed the cutoff length. "Drop" removes them entirely (recommended). "Truncate" cuts from the right, which may produce incomplete responses.', elem_classes=['slim-dropdown'])
-
                                 gradient_checkpointing = gr.Checkbox(label='Gradient checkpointing', value=True, info='Trades ~20-30% training speed for reduced VRAM usage by recomputing activations during the backward pass instead of storing them. No impact on accuracy.')
+                                add_eos_token = gr.Checkbox(label='Add EOS token', value=True, info="Adds EOS token for each document in text datasets.")
                                 higher_rank_limit = gr.Checkbox(label='Enable higher ranks', value=False, info='If checked, changes Rank/Alpha slider above to go much higher. This will not work without a datacenter-class GPU.')
+                                excess_length = gr.Dropdown(label='Excess length', value='drop', choices=['drop', 'truncate'], info='What to do with conversations that exceed the cutoff length. "Drop" removes them entirely (recommended). "Truncate" cuts from the right, which may produce incomplete responses.', elem_classes=['slim-dropdown'])
                                 report_to = gr.Radio(label="Save detailed logs with", value="None", choices=["None", "wandb", "tensorboard"], interactive=True)
 
                 with gr.Column():