Upload 6 files

model.py

@@ -1,8 +1,4 @@
-import os
-from PIL import Image, ImageDraw
-import cv2
-import numpy as np
-from IPython.display import HTML
+from PIL import Image
 from base64 import b64encode
 
 import torch
@@ -14,8 +10,7 @@ from diffusers.schedulers.scheduling_ddim import DDIMScheduler
 #from transformers import CLIPTextModel, CLIPTokenizer
 from tqdm.auto import tqdm
 from huggingface_hub import notebook_login
-
-import weights
+import torch.nn as nn
 
 device = 'cpu'
 
@@ -47,9 +42,6 @@ class MultilingualCLIP(transformers.PreTrainedModel):
         return model, [], [], []
 
 
-import torch
-import torch.nn as nn
-
 # Define the adaptation layer, 'checkpoint_9.pth'
 class AdaptationLayer(nn.Module):
   def __init__(self, input_dim, output_dim):
@@ -87,6 +79,20 @@ adapt_model.to(device)
 state_dict = torch.load('weights/checkpoint_9.pth')
 adapt_model.load_state_dict(state_dict)
 
+from Multilingual_CLIP.multilingual_clip import pt_multilingual_clip
+
+texts = [
+    'قطة تقرأ كتابا'
+]
+
+model_name = 'M-CLIP/LABSE-Vit-L-14'
+
+# Load Model & Tokenizer
+text_model = pt_multilingual_clip.MultilingualCLIP.from_pretrained(model_name)
+text_model = text_model.to(device)
+text_tokenizer = transformers.AutoTokenizer.from_pretrained(model_name)
+
+embeddings= text_model.forward(texts, text_tokenizer, device )
 # 1. Load the autoencoder model which will be used to decode the latents into image space.
 vae = AutoencoderKL.from_pretrained(
     'CompVis/stable-diffusion-v1-4', subfolder='vae', use_auth_token=True)

requirements.txt

@@ -2,4 +2,6 @@ transformers
 diffusers
 torch
 accelerate
-gradio
+gradio
+opencv-python-headless
+tqdm