Update app.py
Browse files
app.py
CHANGED
@@ -107,11 +107,11 @@ trainer = SFTTrainer(
|
|
107 |
train_dataset=dataset,
|
108 |
dataset_text_field="text",
|
109 |
max_seq_length=max_seq_length,
|
110 |
-
dataset_num_proc=
|
111 |
packing=False,
|
112 |
args=TrainingArguments(
|
113 |
-
per_device_train_batch_size=
|
114 |
-
gradient_accumulation_steps=
|
115 |
learning_rate=2e-4,
|
116 |
fp16=not is_bfloat16_supported(),
|
117 |
bf16=is_bfloat16_supported(),
|
@@ -141,13 +141,4 @@ model.push_to_hub_merged(
|
|
141 |
save_method="merged_16bit",
|
142 |
token=True
|
143 |
)
|
144 |
-
print("Model pushed to hub successfully.")
|
145 |
-
|
146 |
-
# Gradio app
|
147 |
-
print("Launching Gradio app...")
|
148 |
-
def greet(name):
|
149 |
-
return "Hello " + name + "!!"
|
150 |
-
|
151 |
-
demo = gr.Interface(fn=greet, inputs="text", outputs="text")
|
152 |
-
demo.launch()
|
153 |
-
print("Gradio app launched.")
|
|
|
107 |
train_dataset=dataset,
|
108 |
dataset_text_field="text",
|
109 |
max_seq_length=max_seq_length,
|
110 |
+
dataset_num_proc=20,
|
111 |
packing=False,
|
112 |
args=TrainingArguments(
|
113 |
+
per_device_train_batch_size=20,
|
114 |
+
gradient_accumulation_steps=20,
|
115 |
learning_rate=2e-4,
|
116 |
fp16=not is_bfloat16_supported(),
|
117 |
bf16=is_bfloat16_supported(),
|
|
|
141 |
save_method="merged_16bit",
|
142 |
token=True
|
143 |
)
|
144 |
+
print("Model pushed to hub successfully.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|