testing model precision type option

app.py

@@ -4,7 +4,7 @@ import random
 from transformers import T5Tokenizer, T5ForConditionalGeneration
 
 tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-small")
-model = T5ForConditionalGeneration.from_pretrained("roborovski/superprompt-v1", torch_dtype=torch.float16)
+
 
 if torch.cuda.is_available():
     device = "cuda"
@@ -24,8 +24,11 @@ def generate(
     top_p,
     top_k,
     seed,
+    precision_model
 ):
 
+    model = T5ForConditionalGeneration.from_pretrained("roborovski/superprompt-v1", torch_dtype=precision_model)
+
     input_text = f"{system_prompt}, {prompt}"
     input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)
 
@@ -49,6 +52,9 @@ def generate(
     better_prompt = better_prompt.replace("<pad»", "").replace("</s>", "")
     return better_prompt
 
+
+precision_model = gr.Radio([('fp32', torch.float32), ('fp16', toch.float16)], label="Model Precision Type", info="fp32 is more precised but slower, fp16 is faster and less resource consuming but less pricse")
+
 prompt = gr.Textbox(label="Prompt", interactive=True)
 
 system_prompt = gr.Textbox(label="System Prompt", interactive=True)
@@ -65,7 +71,6 @@ top_k = gr.Slider(value=1, minimum=1, maximum=100, step=1, interactive=True, lab
 
 seed = gr.Number(value=42, interactive=True, label="Seed", info="A starting point to initiate the generation process, put 0 for a random one")
 
-
 examples = [
     [
         "A storefront with 'Text to Image' written on it.",