Kevin Fink
commited on
Commit
·
3c24b96
1
Parent(s):
d06ead9
init
Browse files
app.py
CHANGED
@@ -51,7 +51,7 @@ def fine_tune_model(model_name, dataset_name, hub_id, api_key, num_epochs, batch
|
|
51 |
save_strategy='epoch',
|
52 |
learning_rate=lr*0.00001,
|
53 |
per_device_train_batch_size=int(batch_size),
|
54 |
-
per_device_eval_batch_size=
|
55 |
num_train_epochs=int(num_epochs),
|
56 |
weight_decay=0.01,
|
57 |
gradient_accumulation_steps=grad*0.1,
|
@@ -63,7 +63,7 @@ def fine_tune_model(model_name, dataset_name, hub_id, api_key, num_epochs, batch
|
|
63 |
#push_to_hub=True,
|
64 |
hub_model_id=hub_id.strip(),
|
65 |
#fp16=True,
|
66 |
-
lr_scheduler_type='cosine',
|
67 |
)
|
68 |
|
69 |
# Create Trainer
|
@@ -99,10 +99,10 @@ try:
|
|
99 |
gr.Textbox(label="Dataset Name (e.g., 'imdb')"),
|
100 |
gr.Textbox(label="HF hub to push to after training"),
|
101 |
gr.Textbox(label="HF API token"),
|
102 |
-
gr.Slider(minimum=1, maximum=10, value=3, label="Number of Epochs"),
|
103 |
-
gr.Slider(minimum=1, maximum=16, value=4, label="Batch Size"),
|
104 |
-
gr.Slider(minimum=1, maximum=1000, value=50, label="Learning Rate (e-5)"),
|
105 |
-
gr.Slider(minimum=1, maximum=100, value=1, label="Gradient accumulation (e-1)"),
|
106 |
],
|
107 |
outputs="text",
|
108 |
title="Fine-Tune Hugging Face Model",
|
|
|
51 |
save_strategy='epoch',
|
52 |
learning_rate=lr*0.00001,
|
53 |
per_device_train_batch_size=int(batch_size),
|
54 |
+
per_device_eval_batch_size=1,
|
55 |
num_train_epochs=int(num_epochs),
|
56 |
weight_decay=0.01,
|
57 |
gradient_accumulation_steps=grad*0.1,
|
|
|
63 |
#push_to_hub=True,
|
64 |
hub_model_id=hub_id.strip(),
|
65 |
#fp16=True,
|
66 |
+
#lr_scheduler_type='cosine',
|
67 |
)
|
68 |
|
69 |
# Create Trainer
|
|
|
99 |
gr.Textbox(label="Dataset Name (e.g., 'imdb')"),
|
100 |
gr.Textbox(label="HF hub to push to after training"),
|
101 |
gr.Textbox(label="HF API token"),
|
102 |
+
gr.Slider(minimum=1, maximum=10, value=3, label="Number of Epochs", step=1),
|
103 |
+
gr.Slider(minimum=1, maximum=16, value=4, label="Batch Size", step=1),
|
104 |
+
gr.Slider(minimum=1, maximum=1000, value=50, label="Learning Rate (e-5)", step=1),
|
105 |
+
gr.Slider(minimum=1, maximum=100, value=1, label="Gradient accumulation (e-1)", step=1),
|
106 |
],
|
107 |
outputs="text",
|
108 |
title="Fine-Tune Hugging Face Model",
|