423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-06 05:32:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix webui val size

Browse Source 
Former-commit-id: ad6e7c76c7b5590ea6ca55c4f76db2d3206b5987
This commit is contained in:
hiyouga 2023-08-10 15:20:44 +08:00
parent 0dc9b41b16
commit 58e95776e1
3 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/llmtuner/webui/components/sft.py
View File

@ -38,7 +38,7 @@ def create_sft_tab(top_elems: Dict[str, "Component"], runner: "Runner") -> Dict[
value="cosine", choices=[scheduler.value for scheduler in SchedulerType] value="cosine", choices=[scheduler.value for scheduler in SchedulerType]
) )
max_grad_norm = gr.Textbox(value="1.0") max_grad_norm = gr.Textbox(value="1.0")
dev_ratio = gr.Slider(value=0, minimum=0, maximum=1, step=0.001) val_size = gr.Slider(value=0, minimum=0, maximum=1, step=0.001)
with gr.Accordion(label="Advanced config", open=False) as advanced_tab: with gr.Accordion(label="Advanced config", open=False) as advanced_tab:
with gr.Row(): with gr.Row():
@ -88,7 +88,7 @@ def create_sft_tab(top_elems: Dict[str, "Component"], runner: "Runner") -> Dict[
gradient_accumulation_steps, gradient_accumulation_steps,
lr_scheduler_type, lr_scheduler_type,
max_grad_norm, max_grad_norm,
dev_ratio, val_size,
logging_steps, logging_steps,
save_steps, save_steps,
warmup_steps, warmup_steps,
@ -122,7 +122,7 @@ def create_sft_tab(top_elems: Dict[str, "Component"], runner: "Runner") -> Dict[
gradient_accumulation_steps=gradient_accumulation_steps, gradient_accumulation_steps=gradient_accumulation_steps,
lr_scheduler_type=lr_scheduler_type, lr_scheduler_type=lr_scheduler_type,
max_grad_norm=max_grad_norm, max_grad_norm=max_grad_norm,
dev_ratio=dev_ratio, val_size=val_size,
advanced_tab=advanced_tab, advanced_tab=advanced_tab,
logging_steps=logging_steps, logging_steps=logging_steps,
save_steps=save_steps, save_steps=save_steps,

4
src/llmtuner/webui/locales.py
View File

@ -227,9 +227,9 @@ LOCALES = {
"info": "用于梯度裁剪的范数。" "info": "用于梯度裁剪的范数。"
} }
}, },
"dev_ratio": { "val_size": {
"en": { "en": {
"label": "Dev ratio", "label": "Val size",
"info": "Proportion of data in the dev set." "info": "Proportion of data in the dev set."
}, },
"zh": { "zh": {

6
src/llmtuner/webui/runner.py
View File

@ -83,7 +83,7 @@ class Runner:
gradient_accumulation_steps: int, gradient_accumulation_steps: int,
lr_scheduler_type: str, lr_scheduler_type: str,
max_grad_norm: str, max_grad_norm: str,
dev_ratio: float, val_size: float,
logging_steps: int, logging_steps: int,
save_steps: int, save_steps: int,
warmup_steps: int, warmup_steps: int,
@ -139,8 +139,8 @@ class Runner:
output_dir=output_dir output_dir=output_dir
) )
if dev_ratio > 1e-6: if val_size > 1e-6:
args["dev_ratio"] = dev_ratio args["val_size"] = val_size
args["evaluation_strategy"] = "steps" args["evaluation_strategy"] = "steps"
args["eval_steps"] = save_steps args["eval_steps"] = save_steps
args["load_best_model_at_end"] = True args["load_best_model_at_end"] = True