Dataset Viewer
Auto-converted to Parquet Duplicate
label
stringlengths
6
189
latent
sequencelengths
5
5
a plane with a wooden handle and a wooden handle
[ [ [ [ 0.8046875, 0.50390625, -2.6875 ], [ -2.515625, -4.0625, -1.34375 ], [ -4.6875, -0.2177734375, -1.078125 ] ], [ [ 0.3046875, -0.0966796875, 0.3984375 ], ...
a collection of comic books on a table
[ [ [ [ 0.390625, 0.0703125, 1.453125 ], [ -3.125, -1.5234375, -0.0205078125 ], [ -1.046875, 2.34375, 1.4296875 ] ], [ [ -4.3125, -2.21875, -1.875 ], [ ...
a brown and white bird standing on gravel
[ [ [ [ -0.375, 1.125, -0.12890625 ], [ 1.328125, 1.9375, 1.84375 ], [ 0.765625, -0.703125, 4 ] ], [ [ -1.1953125, -2.609375, 1.8359375 ], [ -0.4414...
a green plant with a green stem
[ [ [ [ 2.5625, 4.0625, -4.5625 ], [ -2.359375, -1.078125, -2.21875 ], [ -3.84375, 3.59375, -0.359375 ] ], [ [ -2.359375, -1.71875, 6.0625 ], [ -0.9...
a sting ray swimming in the ocean
[ [ [ [ -0.2109375, -0.55859375, -0.2490234375 ], [ -0.48828125, 0.6171875, -0.62890625 ], [ -0.39453125, -1.0234375, -1.1875 ] ], [ [ -0.39453125, -0.0272216796875, -0...
a person is under water
[ [ [ [ -3.8125, -0.2734375, -3.3125 ], [ -0.1904296875, -0.40625, -0.890625 ], [ -0.83203125, -4.28125, -0.83984375 ] ], [ [ -2.171875, -3.828125, -0.30078125 ],...
a woman wearing a brown sweater
[ [ [ [ 0.275390625, 0.314453125, 1.4296875 ], [ -3.03125, -0.00750732421875, -2.828125 ], [ -1.4609375, -0.54296875, -2.4375 ] ], [ [ 1.8671875, 0.33984375, -2.796875 ...
a black grand piano in a living room
[ [ [ [ -2.640625, -1.5078125, -1.9765625 ], [ -1.78125, 0.283203125, -4.75 ], [ -1.46875, 0.34765625, -4.875 ] ], [ [ -4.65625, -2.28125, 2.46875 ], [ ...
a young boy sitting on a balance beam
[ [ [ [ 1.96875, -3.78125, -1.6328125 ], [ 1.0625, 0.08447265625, -0.375 ], [ -0.1162109375, -0.53125, 0.01312255859375 ] ], [ [ 0.76171875, -1.4609375, -0.1552734375 ...
a silver apple with four syringes sticking out of it
[ [ [ [ 3.765625, 1.125, 2.140625 ], [ 2.171875, 5.28125, -2.828125 ], [ 3.75, 2.234375, -2.4375 ] ], [ [ 3.265625, 2.15625, 4 ], [ 6.65625, ...
a trilobite fossil on a rock with a black background
[ [ [ [ -2.796875, 0.734375, 1.953125 ], [ 0.5234375, -0.63671875, -2.71875 ], [ 1.046875, 1.53125, -1.6875 ] ], [ [ 3.265625, -1.2734375, 4.1875 ], [ ...
a man with a hat on
[ [ [ [ 0.40234375, -0.34375, 0.73828125 ], [ 2.6875, -0.87890625, 1.046875 ], [ -0.251953125, 0.50390625, 0.0244140625 ] ], [ [ 2.984375, -3.0625, 1.5859375 ], ...
a person sitting on a chair with their feet on the ground
[ [ [ [ 0.12255859375, 2.359375, -1.421875 ], [ 2.9375, -1.421875, -0.984375 ], [ 3.703125, 2.3125, 1.640625 ] ], [ [ -0.9453125, 2.78125, -1.1875 ], [ ...
End of preview. Expand in Data Studio
class ImageNet96Dataset(torch.utils.data.Dataset):
    
    def __init__(
        self, hf_ds, text_enc, tokenizer, bs, ddp, col_label="label", col_latent="latent"
    ):
        self.hf_ds=hf_ds
        self.col_label, self.col_latent = col_label, col_latent
        self.text_enc, self.tokenizer =  text_enc, tokenizer
        self.tokenizer.padding_side = "right"
        self.prompt_len = 50
        
        if ddp: 
            self.sampler = DistributedSampler(hf_ds, shuffle = True, seed = seed)
        else: 
            self.sampler = RandomSampler(hf_ds, generator = torch.manual_seed(seed))
        self.dataloader = DataLoader(
            hf_ds, sampler=self.sampler, collate_fn=self.collate, batch_size=bs, num_workers=4, prefetch_factor=2
        )
    
    def collate(self, items):
        labels = [i[self.col_label] for i in items]
        # latents shape [B, num_aug, 32, 3, 3]
        latents = torch.Tensor([i[self.col_latent] for i in items])
        B, num_aug, _, _, _ = latents.shape

        # pick random augmentation -> latents shape [B, 32, 3, 3]
        aug_idx = torch.randint(0, num_aug, (B,))  # Random int between 0-4 for each batch item
        batch_idx = torch.arange(B)
        latents = latents[batch_idx, aug_idx] 

        return labels, latents
        
    def __iter__(self):
        for labels, latents in self.dataloader:
            label_embs, label_atnmasks = self.encode_prompts(labels)
            latents = latents.to(dtype).to(device)
            yield labels, latents, label_embs, label_atnmasks
    
    def encode_prompts(self, prompts):
        prompts_tok = self.tokenizer(
            prompts, padding="max_length", truncation=True, max_length=self.prompt_len, return_tensors="pt"
        )
        with torch.no_grad():
            prompts_encoded = self.text_enc(**prompts_tok.to(self.text_enc.device))
        return prompts_encoded.last_hidden_state, prompts_tok.attention_mask

    def __len__(self):
        return len(self.dataloader)
Downloads last month
16