Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
bytes: int64
created_utc: timestamp[s]
labels_max: int64
labels_min: int64
latents_dtype: string
latents_shape: list<item: int64>
child 0, item: int64
num_samples: int64
path: string
save_seconds: double
sha256: null
shard_index: int64
verify_ok: bool
num_new_shards: int64
output_dir: string
new_encoded: int64
repo_id: string
vae_backend: string
diffusers_class: string
d2h_time_total_sec: double
encoded_total: int64
latent_mode: string
encode_time_total_sec: double
num_shards_total: int64
h2d_time_total_sec: double
save_time_total_sec: double
elapsed_sec: double
input_layout: string
target_total: int64
num_existing_shards_at_start: int64
last_batch_timing: struct<d2h_sec: double, encode_sec: double, h2d_sec: double, latent_dtype_before_save: string, laten (... 119 chars omitted)
child 0, d2h_sec: double
child 1, encode_sec: double
child 2, h2d_sec: double
child 3, latent_dtype_before_save: string
child 4, latent_flip_mean: double
child 5, latent_flip_std: double
child 6, latent_mean: double
child 7, latent_shape: list<item: int64>
child 0, item: int64
child 8, latent_std: double
shards_tail: list<item: struct<bytes: int64, created_utc: timestamp[s], labels_max: int64, labels_min: int64, lat (... 161 chars omitted)
child 0, item: struct<bytes: int64, created_utc: timestamp[s], labels_max: int64, labels_min: int64, latents_dtype: (... 149 chars omitted)
child 0, bytes: int64
child 1, created_utc: timestamp[s]
child 2, labels_max: int64
child 3, labels_min: int64
child 4, latents_dtype: string
child 5, latents_shape: list<item: int64>
child 0, item: int64
child 6, num_samples: int64
child 7, path: string
child 8, save_seconds: double
child 9, sha256: null
child 10, shard_index: int64
child 11, verify_ok: bool
avg_new_samples_per_sec: double
cuda_peak_allocated_mb: double
finished_utc: timestamp[s]
crop_cache_dir: string
cuda_allocated_mb: double
ok: bool
cuda_reserved_mb: double
status: string
first_latent_shape: list<item: int64>
child 0, item: int64
to
{'avg_new_samples_per_sec': Value('float64'), 'crop_cache_dir': Value('string'), 'cuda_allocated_mb': Value('float64'), 'cuda_peak_allocated_mb': Value('float64'), 'cuda_reserved_mb': Value('float64'), 'd2h_time_total_sec': Value('float64'), 'diffusers_class': Value('string'), 'elapsed_sec': Value('float64'), 'encode_time_total_sec': Value('float64'), 'encoded_total': Value('int64'), 'finished_utc': Value('timestamp[s]'), 'first_latent_shape': List(Value('int64')), 'h2d_time_total_sec': Value('float64'), 'input_layout': Value('string'), 'last_batch_timing': {'d2h_sec': Value('float64'), 'encode_sec': Value('float64'), 'h2d_sec': Value('float64'), 'latent_dtype_before_save': Value('string'), 'latent_flip_mean': Value('float64'), 'latent_flip_std': Value('float64'), 'latent_mean': Value('float64'), 'latent_shape': List(Value('int64')), 'latent_std': Value('float64')}, 'latent_mode': Value('string'), 'new_encoded': Value('int64'), 'num_existing_shards_at_start': Value('int64'), 'num_new_shards': Value('int64'), 'num_shards_total': Value('int64'), 'ok': Value('bool'), 'output_dir': Value('string'), 'repo_id': Value('string'), 'save_time_total_sec': Value('float64'), 'shards_tail': List({'bytes': Value('int64'), 'created_utc': Value('timestamp[s]'), 'labels_max': Value('int64'), 'labels_min': Value('int64'), 'latents_dtype': Value('string'), 'latents_shape': List(Value('int64')), 'num_samples': Value('int64'), 'path': Value('string'), 'save_seconds': Value('float64'), 'sha256': Value('null'), 'shard_index': Value('int64'), 'verify_ok': Value('bool')}), 'status': Value('string'), 'target_total': Value('int64'), 'vae_backend': Value('string')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
return get_rows(
^^^^^^^^^
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2690, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2227, in __iter__
for key, pa_table in self._iter_arrow():
^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2251, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 494, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 384, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 299, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 128, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2321, in table_cast
return cast_table_to_schema(table, schema)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2249, in cast_table_to_schema
raise CastError(
datasets.table.CastError: Couldn't cast
bytes: int64
created_utc: timestamp[s]
labels_max: int64
labels_min: int64
latents_dtype: string
latents_shape: list<item: int64>
child 0, item: int64
num_samples: int64
path: string
save_seconds: double
sha256: null
shard_index: int64
verify_ok: bool
num_new_shards: int64
output_dir: string
new_encoded: int64
repo_id: string
vae_backend: string
diffusers_class: string
d2h_time_total_sec: double
encoded_total: int64
latent_mode: string
encode_time_total_sec: double
num_shards_total: int64
h2d_time_total_sec: double
save_time_total_sec: double
elapsed_sec: double
input_layout: string
target_total: int64
num_existing_shards_at_start: int64
last_batch_timing: struct<d2h_sec: double, encode_sec: double, h2d_sec: double, latent_dtype_before_save: string, laten (... 119 chars omitted)
child 0, d2h_sec: double
child 1, encode_sec: double
child 2, h2d_sec: double
child 3, latent_dtype_before_save: string
child 4, latent_flip_mean: double
child 5, latent_flip_std: double
child 6, latent_mean: double
child 7, latent_shape: list<item: int64>
child 0, item: int64
child 8, latent_std: double
shards_tail: list<item: struct<bytes: int64, created_utc: timestamp[s], labels_max: int64, labels_min: int64, lat (... 161 chars omitted)
child 0, item: struct<bytes: int64, created_utc: timestamp[s], labels_max: int64, labels_min: int64, latents_dtype: (... 149 chars omitted)
child 0, bytes: int64
child 1, created_utc: timestamp[s]
child 2, labels_max: int64
child 3, labels_min: int64
child 4, latents_dtype: string
child 5, latents_shape: list<item: int64>
child 0, item: int64
child 6, num_samples: int64
child 7, path: string
child 8, save_seconds: double
child 9, sha256: null
child 10, shard_index: int64
child 11, verify_ok: bool
avg_new_samples_per_sec: double
cuda_peak_allocated_mb: double
finished_utc: timestamp[s]
crop_cache_dir: string
cuda_allocated_mb: double
ok: bool
cuda_reserved_mb: double
status: string
first_latent_shape: list<item: int64>
child 0, item: int64
to
{'avg_new_samples_per_sec': Value('float64'), 'crop_cache_dir': Value('string'), 'cuda_allocated_mb': Value('float64'), 'cuda_peak_allocated_mb': Value('float64'), 'cuda_reserved_mb': Value('float64'), 'd2h_time_total_sec': Value('float64'), 'diffusers_class': Value('string'), 'elapsed_sec': Value('float64'), 'encode_time_total_sec': Value('float64'), 'encoded_total': Value('int64'), 'finished_utc': Value('timestamp[s]'), 'first_latent_shape': List(Value('int64')), 'h2d_time_total_sec': Value('float64'), 'input_layout': Value('string'), 'last_batch_timing': {'d2h_sec': Value('float64'), 'encode_sec': Value('float64'), 'h2d_sec': Value('float64'), 'latent_dtype_before_save': Value('string'), 'latent_flip_mean': Value('float64'), 'latent_flip_std': Value('float64'), 'latent_mean': Value('float64'), 'latent_shape': List(Value('int64')), 'latent_std': Value('float64')}, 'latent_mode': Value('string'), 'new_encoded': Value('int64'), 'num_existing_shards_at_start': Value('int64'), 'num_new_shards': Value('int64'), 'num_shards_total': Value('int64'), 'ok': Value('bool'), 'output_dir': Value('string'), 'repo_id': Value('string'), 'save_time_total_sec': Value('float64'), 'shards_tail': List({'bytes': Value('int64'), 'created_utc': Value('timestamp[s]'), 'labels_max': Value('int64'), 'labels_min': Value('int64'), 'latents_dtype': Value('string'), 'latents_shape': List(Value('int64')), 'num_samples': Value('int64'), 'path': Value('string'), 'save_seconds': Value('float64'), 'sha256': Value('null'), 'shard_index': Value('int64'), 'verify_ok': Value('bool')}), 'status': Value('string'), 'target_total': Value('int64'), 'vae_backend': Value('string')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
PDM-3-HT FLUX.2 VAE latents for ImageNet-256 train
Public research artifact for PDM-3-HT VAE-backend experiments. This repository contains latent cache shards only. It intentionally does not contain raw ImageNet images, ADM-cropped uint8 images, PAE latents, PAE checkpoints, or training checkpoints.
Source and preprocessing
- Source dataset: ImageNet-1k train via
ILSVRC/imagenet-1k; access requires accepting the upstream ImageNet terms. - Image preprocessing before VAE encode: ADM-style center crop to 256x256, then uint8
[0,255]converted to float[-1,1]. - Raw/cropped images are not redistributed here.
Encoder
- Diffusers model repo:
diffusers/FLUX.2-dev-bnb-4bit - Subfolder:
vae - Diffusers class:
AutoencoderKLFlux2 - Latent mode: deterministic posterior
mode() - VAE forward dtype used for cache build: fp32
- Saved latent dtype: bf16
Schema
Each latents_rank00_shard*.safetensors file contains:
| key | dtype | shape | meaning |
|---|---|---|---|
latents |
bf16 | [N, 32, 32, 32] |
VAE latent for the cropped image |
latents_flip |
bf16 | [N, 32, 32, 32] |
VAE latent for the horizontal flip |
labels |
int64 | [N] |
ImageNet class label |
Shard size is 4096 samples except the final partial shard (3215 samples).
Verified cache summary
- Samples:
1,281,167 - Shards:
313 - Total safetensors size:
167.936GB decimal - Last shard:
latents_rank00_shard000312.safetensors - Generated/upload-prepared UTC:
2026-05-28T03:18:07+00:00
Intended use
This cache is for testing whether PDM-3-HT's HT + MeanFlow trajectory mechanism is VAE-backend agnostic. PAE is treated as a baseline latent backend; this repository is the FLUX.2 VAE backend cache for swap-in ablations.
- Downloads last month
- 1,260