is_expendable argument reduces memory usage for command line script

image_from_text.py

@@ -15,7 +15,7 @@ parser.set_defaults(torch=False)
 parser.add_argument('--text', type=str, default='alien life')
 parser.add_argument('--seed', type=int, default=7)
 parser.add_argument('--image_path', type=str, default='generated')
-parser.add_argument('--sample_token_count', type=int, default=256) # for debugging
+parser.add_argument('--token_count', type=int, default=256) # for debugging
 
 
 def ascii_from_image(image: Image.Image, size: int) -> str:
@@ -42,20 +42,21 @@ def generate_image(
     text: str,
     seed: int,
     image_path: str,
-    sample_token_count: int
+    token_count: int
 ):
+    is_expendable = True
     if is_torch:
-        image_generator = MinDalleTorch(is_mega, sample_token_count)
+        image_generator = MinDalleTorch(is_mega, is_expendable, token_count)
-        image_tokens = image_generator.generate_image_tokens(text, seed)
 
-        if sample_token_count < image_generator.config['image_length']:
+        if token_count < image_generator.config['image_length']:
+            image_tokens = image_generator.generate_image_tokens(text, seed)
             print('image tokens', list(image_tokens.to('cpu').detach().numpy()))
             return
         else:
             image = image_generator.generate_image(text, seed)
 
     else:
-        image_generator = MinDalleFlax(is_mega)
+        image_generator = MinDalleFlax(is_mega, is_expendable=True)
         image = image_generator.generate_image(text, seed)
 
     save_image(image, image_path)
@@ -71,5 +72,5 @@ if __name__ == '__main__':
         text=args.text,
         seed=args.seed,
         image_path=args.image_path,
-        sample_token_count=args.sample_token_count
+        token_count=args.token_count
     )

min_dalle/min_dalle_base.py

@@ -6,7 +6,7 @@ from .text_tokenizer import TextTokenizer
 from .load_params import load_vqgan_torch_params, load_dalle_bart_flax_params
 from .models.vqgan_detokenizer import VQGanDetokenizer
 
-class MinDalle:
+class MinDalleBase:
     def __init__(self, is_mega: bool):
         self.is_mega = is_mega
         model_name = 'dalle_bart_{}'.format('mega' if is_mega else 'mini')
@@ -25,11 +25,15 @@ class MinDalle:
             merges = f.read().split("\n")[1:-1]
             
         self.model_params = load_dalle_bart_flax_params(model_path)
-
         self.tokenizer = TextTokenizer(vocab, merges)
+
+
+    def init_detokenizer(self):
+        print("initializing VQGanDetokenizer")
+        params = load_vqgan_torch_params('./pretrained/vqgan')
         self.detokenizer = VQGanDetokenizer()
-        vqgan_params = load_vqgan_torch_params('./pretrained/vqgan')
+        self.detokenizer.load_state_dict(params)
-        self.detokenizer.load_state_dict(vqgan_params)
+        del params
 
 
     def tokenize_text(self, text: str) -> numpy.ndarray:

min_dalle/min_dalle_flax.py

@@ -3,18 +3,25 @@ import numpy
 from PIL import Image
 import torch
 
-from .min_dalle import MinDalle
+from .min_dalle_base import MinDalleBase
 from .models.dalle_bart_encoder_flax import DalleBartEncoderFlax
 from .models.dalle_bart_decoder_flax import DalleBartDecoderFlax
 
 
-class MinDalleFlax(MinDalle):
+class MinDalleFlax(MinDalleBase):
-    def __init__(self, is_mega: bool):
+    def __init__(self, is_mega: bool, is_expendable: bool = False):
         super().__init__(is_mega)
+        self.is_expendable = is_expendable
         print("initializing MinDalleFlax")
+        if not is_expendable:
+            self.init_encoder()
+            self.init_decoder()
+            self.init_detokenizer()
 
-        print("loading encoder")
+
-        self.encoder = DalleBartEncoderFlax(
+    def init_encoder(self):
+        print("initializing DalleBartEncoderFlax")
+        self.encoder: DalleBartEncoderFlax = DalleBartEncoderFlax(
             attention_head_count = self.config['encoder_attention_heads'],
             embed_count = self.config['d_model'],
             glu_embed_count = self.config['encoder_ffn_dim'],
@@ -23,7 +30,9 @@ class MinDalleFlax(MinDalle):
             layer_count = self.config['encoder_layers']
         ).bind({'params': self.model_params.pop('encoder')})
 
-        print("loading decoder")
+
+    def init_decoder(self):
+        print("initializing DalleBartDecoderFlax")
         self.decoder = DalleBartDecoderFlax(
             image_token_count = self.config['image_length'],
             text_token_count = self.config['max_text_length'],
@@ -39,20 +48,30 @@ class MinDalleFlax(MinDalle):
     def generate_image(self, text: str, seed: int) -> Image.Image:
         text_tokens = self.tokenize_text(text)
 
+        if self.is_expendable: self.init_encoder()
         print("encoding text tokens")
         encoder_state = self.encoder(text_tokens)
+        if self.is_expendable: del self.encoder
 
+        if self.is_expendable:
+            self.init_decoder()
+            params = self.model_params.pop('decoder')
+        else:
+            params = self.model_params['decoder']
         print("sampling image tokens")
         image_tokens = self.decoder.sample_image_tokens(
             text_tokens,
             encoder_state,
             jax.random.PRNGKey(seed),
-            self.model_params['decoder']
+            params
         )
+        if self.is_expendable: del self.decoder
 
         image_tokens = torch.tensor(numpy.array(image_tokens))
 
+        if self.is_expendable: self.init_detokenizer()
         print("detokenizing image")
         image = self.detokenizer.forward(image_tokens).to(torch.uint8)
+        if self.is_expendable: del self.detokenizer
         image = Image.fromarray(image.to('cpu').detach().numpy())
         return image

min_dalle/min_dalle_torch.py

@@ -9,17 +9,30 @@ torch.set_grad_enabled(False)
 torch.set_num_threads(os.cpu_count())
 
 from .load_params import convert_dalle_bart_torch_from_flax_params
-from .min_dalle import MinDalle
+from .min_dalle_base import MinDalleBase
 from .models.dalle_bart_encoder_torch import DalleBartEncoderTorch
 from .models.dalle_bart_decoder_torch import DalleBartDecoderTorch
 
 
-class MinDalleTorch(MinDalle):
+class MinDalleTorch(MinDalleBase):
-    def __init__(self, is_mega: bool, sample_token_count: int = 256):
+    def __init__(
+        self, 
+        is_mega: bool, 
+        is_expendable: bool = False,
+        token_count: int = 256
+    ):
         super().__init__(is_mega)
+        self.is_expendable = is_expendable
+        self.token_count = token_count
         print("initializing MinDalleTorch")
+        if not is_expendable:
+            self.init_encoder()
+            self.init_decoder()
+            self.init_detokenizer()
 
-        print("loading encoder")
+
+    def init_encoder(self):
+        print("initializing DalleBartEncoderTorch")
         self.encoder = DalleBartEncoderTorch(
             layer_count = self.config['encoder_layers'],
             embed_count = self.config['d_model'],
@@ -28,18 +41,22 @@ class MinDalleTorch(MinDalle):
             text_token_count = self.config['max_text_length'],
             glu_embed_count = self.config['encoder_ffn_dim']
         )
-        encoder_params = convert_dalle_bart_torch_from_flax_params(
+        params = convert_dalle_bart_torch_from_flax_params(
             self.model_params.pop('encoder'), 
             layer_count=self.config['encoder_layers'], 
             is_encoder=True
         )
-        self.encoder.load_state_dict(encoder_params, strict=False)
+        self.encoder.load_state_dict(params, strict=False)
+        if torch.cuda.is_available(): self.encoder = self.encoder.cuda()
+        del params
 
-        print("loading decoder")
+
+    def init_decoder(self):
+        print("initializing DalleBartDecoderTorch")
         self.decoder = DalleBartDecoderTorch(
             image_vocab_size = self.config['image_vocab_size'],
             image_token_count = self.config['image_length'],
-            sample_token_count = sample_token_count,
+            sample_token_count = self.token_count,
             embed_count = self.config['d_model'],
             attention_head_count = self.config['decoder_attention_heads'],
             glu_embed_count = self.config['decoder_ffn_dim'],
@@ -48,16 +65,19 @@ class MinDalleTorch(MinDalle):
             start_token = self.config['decoder_start_token_id'],
             is_verbose = True
         )
-        decoder_params = convert_dalle_bart_torch_from_flax_params(
+        params = convert_dalle_bart_torch_from_flax_params(
             self.model_params.pop('decoder'), 
             layer_count=self.config['decoder_layers'],
             is_encoder=False
         )
-        self.decoder.load_state_dict(decoder_params, strict=False)
+        self.decoder.load_state_dict(params, strict=False)
+        if torch.cuda.is_available(): self.decoder = self.decoder.cuda()
+        del params
 
+    
+    def init_detokenizer(self):
+        super().init_detokenizer()
         if torch.cuda.is_available(): 
-            self.encoder = self.encoder.cuda()
-            self.decoder = self.decoder.cuda()
             self.detokenizer = self.detokenizer.cuda()
             
 
@@ -66,18 +86,24 @@ class MinDalleTorch(MinDalle):
         text_tokens = torch.tensor(text_tokens).to(torch.long)
         if torch.cuda.is_available(): text_tokens = text_tokens.cuda()
 
+        if self.is_expendable: self.init_encoder()
         print("encoding text tokens")
         encoder_state = self.encoder.forward(text_tokens)
+        if self.is_expendable: del self.encoder
 
+        if self.is_expendable: self.init_decoder()
         print("sampling image tokens")
         torch.manual_seed(seed)
         image_tokens = self.decoder.forward(text_tokens, encoder_state)
+        if self.is_expendable: del self.decoder
         return image_tokens
         
 
     def generate_image(self, text: str, seed: int) -> Image.Image:
         image_tokens = self.generate_image_tokens(text, seed)
+        if self.is_expendable: self.init_detokenizer()
         print("detokenizing image")
         image = self.detokenizer.forward(image_tokens).to(torch.uint8)
+        if self.is_expendable: del self.detokenizer
         image = Image.fromarray(image.to('cpu').detach().numpy())
         return image