v1

README.md

@@ -1,6 +1,6 @@
 ---
 title: Phantom
-emoji: ⛰️
+emoji: 👻
 colorFrom: yellow
 colorTo: purple
 sdk: gradio

app.py

@@ -9,9 +9,7 @@ subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENT
 import torch
 from PIL import Image
 from utils.utils import *
-import torch.nn.functional as F
 from model.load_model import load_model
-from torchvision.transforms.functional import pil_to_tensor
 
 # Gradio Package
 import time
@@ -49,7 +47,7 @@ def threading_function(inputs, streamer, device, model, tokenizer, temperature,
     generation_kwargs.update({'use_cache': True})
     return model.generate(**generation_kwargs)
 
-# @spaces.GPU
+@spaces.GPU
 def bot_streaming(message, history, link, temperature, new_max_token, top_p):
 
     # model selection
@@ -63,7 +61,7 @@ def bot_streaming(message, history, link, temperature, new_max_token, top_p):
         model = model_7
         tokenizer = tokenizer_7
     
-    # X -> float16 conversion 
+    # X -> bfloat16 conversion 
     for param in model.parameters():
         if 'float32' in str(param.dtype).lower() or 'float16' in str(param.dtype).lower():
             param.data = param.data.to(torch.bfloat16)