codeShare commited on
Commit
56285d0
Β·
verified Β·
1 Parent(s): d49f46c

Upload fusion_t2i_CLIP_interrogator.ipynb

Browse files
Google Colab Jupyter Notebooks/fusion_t2i_CLIP_interrogator.ipynb CHANGED
@@ -90,21 +90,7 @@
90
  "processor = CLIPProcessor.from_pretrained(\"openai/clip-vit-large-patch14\" , clean_up_tokenization_spaces = True)\n",
91
  "model = CLIPModel.from_pretrained(\"openai/clip-vit-large-patch14\")\n",
92
  "logit_scale = model.logit_scale.exp() #logit_scale = 100.00000762939453\n",
93
- "f_add = torch.nn.quantized.FloatFunctional()\n",
94
  "\n",
95
- "\n",
96
- "\n",
97
- "\n"
98
- ],
99
- "metadata": {
100
- "id": "TC5lMJrS1HCC"
101
- },
102
- "execution_count": null,
103
- "outputs": []
104
- },
105
- {
106
- "cell_type": "code",
107
- "source": [
108
  "index = 0\n",
109
  "%cd {home_directory + 'fusion-t2i-generator-data/' + 'vocab'}\n",
110
  "vocab_encodings = torch.load('vocab_encodings.pt', weights_only=False)\n",
@@ -118,10 +104,11 @@
118
  "for key in torch.load('reference_text_and_image_encodings.pt', weights_only=False):\n",
119
  " index = index + 1;\n",
120
  "#------#\n",
121
- "NUM_REFERENCE_ITEMS = index"
 
122
  ],
123
  "metadata": {
124
- "id": "Z8x3Y7IsnGlT"
125
  },
126
  "execution_count": null,
127
  "outputs": []
@@ -131,7 +118,7 @@
131
  "source": [
132
  "# @title \tβš„ Use a pre-encoded prompt + image pair from the fusion gen (note: NSFW!)\n",
133
  "# @markdown Choose a pre-encoded reference\n",
134
- "index = 457 # @param {type:\"slider\", min:0, max:1666, step:1}\n",
135
  "PROMPT_INDEX = index\n",
136
  "prompt = target_prompts[f'{PROMPT_INDEX}']\n",
137
  "url = target_urls[f'{PROMPT_INDEX}']\n",
@@ -139,24 +126,24 @@
139
  " image = Image.open(requests.get(url, stream=True).raw)\n",
140
  "#------#\n",
141
  "# @markdown βš–οΈ πŸ–ΌοΈ encoding <-----?-----> πŸ“ encoding </div> <br>\n",
142
- "C = 1 # @param {type:\"slider\", min:0, max:1, step:0.01}\n",
143
- "log_strength_1 = 1.06 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
144
- "prompt_strength = round(math.pow(10 ,log_strength_1-1),2)\n",
145
- "reference = torch.zeros(768)\n",
146
  "\n",
147
  "%cd {home_directory + 'fusion-t2i-generator-data/' + 'reference'}\n",
148
  "references = torch.load('reference_text_and_image_encodings.pt' , weights_only=False)\n",
149
- "reference = torch.add(reference, prompt_strength * C * references[index][0].dequantize())\n",
150
- "reference = torch.add(reference, prompt_strength * (1-C) * references[index][1].dequantize())\n",
151
  "references = '' # Clear up memory\n",
152
  "# @markdown -----------\n",
153
  "# @markdown πŸ“βž• 1st Enhance similarity to prompt(s)\n",
154
  "POS_2 = '' # @param {type:'string' ,placeholder:'item1 , item2 , ...'}\n",
155
  "log_strength_2 = 1.03 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
156
- "pos_strength = round(math.pow(10 ,log_strength_2-1),2)\n",
157
- "for _POS in POS_2.split(','):\n",
158
  " inputs = tokenizer(text = _POS.strip(), truncation = True , padding=True, return_tensors=\"pt\")\n",
159
- " text_features_POS = model.get_text_features(**inputs)\n",
160
  " text_features_POS = text_features_POS/text_features_POS.norm(p=2, dim=-1, keepdim=True)\n",
161
  " reference = torch.add(reference, pos_strength * text_features_POS)\n",
162
  "# @markdown -----------\n",
@@ -165,10 +152,10 @@
165
  "# @markdown πŸ“βž• 2nd Enhance similarity to prompt(s)\n",
166
  "POS = '' # @param {type:'string' ,placeholder:'item1 , item2 , ...'}\n",
167
  "log_strength_3 = 1.06 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
168
- "pos_strength = round(math.pow(10 ,log_strength_3-1),2)\n",
169
- "for _POS in POS.split(','):\n",
170
  " inputs = tokenizer(text = _POS.strip(), truncation = True , padding=True, return_tensors=\"pt\")\n",
171
- " text_features_POS = model.get_text_features(**inputs)\n",
172
  " text_features_POS = text_features_POS/text_features_POS.norm(p=2, dim=-1, keepdim=True)\n",
173
  " reference = torch.add(reference, pos_strength * text_features_POS)\n",
174
  "# @markdown -----------\n",
@@ -176,25 +163,30 @@
176
  "# @markdown 🚫 Penalize similarity to prompt(s)\n",
177
  "NEG = '' # @param {type:'string' , placeholder:'item1 , item2 , ...'}\n",
178
  "log_strength_4 = 1 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
179
- "neg_strength = round(math.pow(10 ,log_strength_4-1),2)\n",
180
- "for _NEG in NEG.split(','):\n",
181
  " inputs = tokenizer(text = _NEG.strip(), truncation = True , padding=True, return_tensors=\"pt\")\n",
182
- " text_features_NEG = model.get_text_features(**inputs)\n",
183
  " text_features_NEG = text_features_NEG/text_features_NEG.norm(p=2, dim=-1, keepdim=True)\n",
184
  " reference = torch.sub(reference, neg_strength * text_features_NEG)\n",
185
  "# @markdown -----------\n",
186
  "# @markdown ⏩ Skip item(s) containing the word(s)\n",
187
  "SKIP = '' # @param {type:'string' , placeholder:'item1 , item2 , ...'}\n",
188
  "\n",
189
- "def isBlacklisted(txt, blacklist):\n",
190
- " if txt.strip().isnumeric(): return True\n",
191
- " if blacklist.strip() == '': return False\n",
 
 
 
 
 
192
  " for item in list(blacklist.split(',')):\n",
193
  " if item.strip() == '' : continue\n",
194
  " if txt.find(item.strip())> -1 : return True\n",
195
  " #------#\n",
196
  " found = False\n",
197
- " alphabet = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ'\n",
198
  " for letter in alphabet:\n",
199
  " found = txt.find(letter)>-1\n",
200
  " if found:break\n",
@@ -213,7 +205,8 @@
213
  "update_list = True # @param {type:\"boolean\"}\n",
214
  "\n",
215
  "calculate_variance = False # @param {type:\"boolean\"}\n",
216
- "enable = update_list\n",
 
217
  "\n",
218
  "try: first\n",
219
  "except:\n",
@@ -253,12 +246,18 @@
253
  "for _index in range(list_size):\n",
254
  " tmp = prompts[f'{indices[min(_index+start_at_index,NUM_VOCAB_ITEMS-1)].item()}']\n",
255
  " if isBlacklisted(tmp , SKIP): continue\n",
256
- " output = output + fix_bad_symbols(tmp) + '|'\n",
 
 
257
  "#---------#\n",
258
  "output = (output + '}').replace('|}' , '} ')\n",
 
 
259
  "for iter in range(N):\n",
260
  " print(output)\n",
261
  "#-------#\n",
 
 
262
  "image or print('No image found')"
263
  ],
264
  "metadata": {
@@ -649,7 +648,7 @@
649
  "cellView": "form",
650
  "id": "uDzsk02CbMFc"
651
  },
652
- "execution_count": 49,
653
  "outputs": []
654
  },
655
  {
@@ -681,7 +680,7 @@
681
  "cellView": "form",
682
  "id": "Azz1kCza6LB3"
683
  },
684
- "execution_count": 56,
685
  "outputs": []
686
  }
687
  ]
 
90
  "processor = CLIPProcessor.from_pretrained(\"openai/clip-vit-large-patch14\" , clean_up_tokenization_spaces = True)\n",
91
  "model = CLIPModel.from_pretrained(\"openai/clip-vit-large-patch14\")\n",
92
  "logit_scale = model.logit_scale.exp() #logit_scale = 100.00000762939453\n",
 
93
  "\n",
 
 
 
 
 
 
 
 
 
 
 
 
 
94
  "index = 0\n",
95
  "%cd {home_directory + 'fusion-t2i-generator-data/' + 'vocab'}\n",
96
  "vocab_encodings = torch.load('vocab_encodings.pt', weights_only=False)\n",
 
104
  "for key in torch.load('reference_text_and_image_encodings.pt', weights_only=False):\n",
105
  " index = index + 1;\n",
106
  "#------#\n",
107
+ "NUM_REFERENCE_ITEMS = index\n",
108
+ "\n"
109
  ],
110
  "metadata": {
111
+ "id": "TC5lMJrS1HCC"
112
  },
113
  "execution_count": null,
114
  "outputs": []
 
118
  "source": [
119
  "# @title \tβš„ Use a pre-encoded prompt + image pair from the fusion gen (note: NSFW!)\n",
120
  "# @markdown Choose a pre-encoded reference\n",
121
+ "index = 213 # @param {type:\"slider\", min:0, max:1666, step:1}\n",
122
  "PROMPT_INDEX = index\n",
123
  "prompt = target_prompts[f'{PROMPT_INDEX}']\n",
124
  "url = target_urls[f'{PROMPT_INDEX}']\n",
 
126
  " image = Image.open(requests.get(url, stream=True).raw)\n",
127
  "#------#\n",
128
  "# @markdown βš–οΈ πŸ–ΌοΈ encoding <-----?-----> πŸ“ encoding </div> <br>\n",
129
+ "C = 0.3 # @param {type:\"slider\", min:0, max:1, step:0.01}\n",
130
+ "log_strength_1 = 2.17 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
131
+ "prompt_strength = torch.tensor(math.pow(10 ,log_strength_1-1)).to(dtype = torch.float32)\n",
132
+ "reference = torch.zeros(768).to(dtype = torch.float32)\n",
133
  "\n",
134
  "%cd {home_directory + 'fusion-t2i-generator-data/' + 'reference'}\n",
135
  "references = torch.load('reference_text_and_image_encodings.pt' , weights_only=False)\n",
136
+ "reference = torch.add(reference, prompt_strength * C * references[index][0].dequantize().to(dtype = torch.float32))\n",
137
+ "reference = torch.add(reference, prompt_strength * (1-C) * references[index][1].dequantize().to(dtype = torch.float32))\n",
138
  "references = '' # Clear up memory\n",
139
  "# @markdown -----------\n",
140
  "# @markdown πŸ“βž• 1st Enhance similarity to prompt(s)\n",
141
  "POS_2 = '' # @param {type:'string' ,placeholder:'item1 , item2 , ...'}\n",
142
  "log_strength_2 = 1.03 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
143
+ "pos_strength = torch.tensor(math.pow(10 ,log_strength_2-1)).to(dtype = torch.float32)\n",
144
+ "for _POS in POS_2.replace('</w>' , ' ').replace('{' , '').replace('}' , '').replace('|' , ',').split(','):\n",
145
  " inputs = tokenizer(text = _POS.strip(), truncation = True , padding=True, return_tensors=\"pt\")\n",
146
+ " text_features_POS = model.get_text_features(**inputs).to(dtype = torch.float32)\n",
147
  " text_features_POS = text_features_POS/text_features_POS.norm(p=2, dim=-1, keepdim=True)\n",
148
  " reference = torch.add(reference, pos_strength * text_features_POS)\n",
149
  "# @markdown -----------\n",
 
152
  "# @markdown πŸ“βž• 2nd Enhance similarity to prompt(s)\n",
153
  "POS = '' # @param {type:'string' ,placeholder:'item1 , item2 , ...'}\n",
154
  "log_strength_3 = 1.06 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
155
+ "pos_strength = torch.tensor(math.pow(10 ,log_strength_3-1)).to(dtype = torch.float32)\n",
156
+ "for _POS in POS.replace('</w>' , ' ').replace('{' , '').replace('}' , '').replace('|' , ',').split(','):\n",
157
  " inputs = tokenizer(text = _POS.strip(), truncation = True , padding=True, return_tensors=\"pt\")\n",
158
+ " text_features_POS = model.get_text_features(**inputs).to(dtype = torch.float32)\n",
159
  " text_features_POS = text_features_POS/text_features_POS.norm(p=2, dim=-1, keepdim=True)\n",
160
  " reference = torch.add(reference, pos_strength * text_features_POS)\n",
161
  "# @markdown -----------\n",
 
163
  "# @markdown 🚫 Penalize similarity to prompt(s)\n",
164
  "NEG = '' # @param {type:'string' , placeholder:'item1 , item2 , ...'}\n",
165
  "log_strength_4 = 1 # @param {type:\"slider\", min:-5, max:5, step:0.01}\n",
166
+ "neg_strength = torch.tensor(math.pow(10 ,log_strength_4-1)).to(dtype = torch.float32)\n",
167
+ "for _NEG in NEG.replace('</w>' , ' ').replace('{' , '').replace('}' , '').replace('|' , ',').split(','):\n",
168
  " inputs = tokenizer(text = _NEG.strip(), truncation = True , padding=True, return_tensors=\"pt\")\n",
169
+ " text_features_NEG = model.get_text_features(**inputs).to(dtype = torch.float32)\n",
170
  " text_features_NEG = text_features_NEG/text_features_NEG.norm(p=2, dim=-1, keepdim=True)\n",
171
  " reference = torch.sub(reference, neg_strength * text_features_NEG)\n",
172
  "# @markdown -----------\n",
173
  "# @markdown ⏩ Skip item(s) containing the word(s)\n",
174
  "SKIP = '' # @param {type:'string' , placeholder:'item1 , item2 , ...'}\n",
175
  "\n",
176
+ "min_wordcount = 0 # @param {type:\"slider\", min:0, max:20, step:1}\n",
177
+ "\n",
178
+ "def isBlacklisted(_txt, _blacklist):\n",
179
+ " blacklist = _blacklist.lower().replace('</w>' , ' ').replace('{' , '').replace('}' , '').replace('|' , ',').strip()\n",
180
+ " txt = _txt.lower().strip()\n",
181
+ " if len(txt)<min_wordcount: return True\n",
182
+ " if txt.isnumeric(): return True\n",
183
+ " if blacklist == '': return False\n",
184
  " for item in list(blacklist.split(',')):\n",
185
  " if item.strip() == '' : continue\n",
186
  " if txt.find(item.strip())> -1 : return True\n",
187
  " #------#\n",
188
  " found = False\n",
189
+ " alphabet = 'abcdefghijklmnopqrstuvxyz'\n",
190
  " for letter in alphabet:\n",
191
  " found = txt.find(letter)>-1\n",
192
  " if found:break\n",
 
205
  "update_list = True # @param {type:\"boolean\"}\n",
206
  "\n",
207
  "calculate_variance = False # @param {type:\"boolean\"}\n",
208
+ "\n",
209
+ "ne = update_list\n",
210
  "\n",
211
  "try: first\n",
212
  "except:\n",
 
246
  "for _index in range(list_size):\n",
247
  " tmp = prompts[f'{indices[min(_index+start_at_index,NUM_VOCAB_ITEMS-1)].item()}']\n",
248
  " if isBlacklisted(tmp , SKIP): continue\n",
249
+ " tmp = fix_bad_symbols(tmp)\n",
250
+ " if output.find(tmp)>-1:continue\n",
251
+ " output = output + tmp + '|'\n",
252
  "#---------#\n",
253
  "output = (output + '}').replace('|}' , '} ')\n",
254
+ "print('')\n",
255
+ "print('')\n",
256
  "for iter in range(N):\n",
257
  " print(output)\n",
258
  "#-------#\n",
259
+ "print('')\n",
260
+ "print('')\n",
261
  "image or print('No image found')"
262
  ],
263
  "metadata": {
 
648
  "cellView": "form",
649
  "id": "uDzsk02CbMFc"
650
  },
651
+ "execution_count": null,
652
  "outputs": []
653
  },
654
  {
 
680
  "cellView": "form",
681
  "id": "Azz1kCza6LB3"
682
  },
683
+ "execution_count": null,
684
  "outputs": []
685
  }
686
  ]