From a67b736d0c9c03cd1636ff1597fbd77159008e0c Mon Sep 17 00:00:00 2001 From: IMvision12 Date: Sun, 30 Aug 2026 15:58:45 -0700 Subject: [PATCH 1/5] tests --- .github/workflows/test_code.yml | 17 ++- tests/base/model_test_registry.py | 43 ++++++ tests/conftest.py | 122 ++++++++++++++++-- .../integration/test_backend_compatibility.py | 6 +- tests/integration/test_data_formats.py | 88 +++++++------ tests/integration/test_model_saving.py | 3 +- tests/integration/test_serialization.py | 6 +- 7 files changed, 227 insertions(+), 58 deletions(-) diff --git a/.github/workflows/test_code.yml b/.github/workflows/test_code.yml index 994508ee..ee98e3be 100644 --- a/.github/workflows/test_code.yml +++ b/.github/workflows/test_code.yml @@ -18,13 +18,25 @@ jobs: - run: ruff format --check . test-code: - name: Test the code (${{ matrix.backend }}) + name: Test the code (${{ matrix.backend }} ${{ matrix.shard }}) needs: lint-and-format runs-on: ubuntu-latest strategy: fail-fast: false + # Shard the model suite BY MODEL across parallel jobs (see the --shard + # option in tests/conftest.py). torch is fast enough unsharded; the JAX and + # TF legs trace + XLA-compile every model, so they are split to stay under + # the per-job time cap. Stacks with the per-model build-once reuse. matrix: - backend: [torch, tensorflow, jax] + include: + - { backend: torch, shard: "1/1" } + - { backend: tensorflow, shard: "1/3" } + - { backend: tensorflow, shard: "2/3" } + - { backend: tensorflow, shard: "3/3" } + - { backend: jax, shard: "1/4" } + - { backend: jax, shard: "2/4" } + - { backend: jax, shard: "3/4" } + - { backend: jax, shard: "4/4" } env: KERAS_BACKEND: ${{ matrix.backend }} steps: @@ -59,6 +71,7 @@ jobs: tests/integration/test_serialization.py \ tests/integration/test_model_saving.py \ tests/integration/test_data_formats.py \ + --shard ${{ matrix.shard }} \ -v --durations=20 --cov=zeromodels --cov-append \ -m "not slow and not gpu" - name: Generate coverage diff --git a/tests/base/model_test_registry.py b/tests/base/model_test_registry.py index 14e128a3..e667c1e1 100644 --- a/tests/base/model_test_registry.py +++ b/tests/base/model_test_registry.py @@ -4352,6 +4352,49 @@ def instantiate_model(config): return model +# Per-model build cache. The integration suite is reordered (see +# tests/conftest.py) so every test of one model runs consecutively; a READ-ONLY +# test then reuses the model built here instead of rebuilding it. Building + +# XLA-compiling a functional model is the dominant per-test cost on the JAX / TF +# backends (seconds each), and the suite otherwise rebuilds each model ~10x +# across its tests. The cache is cleared when the model changes, so peak memory +# stays at ~one model (the CI RAM cap that the per-test teardown protects). +_MODEL_CACHE = {} + + +def get_cached_model(config): + """Return a shared, build-once model for ``config``'s READ-ONLY tests. + + Keyed by (class, active data_format, quantized) so a channels_first build or + a quantized variant never aliases the plain one. Use this only for tests that + do not mutate the model (forward / shape / NaN / ``get_config`` / ``save*``); + a test that assigns weights, builds a different data format, or calls + ``clear_session`` must call :func:`instantiate_model` for a fresh instance. + """ + import os + + if os.environ.get("ZM_NO_MODEL_CACHE") == "1": + return instantiate_model(config) # escape hatch: force a fresh build + + import keras + + key = ( + config["model_cls"], + keras.config.image_data_format(), + bool(config.get("quantization_config")), + ) + model = _MODEL_CACHE.get(key) + if model is None: + model = instantiate_model(config) + _MODEL_CACHE[key] = model + return model + + +def clear_model_cache(): + """Drop all cached models (called on model change by the conftest teardown).""" + _MODEL_CACHE.clear() + + # Generative VLMs became functional models (#390): text-only factories no longer # satisfy their expanded (image + video + mask + position) input signatures, so flag # them for the model-driven multimodal builder in create_test_input. diff --git a/tests/conftest.py b/tests/conftest.py index d0887f2e..0ed20184 100644 --- a/tests/conftest.py +++ b/tests/conftest.py @@ -1,21 +1,82 @@ import gc +import hashlib import os import pytest -@pytest.fixture(autouse=True) -def _release_backend_state(): - """Release per-test Keras / JAX state to keep CI memory bounded. - - Each parametrized model in the integration suite triggers fresh - XLA / TF function tracing. Without an explicit teardown the JIT - cache, compiled HLO modules, and dead Keras layers accumulate - across the 300+ tests and the JAX matrix entry hits the - ubuntu-latest runner's 7 GB RAM / 60 min timeout (visible as - process SIGTERM, exit code 143). +def _node_model_name(node): + """The ``model_name`` parametrization of a test node, or None if it has none.""" + callspec = getattr(node, "callspec", None) + if callspec is None: + return None + return callspec.params.get("model_name") + + +def _parse_shard(spec): + """Parse a ``k/n`` shard spec (1-based k) into a 0-based (index, count).""" + k_str, n_str = spec.split("/") + k, n = int(k_str), int(n_str) + if not 1 <= k <= n: + raise pytest.UsageError(f"--shard {spec!r}: need 1 <= k <= n") + return k - 1, n + + +def pytest_collection_modifyitems(config, items): + """Group every test of a model together, then optionally keep one shard. + + A stable sort by ``model_name`` puts all of one model's tests (across the + backend-compat / serialization / saving / data-format files) back to back, + so :func:`get_cached_model` can hand out one built model to that model's + read-only tests and it can be released in a single teardown when the model + changes. Tests with no ``model_name`` keep their original order as one + leading group. + + With ``--shard k/n`` the models are round-robin assigned to ``n`` shards and + only shard ``k`` is kept. Sharding is BY MODEL (not by test) so a model's + whole test group stays on one shard: the per-model build-once reuse holds, + and no model is built on more than one CI runner. Non-model tests are + distributed by a stable hash of their node id so each runs on exactly one + shard. Splitting a backend's models across parallel jobs is what brings the + slow JAX / TF legs under the per-job time cap (stacks with the reuse above). """ - yield + original = {id(item): i for i, item in enumerate(items)} + items.sort(key=lambda item: (_node_model_name(item) or "", original[id(item)])) + + spec = config.getoption("shard") + if not spec: + return + shard_index, shard_count = _parse_shard(spec) + if shard_count == 1: + return + model_names = sorted({n for n in map(_node_model_name, items) if n is not None}) + model_shard = {name: i % shard_count for i, name in enumerate(model_names)} + + def item_shard(item): + name = _node_model_name(item) + if name is not None: + return model_shard[name] + digest = hashlib.md5(item.nodeid.encode()).hexdigest() + return int(digest, 16) % shard_count + + selected, deselected = [], [] + for item in items: + (selected if item_shard(item) == shard_index else deselected).append(item) + if deselected: + config.hook.pytest_deselected(items=deselected) + items[:] = selected + + +_CURRENT_MODEL = ["\x00unset"] + + +def _flush_backend_state(): + try: + from tests.base.model_test_registry import clear_model_cache + + clear_model_cache() + except Exception: + pass try: import keras @@ -29,6 +90,34 @@ def _release_backend_state(): gc.collect() +@pytest.fixture(autouse=True) +def _release_backend_state(request): + """Release the previous model's build + XLA compilation when the model changes. + + Each parametrized model triggers fresh XLA / TF tracing; the JIT cache, + compiled HLO, and dead layers otherwise accumulate across the 300+ tests and + the JAX matrix entry hits the ubuntu-latest 7 GB RAM / 60 min cap (SIGTERM, + exit 143). The old fix cleared after *every* test, which also threw away the + build + compile so each of a model's ~10 tests paid them again (hours on JAX). + + Because tests are now grouped per model, clearing only when the model changes + keeps peak memory at ~one model *and* lets that model's build + compile be + reused across its tests. Non-model tests (``model_name is None``) clear every + time, preserving the original bounded-memory behavior for them. + """ + # ZM_LEGACY_CLEAR=1 restores the old clear-after-every-test behavior, for + # A/B timing against the per-model reuse (pair with ZM_NO_MODEL_CACHE=1). + if os.environ.get("ZM_LEGACY_CLEAR") == "1": + yield + _flush_backend_state() + return + name = _node_model_name(request.node) + if name is None or name != _CURRENT_MODEL[0]: + _flush_backend_state() + _CURRENT_MODEL[0] = name + yield + + def pytest_addoption(parser): parser.addoption( "--backend", @@ -42,6 +131,17 @@ def pytest_addoption(parser): default=None, help="Image data format: channels_first, channels_last", ) + parser.addoption( + "--shard", + action="store", + dest="shard", + default=None, + help=( + "Run only shard k of n (format 'k/n', 1-based), sharded BY MODEL so " + "a model's tests stay together. Splits a backend's models across " + "parallel CI jobs." + ), + ) def pytest_configure(config): diff --git a/tests/integration/test_backend_compatibility.py b/tests/integration/test_backend_compatibility.py index 3fb5ed7f..dbad7f01 100644 --- a/tests/integration/test_backend_compatibility.py +++ b/tests/integration/test_backend_compatibility.py @@ -6,7 +6,7 @@ from tests.base.model_test_registry import ( MODEL_TEST_CONFIGS, create_test_input, - instantiate_model, + get_cached_model, ) BACKEND = os.environ.get("KERAS_BACKEND", "torch") @@ -31,7 +31,7 @@ def _skip_if_incompatible(model_name): def test_model_forward_pass(model_name): _skip_if_incompatible(model_name) config = MODEL_TEST_CONFIGS[model_name] - model = instantiate_model(config) + model = get_cached_model(config) input_data = create_test_input(config, model=model) output = model(input_data) @@ -72,7 +72,7 @@ def test_model_forward_pass(model_name): def test_model_no_nans(model_name): _skip_if_incompatible(model_name) config = MODEL_TEST_CONFIGS[model_name] - model = instantiate_model(config) + model = get_cached_model(config) input_data = create_test_input(config, model=model) output = model(input_data) diff --git a/tests/integration/test_data_formats.py b/tests/integration/test_data_formats.py index 13e55ba2..343c790d 100644 --- a/tests/integration/test_data_formats.py +++ b/tests/integration/test_data_formats.py @@ -8,6 +8,7 @@ from tests.base.model_test_registry import ( MODEL_TEST_CONFIGS, create_test_input, + get_cached_model, import_model_class, ) from zeromodels.base import BaseProcessor @@ -70,18 +71,29 @@ def _adapt_input_shape_for_format(init_kwargs, data_format): return kwargs +def _is_channels_last_image(key, value): + if not (hasattr(value, "shape") and len(value.shape) == 4): + return False + if int(value.shape[-1]) not in (1, 3, 4): + return False + return key is None or "pixel" in key or "image" in key + + +def _has_transposable_image(input_data): + if isinstance(input_data, dict): + return any(_is_channels_last_image(k, v) for k, v in input_data.items()) + return _is_channels_last_image(None, input_data) + + def _transpose_input(input_data, data_format): if data_format != "channels_first": return input_data if isinstance(input_data, dict): - result = {} - for k, v in input_data.items(): - if k in ("pixel_values", "images") and len(v.shape) == 4: - result[k] = ops.transpose(v, (0, 3, 1, 2)) - else: - result[k] = v - return result - if len(input_data.shape) == 4: + return { + k: (ops.transpose(v, (0, 3, 1, 2)) if _is_channels_last_image(k, v) else v) + for k, v in input_data.items() + } + if _is_channels_last_image(None, input_data): return ops.transpose(input_data, (0, 3, 1, 2)) return input_data @@ -96,8 +108,7 @@ def test_channels_last(model_name): try: keras.config.set_image_data_format("channels_last") config = MODEL_TEST_CONFIGS[model_name] - model_cls = import_model_class(config) - model = model_cls(**config["init_kwargs"]) + model = get_cached_model(config) input_data = create_test_input(config, model=model) output = model(input_data) @@ -195,22 +206,6 @@ def _output_rel(a, b): return min(rels) if rels else None -def _has_image_input(config): - """True if the model's test input carries a 4D spatial image tensor. - - channels_first only affects models that consume a spatial ``(B, H, W, C)`` / - ``(B, C, H, W)`` image. Text LLMs, generative VLMs (pre-patchified / - token-id inputs), and ASR have nothing to transpose, so the parity check - does not apply to them and they are skipped. - """ - try: - x = create_test_input(config) - except Exception: - return False - tensors = list(x.values()) if isinstance(x, dict) else [x] - return any(hasattr(t, "shape") and len(t.shape) == 4 for t in tensors) - - @pytest.mark.data_format @pytest.mark.parametrize("model_name", MODEL_IDS) def test_channels_first_matches_channels_last(model_name): @@ -220,21 +215,20 @@ def test_channels_first_matches_channels_last(model_name): ``Reshape`` at a token<->grid boundary silently scrambles the data (it stays finite and keeps the right shape), so equivalence needs a direct comparison: build one model per format with the *same* weights (conv kernels are - ``(kh, kw, in, out)`` regardless of format) and assert the outputs match - after transposing the channels_first result back to channels_last. - - Scoped to models with a spatial image input (vision backbones, detection, - segmentation, depth, DINO, SAM, and the CLIP / SigLIP / MetaCLIP 2 / TIPS - dual encoders); text LLMs and generative VLMs are skipped by - :func:`_has_image_input`. + ``(kh, kw, in, out)`` regardless of format), feed the same input with its + spatial image transposed, and assert the outputs match (transposing a + channels_first feature map back to channels_last). + + Covers every model whose *built* input carries a spatial image: vision + backbones, detection, segmentation, depth, DINO, SAM, the CLIP / SigLIP / + MetaCLIP 2 / TIPS dual encoders, and the raw-image generative VLMs (Gemma 3, + Mistral 3, DeepSeek-VL, InternVL, Janus). Text LLMs, ASR, and pre-patchified + VLMs (Qwen-VL, GLM-4V, Kimi) have no spatial image to transpose and skip via + :func:`_has_transposable_image`. """ if model_name in SKIP_DATA_FORMAT: pytest.skip(f"{model_name} doesn't support data format switching") - config = MODEL_TEST_CONFIGS[model_name] - if not _has_image_input(config): - pytest.skip(f"{model_name}: no spatial image input; channels_first is a no-op") - if BACKEND == "tensorflow": try: import tensorflow as tf @@ -244,13 +238,31 @@ def test_channels_first_matches_channels_last(model_name): except ImportError: pytest.skip("TensorFlow not installed") + config = MODEL_TEST_CONFIGS[model_name] original = keras.config.image_data_format() try: model_cls = import_model_class(config) + # Start from a clean global layer-name counter. The per-model build cache + # leaves a model alive in this group, and keras dedups auto-named layers + # against a global counter; without this reset the cl build gets + # "_1"-suffixed layer names that the post-clear cf build does not, which + # breaks the exact-path weight copy below. + keras.backend.clear_session() keras.config.set_image_data_format("channels_last") model_cl = model_cls(**config["init_kwargs"]) - input_data = create_test_input(config) + # Generative VLM inputs are built from the model's own input signature; + # every other model's input comes straight from its config. + if config.get("multimodal_vlm"): + input_data = create_test_input(config, model=model_cl) + else: + input_data = create_test_input(config) + + if not _has_transposable_image(input_data): + pytest.skip( + f"{model_name}: no spatial image input; channels_first is a no-op" + ) + out_cl = [ ops.convert_to_numpy(t) for t in _flatten_outputs(model_cl(input_data)) ] diff --git a/tests/integration/test_model_saving.py b/tests/integration/test_model_saving.py index cb69a1fa..27421b03 100644 --- a/tests/integration/test_model_saving.py +++ b/tests/integration/test_model_saving.py @@ -5,6 +5,7 @@ from tests.base.model_test_registry import ( MODEL_TEST_CONFIGS, create_test_input, + get_cached_model, instantiate_model, ) @@ -53,7 +54,7 @@ def test_save_weights_h5_roundtrip(model_name, tmp_path): if model_name in SKIP_SAVING: pytest.skip(f"{model_name} is subclassed; load_weights needs a built model") config = MODEL_TEST_CONFIGS[model_name] - model = instantiate_model(config) + model = get_cached_model(config) input_data = create_test_input(config, model=model) original_output = model(input_data) diff --git a/tests/integration/test_serialization.py b/tests/integration/test_serialization.py index 46d6e991..18e466ba 100644 --- a/tests/integration/test_serialization.py +++ b/tests/integration/test_serialization.py @@ -7,7 +7,7 @@ from tests.base.model_test_registry import ( MODEL_TEST_CONFIGS, create_test_input, - instantiate_model, + get_cached_model, ) BACKEND = os.environ.get("KERAS_BACKEND", "torch") @@ -24,7 +24,7 @@ def test_config_roundtrip(model_name): pytest.skip(f"{model_name} causes TF backend segfault during serialization") config = MODEL_TEST_CONFIGS[model_name] - model = instantiate_model(config) + model = get_cached_model(config) cfg = model.get_config() revived = model.__class__.from_config(cfg) @@ -178,7 +178,7 @@ def test_keras_serialization_roundtrip(model_name): pytest.skip(f"{model_name} causes TF backend segfault during serialization") config = MODEL_TEST_CONFIGS[model_name] - model = instantiate_model(config) + model = get_cached_model(config) serialized = keras.saving.serialize_keras_object(model) json_str = json.dumps(serialized, indent=4, default=str) From 3e02ddae27af335c2d575f80959858b5dcd43d33 Mon Sep 17 00:00:00 2001 From: IMvision12 Date: Sun, 30 Aug 2026 16:06:33 -0700 Subject: [PATCH 2/5] format --- docs/dinov2.md | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) diff --git a/docs/dinov2.md b/docs/dinov2.md index 40abd6fa..30187376 100644 --- a/docs/dinov2.md +++ b/docs/dinov2.md @@ -110,9 +110,7 @@ from PIL import Image from zeromodels.models.dino_v2 import DinoV2ImageProcessor, DinoV2Model size, patch = 896, 14 -model = DinoV2Model.from_weights( - "zeromodels/dinov2-giant", image_size=size -) +model = DinoV2Model.from_weights("zeromodels/dinov2-giant", image_size=size) processor = DinoV2ImageProcessor.from_weights( "zeromodels/dinov2-giant", resize_size=1024, crop_size=size ) @@ -161,9 +159,7 @@ import torch from zeromodels.models.dino_v2 import DinoV2ImageProcessor, DinoV2Model size = 896 -model = DinoV2Model.from_weights( - "zeromodels/dinov2-giant", image_size=size -) +model = DinoV2Model.from_weights("zeromodels/dinov2-giant", image_size=size) processor = DinoV2ImageProcessor.from_weights( "zeromodels/dinov2-giant", resize_size=1024, crop_size=size ) From 012cfcc056bcbb5d232810d73b32269ed21b8149 Mon Sep 17 00:00:00 2001 From: IMvision12 Date: Sun, 30 Aug 2026 16:19:52 -0700 Subject: [PATCH 3/5] fix --- tests/fixtures/image_processor_snapshots.json | 4612 +++++++++++++---- 1 file changed, 3495 insertions(+), 1117 deletions(-) diff --git a/tests/fixtures/image_processor_snapshots.json b/tests/fixtures/image_processor_snapshots.json index 0321bff0..a750cbf2 100644 --- a/tests/fixtures/image_processor_snapshots.json +++ b/tests/fixtures/image_processor_snapshots.json @@ -1,4 +1,62 @@ { + "BeitImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.96529, + "mean": -0.00404, + "min": -0.98002, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.27224 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.20094, + "mean": -0.00492, + "min": -1.23322, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.46871 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24145, + "mean": -0.00304, + "min": -1.36308, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47369 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27045, + "mean": -0.00303, + "min": -1.26931, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47238 + } + } + }, "CLIPImageProcessor": { "large_300x500": { "pixel_values": { @@ -57,6 +115,238 @@ } } }, + "CaiTImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "ConvMixerImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "ConvNeXtImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "ConvNeXtV2ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, "DETRImageProcessor": { "large_300x500": { "pixel_values": { @@ -395,33 +685,149 @@ } } }, - "DepthAnythingV1ImageProcessor": { + "DeiTImageProcessor": { "large_300x500": { - "original_size": { - "max": 500.0, - "mean": 400.0, - "min": 300.0, - "shape": [ - 2 - ], - "std": 100.0 - }, "pixel_values": { - "max": 3.14196, - "mean": 0.21764, - "min": -2.69672, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 518, - 518, + 224, + 224, 3 ], - "std": 1.05085 - }, - "reshaped_size": { - "max": 518.0, - "mean": 518.0, - "min": 518.0, + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "DenseNetImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "DepthAnythingV1ImageProcessor": { + "large_300x500": { + "original_size": { + "max": 500.0, + "mean": 400.0, + "min": 300.0, + "shape": [ + 2 + ], + "std": 100.0 + }, + "pixel_values": { + "max": 3.14196, + "mean": 0.21764, + "min": -2.69672, + "shape": [ + 1, + 518, + 518, + 3 + ], + "std": 1.05085 + }, + "reshaped_size": { + "max": 518.0, + "mean": 518.0, + "min": 518.0, "shape": [ 2 ], @@ -963,1163 +1369,2859 @@ } } }, - "EoMTImageProcessor": { + "EfficientFormerImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.59862, - "mean": -0.66389, - "min": -2.1179, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 1.27328 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.60861, - "mean": 0.19494, - "min": -2.1179, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.88565 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.59265, - "mean": -0.88322, - "min": -2.1179, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 1.26605 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.60203, - "mean": -0.88321, - "min": -2.1179, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 1.26503 + "std": 1.05204 } } }, - "Gemma3ImageProcessor": { + "EfficientNetImageProcessor": { "large_300x500": { "pixel_values": { - "max": 1.0, - "mean": -0.00398, - "min": -1.0, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 896, - 896, + 224, + 224, 3 ], - "std": 0.46908 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 1.0, - "mean": -0.00485, - "min": -1.0, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 896, - 896, + 224, + 224, 3 ], - "std": 0.46887 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 1.0, - "mean": -0.00291, - "min": -1.0, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 896, - 896, + 224, + 224, 3 ], - "std": 0.47187 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 1.0, - "mean": -0.00299, - "min": -1.0, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 896, - 896, + 224, + 224, 3 ], - "std": 0.47074 + "std": 1.05204 } } }, - "Gemma3nImageProcessor": { + "EfficientNetLiteImageProcessor": { "large_300x500": { "pixel_values": { - "max": 0.99216, - "mean": -0.00396, - "min": -1.0, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.3855 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 0.99216, - "mean": -0.0049, - "min": -1.0, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.38557 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 0.98431, - "mean": -0.00268, - "min": -0.99216, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.38849 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 0.97647, - "mean": -0.00268, - "min": -1.0, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.3871 + "std": 1.05204 } } }, - "Gemma4ImageProcessor": { + "EfficientNetV2ImageProcessor": { "large_300x500": { - "image_position_ids": { - "max": 62.0, - "mean": 21.725, - "min": -1.0, - "shape": [ - 1, - 2520, - 2 - ], - "std": 17.19737 - }, - "num_soft_tokens_per_image": { - "max": 252.0, - "mean": 252.0, - "min": 252.0, - "shape": [ - 1 - ], - "std": 0.0 - }, "pixel_values": { - "max": 1.0, - "mean": 0.44821, - "min": 0.0, + "max": 0.96529, + "mean": -0.00404, + "min": -0.98002, "shape": [ 1, - 2520, - 768 + 224, + 224, + 3 ], - "std": 0.268 + "std": 0.27224 } }, "odd_223x225": { - "image_position_ids": { - "max": 47.0, - "mean": 21.4, - "min": -1.0, - "shape": [ - 1, - 2520, + "pixel_values": { + "max": 1.20094, + "mean": -0.00492, + "min": -1.23322, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.46871 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24145, + "mean": -0.00304, + "min": -1.36308, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47369 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27045, + "mean": -0.00303, + "min": -1.26931, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47238 + } + } + }, + "EoMTImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.59862, + "mean": -0.66389, + "min": -2.1179, + "shape": [ + 1, + 640, + 640, + 3 + ], + "std": 1.27328 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 2.60861, + "mean": 0.19494, + "min": -2.1179, + "shape": [ + 1, + 640, + 640, + 3 + ], + "std": 0.88565 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 2.59265, + "mean": -0.88322, + "min": -2.1179, + "shape": [ + 1, + 640, + 640, + 3 + ], + "std": 1.26605 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 2.60203, + "mean": -0.88321, + "min": -2.1179, + "shape": [ + 1, + 640, + 640, + 3 + ], + "std": 1.26503 + } + } + }, + "FlexiViTImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "Gemma3ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 1.0, + "mean": -0.00398, + "min": -1.0, + "shape": [ + 1, + 896, + 896, + 3 + ], + "std": 0.46908 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.0, + "mean": -0.00485, + "min": -1.0, + "shape": [ + 1, + 896, + 896, + 3 + ], + "std": 0.46887 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.0, + "mean": -0.00291, + "min": -1.0, + "shape": [ + 1, + 896, + 896, + 3 + ], + "std": 0.47187 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.0, + "mean": -0.00299, + "min": -1.0, + "shape": [ + 1, + 896, + 896, + 3 + ], + "std": 0.47074 + } + } + }, + "Gemma3nImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.99216, + "mean": -0.00396, + "min": -1.0, + "shape": [ + 1, + 768, + 768, + 3 + ], + "std": 0.3855 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 0.99216, + "mean": -0.0049, + "min": -1.0, + "shape": [ + 1, + 768, + 768, + 3 + ], + "std": 0.38557 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 0.98431, + "mean": -0.00268, + "min": -0.99216, + "shape": [ + 1, + 768, + 768, + 3 + ], + "std": 0.38849 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 0.97647, + "mean": -0.00268, + "min": -1.0, + "shape": [ + 1, + 768, + 768, + 3 + ], + "std": 0.3871 + } + } + }, + "Gemma4ImageProcessor": { + "large_300x500": { + "image_position_ids": { + "max": 62.0, + "mean": 21.725, + "min": -1.0, + "shape": [ + 1, + 2520, + 2 + ], + "std": 17.19737 + }, + "num_soft_tokens_per_image": { + "max": 252.0, + "mean": 252.0, + "min": 252.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.44821, + "min": 0.0, + "shape": [ + 1, + 2520, + 768 + ], + "std": 0.268 + } + }, + "odd_223x225": { + "image_position_ids": { + "max": 47.0, + "mean": 21.4, + "min": -1.0, + "shape": [ + 1, + 2520, + 2 + ], + "std": 14.91666 + }, + "num_soft_tokens_per_image": { + "max": 256.0, + "mean": 256.0, + "min": 256.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.45492, + "min": 0.0, + "shape": [ + 1, + 2520, + 768 + ], + "std": 0.26392 + } + }, + "tall_96x48": { + "image_position_ids": { + "max": 68.0, + "mean": 22.49286, + "min": -1.0, + "shape": [ + 1, + 2520, + 2 + ], + "std": 18.76858 + }, + "num_soft_tokens_per_image": { + "max": 253.0, + "mean": 253.0, + "min": 253.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.45048, + "min": 0.0, + "shape": [ + 1, + 2520, + 768 + ], + "std": 0.26824 + } + }, + "wide_48x96": { + "image_position_ids": { + "max": 68.0, + "mean": 22.49286, + "min": -1.0, + "shape": [ + 1, + 2520, + 2 + ], + "std": 18.76858 + }, + "num_soft_tokens_per_image": { + "max": 253.0, + "mean": 253.0, + "min": 253.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.45044, + "min": 0.0, + "shape": [ + 1, + 2520, + 768 + ], + "std": 0.26779 + } + } + }, + "Gemma4UnifiedImageProcessor": { + "large_300x500": { + "image_position_ids": { + "max": 20.0, + "mean": 6.875, + "min": -1.0, + "shape": [ + 1, + 280, + 2 + ], + "std": 5.77143 + }, + "num_soft_tokens_per_image": { + "max": 252.0, + "mean": 252.0, + "min": 252.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.44821, + "min": 0.0, + "shape": [ + 1, + 280, + 6912 + ], + "std": 0.268 + } + }, + "odd_223x225": { + "image_position_ids": { + "max": 15.0, + "mean": 6.77143, + "min": -1.0, + "shape": [ + 1, + 280, + 2 + ], + "std": 5.00905 + }, + "num_soft_tokens_per_image": { + "max": 256.0, + "mean": 256.0, + "min": 256.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.45492, + "min": 0.0, + "shape": [ + 1, + 280, + 6912 + ], + "std": 0.26392 + } + }, + "tall_96x48": { + "image_position_ids": { + "max": 22.0, + "mean": 7.13214, + "min": -1.0, + "shape": [ + 1, + 280, + 2 + ], + "std": 6.29175 + }, + "num_soft_tokens_per_image": { + "max": 253.0, + "mean": 253.0, + "min": 253.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.45048, + "min": 0.0, + "shape": [ + 1, + 280, + 6912 + ], + "std": 0.26824 + } + }, + "wide_48x96": { + "image_position_ids": { + "max": 22.0, + "mean": 7.13214, + "min": -1.0, + "shape": [ + 1, + 280, + 2 + ], + "std": 6.29175 + }, + "num_soft_tokens_per_image": { + "max": 253.0, + "mean": 253.0, + "min": 253.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 1.0, + "mean": 0.45044, + "min": 0.0, + "shape": [ + 1, + 280, + 6912 + ], + "std": 0.26779 + } + } + }, + "Glm4vImageProcessor": { + "large_300x500": { + "image_grid_thw": { + "max": 36.0, + "mean": 19.66667, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 14.38363 + }, + "pixel_values": { + "max": 2.1459, + "mean": 0.18024, + "min": -1.79226, + "shape": [ + 792, + 1176 + ], + "std": 0.88065 + } + }, + "odd_223x225": { + "image_grid_thw": { + "max": 16.0, + "mean": 11.0, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 7.07107 + }, + "pixel_values": { + "max": 2.1459, + "mean": 0.17867, + "min": -1.79226, + "shape": [ + 256, + 1176 + ], + "std": 0.8767 + } + }, + "tall_96x48": { + "image_grid_thw": { + "max": 12.0, + "mean": 6.33333, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 4.49691 + }, + "pixel_values": { + "max": 2.1459, + "mean": 0.18258, + "min": -1.79226, + "shape": [ + 72, + 1176 + ], + "std": 0.8836 + } + }, + "wide_48x96": { + "image_grid_thw": { + "max": 12.0, + "mean": 6.33333, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 4.49691 + }, + "pixel_values": { + "max": 2.1459, + "mean": 0.18233, + "min": -1.79226, + "shape": [ + 72, + 1176 + ], + "std": 0.88152 + } + } + }, + "GroundingDinoImageProcessor": { + "large_300x500": { + "pixel_mask": { + "max": 1.0, + "mean": 1.0, + "min": 1.0, + "shape": [ + 1, + 800, + 1333 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.62257, + "mean": 0.21811, + "min": -2.10078, + "shape": [ + 1, + 800, + 1333, + 3 + ], + "std": 0.86341 + } + }, + "odd_223x225": { + "pixel_mask": { + "max": 1.0, + "mean": 1.0, + "min": 1.0, + "shape": [ + 1, + 800, + 807 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.62257, + "mean": 0.2156, + "min": -2.1179, + "shape": [ + 1, + 800, + 807, + 3 + ], + "std": 0.86402 + } + }, + "tall_96x48": { + "pixel_mask": { + "max": 1.0, + "mean": 1.0, + "min": 1.0, + "shape": [ + 1, + 1333, + 666 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.62257, + "mean": 0.21961, + "min": -2.1179, + "shape": [ + 1, + 1333, + 666, + 3 + ], + "std": 0.86992 + } + }, + "wide_48x96": { + "pixel_mask": { + "max": 1.0, + "mean": 1.0, + "min": 1.0, + "shape": [ + 1, + 666, + 1333 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.60514, + "mean": 0.21961, + "min": -2.10078, + "shape": [ + 1, + 666, + 1333, + 3 + ], + "std": 0.86695 + } + } + }, + "InceptionNextImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.96529, + "mean": -0.00404, + "min": -0.98002, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.27224 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.20094, + "mean": -0.00492, + "min": -1.23322, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.46871 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24145, + "mean": -0.00304, + "min": -1.36308, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47369 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27045, + "mean": -0.00303, + "min": -1.26931, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47238 + } + } + }, + "InceptionResNetV2ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 1.11938, + "mean": -0.00404, + "min": -1.11777, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.36307 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.25057, + "mean": -0.00492, + "min": -1.20156, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47012 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24844, + "mean": -0.00304, + "min": -1.37676, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47368 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.26429, + "mean": -0.00304, + "min": -1.27812, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47237 + } + } + }, + "InceptionV3ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 1.11938, + "mean": -0.00404, + "min": -1.11777, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.36307 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.25057, + "mean": -0.00492, + "min": -1.20156, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47012 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24844, + "mean": -0.00304, + "min": -1.37676, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47368 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.26429, + "mean": -0.00304, + "min": -1.27812, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47237 + } + } + }, + "InceptionV4ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 1.11938, + "mean": -0.00404, + "min": -1.11777, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.36307 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.25057, + "mean": -0.00492, + "min": -1.20156, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47012 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24844, + "mean": -0.00304, + "min": -1.37676, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47368 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.26429, + "mean": -0.00304, + "min": -1.27812, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47237 + } + } + }, + "InternVLImageProcessor": { + "large_300x500": { + "num_patches": { + "max": 7.0, + "mean": 7.0, + "min": 7.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.64, + "mean": 0.21774, + "min": -2.1179, + "shape": [ + 7, + 448, + 448, + 3 + ], + "std": 1.04048 + } + }, + "odd_223x225": { + "num_patches": { + "max": 1.0, + "mean": 1.0, + "min": 1.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.64, + "mean": 0.21573, + "min": -2.1179, + "shape": [ + 1, + 448, + 448, + 3 + ], + "std": 1.04644 + } + }, + "tall_96x48": { + "num_patches": { + "max": 3.0, + "mean": 3.0, + "min": 3.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.64, + "mean": 0.21988, + "min": -2.1179, + "shape": [ + 3, + 448, + 448, + 3 + ], + "std": 1.0509 + } + }, + "wide_48x96": { + "num_patches": { + "max": 3.0, + "mean": 3.0, + "min": 3.0, + "shape": [ + 1 + ], + "std": 0.0 + }, + "pixel_values": { + "max": 2.64, + "mean": 0.21974, + "min": -2.1179, + "shape": [ + 3, + 448, + 448, + 3 + ], + "std": 1.04848 + } + } + }, + "JanusImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 1.11482, + "mean": -0.00399, + "min": -1.15907, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 0.27936 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.23231, + "mean": -0.0049, + "min": -1.24259, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 0.46848 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24756, + "mean": -0.00348, + "min": -1.37134, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 0.33484 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27612, + "mean": -0.00348, + "min": -1.28159, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 0.33391 + } + } + }, + "KimiK25ImageProcessor": { + "large_300x500": { + "image_grid_thw": { + "max": 36.0, + "mean": 19.66667, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 14.38363 + }, + "pixel_values": { + "max": 0.99216, + "mean": -0.0376, + "min": -1.0, + "shape": [ + 792, + 3, + 14, + 14 + ], + "std": 0.59538 + } + }, + "odd_223x225": { + "image_grid_thw": { + "max": 18.0, + "mean": 11.66667, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 7.58654 + }, + "pixel_values": { + "max": 0.99216, + "mean": -0.1155, + "min": -1.0, + "shape": [ + 288, + 3, + 14, + 14 + ], + "std": 0.62716 + } + }, + "tall_96x48": { + "image_grid_thw": { + "max": 8.0, + "mean": 4.33333, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 2.86744 + }, + "pixel_values": { + "max": 0.99216, + "mean": -0.26755, + "min": -1.0, + "shape": [ + 32, + 3, + 14, + 14 + ], + "std": 0.66262 + } + }, + "wide_48x96": { + "image_grid_thw": { + "max": 8.0, + "mean": 4.33333, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 2.86744 + }, + "pixel_values": { + "max": 0.99216, + "mean": -0.26755, + "min": -1.0, + "shape": [ + 32, + 3, + 14, + 14 + ], + "std": 0.66262 + } + } + }, + "LevitImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "LocateAnythingImageProcessor": { + "large_300x500": { + "image_grid_hws": { + "max": 36.0, + "mean": 29.0, + "min": 22.0, + "shape": [ + 1, + 2 + ], + "std": 7.0 + }, + "pixel_values": { + "max": 1.0, + "mean": -0.004, + "min": -1.0, + "shape": [ + 792, + 3, + 14, + 14 + ], + "std": 0.46899 + } + }, + "odd_223x225": { + "image_grid_hws": { + "max": 18.0, + "mean": 17.0, + "min": 16.0, + "shape": [ + 1, + 2 + ], + "std": 1.0 + }, + "pixel_values": { + "max": 1.0, + "mean": -0.00483, + "min": -1.0, + "shape": [ + 288, + 3, + 14, + 14 + ], + "std": 0.46811 + } + }, + "tall_96x48": { + "image_grid_hws": { + "max": 8.0, + "mean": 6.0, + "min": 4.0, + "shape": [ + 1, + 2 + ], + "std": 2.0 + }, + "pixel_values": { + "max": 1.0, + "mean": -0.00286, + "min": -1.0, + "shape": [ + 32, + 3, + 14, + 14 + ], + "std": 0.47032 + } + }, + "wide_48x96": { + "image_grid_hws": { + "max": 8.0, + "mean": 6.0, + "min": 4.0, + "shape": [ + 1, 2 ], - "std": 14.91666 - }, - "num_soft_tokens_per_image": { - "max": 256.0, - "mean": 256.0, - "min": 256.0, + "std": 2.0 + }, + "pixel_values": { + "max": 1.0, + "mean": -0.00292, + "min": -1.0, + "shape": [ + 32, + 3, + 14, + 14 + ], + "std": 0.46883 + } + } + }, + "MLPMixerImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "Mask2FormerImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.61525, + "mean": -0.66613, + "min": -2.1179, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 1.27301 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 2.61651, + "mean": 0.19259, + "min": -2.1179, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 0.88845 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 2.52889, + "mean": -0.88322, + "min": -2.1179, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 1.26184 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 2.55503, + "mean": -0.88322, + "min": -2.1179, + "shape": [ + 1, + 384, + 384, + 3 + ], + "std": 1.26078 + } + } + }, + "MaskFormerImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.60677, + "mean": -0.6647, + "min": -2.1179, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 1.27298 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 2.61896, + "mean": 0.19412, + "min": -2.1179, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 0.88675 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 2.58792, + "mean": -0.88321, + "min": -2.1179, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 1.26595 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 2.57913, + "mean": -0.8832, + "min": -2.1179, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 1.26493 + } + } + }, + "MaxViTImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "MetaClip2ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.0748, + "mean": 0.18018, + "min": -1.66088, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.51898 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 2.1459, + "mean": 0.17867, + "min": -1.79226, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.8767 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 2.1459, + "mean": 0.18242, + "min": -1.79226, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.88356 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 2.1459, + "mean": 0.18227, + "min": -1.79226, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.88139 + } + } + }, + "MiTImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ - 1 + 1, + 224, + 224, + 3 ], - "std": 0.0 + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "MiniMaxM3VLImageProcessor": { + "large_300x500": { + "image_grid_thw": { + "max": 36.0, + "mean": 19.66667, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 14.38363 }, "pixel_values": { - "max": 1.0, - "mean": 0.45492, - "min": 0.0, + "max": 2.1459, + "mean": 0.18024, + "min": -1.79226, + "shape": [ + 792, + 1176 + ], + "std": 0.88065 + } + }, + "odd_223x225": { + "image_grid_thw": { + "max": 16.0, + "mean": 11.0, + "min": 1.0, "shape": [ 1, - 2520, - 768 + 3 ], - "std": 0.26392 + "std": 7.07107 + }, + "pixel_values": { + "max": 2.1459, + "mean": 0.17867, + "min": -1.79226, + "shape": [ + 256, + 1176 + ], + "std": 0.8767 } }, "tall_96x48": { - "image_position_ids": { - "max": 68.0, - "mean": 22.49286, - "min": -1.0, + "image_grid_thw": { + "max": 6.0, + "mean": 3.66667, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 2.0548 + }, + "pixel_values": { + "max": 2.1459, + "mean": 0.18243, + "min": -1.79226, + "shape": [ + 24, + 1176 + ], + "std": 0.82552 + } + }, + "wide_48x96": { + "image_grid_thw": { + "max": 6.0, + "mean": 3.66667, + "min": 1.0, + "shape": [ + 1, + 3 + ], + "std": 2.0548 + }, + "pixel_values": { + "max": 2.1459, + "mean": 0.18232, + "min": -1.79226, + "shape": [ + 24, + 1176 + ], + "std": 0.82587 + } + } + }, + "Mistral3ImageProcessor": { + "large_300x500": { + "image_sizes": { + "max": 504.0, + "mean": 406.0, + "min": 308.0, "shape": [ 1, - 2520, 2 ], - "std": 18.76858 + "std": 98.0 }, - "num_soft_tokens_per_image": { - "max": 253.0, - "mean": 253.0, - "min": 253.0, + "pixel_values": { + "max": 2.1459, + "mean": 0.18024, + "min": -1.79226, "shape": [ - 1 + 1, + 308, + 504, + 3 ], - "std": 0.0 + "std": 0.88065 + } + }, + "odd_223x225": { + "image_sizes": { + "max": 238.0, + "mean": 231.0, + "min": 224.0, + "shape": [ + 1, + 2 + ], + "std": 7.0 }, "pixel_values": { - "max": 1.0, - "mean": 0.45048, - "min": 0.0, + "max": 2.1459, + "mean": 0.17866, + "min": -1.79226, "shape": [ 1, - 2520, - 768 + 224, + 238, + 3 ], - "std": 0.26824 + "std": 0.87866 } }, - "wide_48x96": { - "image_position_ids": { - "max": 68.0, - "mean": 22.49286, - "min": -1.0, + "tall_96x48": { + "image_sizes": { + "max": 98.0, + "mean": 77.0, + "min": 56.0, "shape": [ 1, - 2520, 2 ], - "std": 18.76858 + "std": 21.0 }, - "num_soft_tokens_per_image": { - "max": 253.0, - "mean": 253.0, - "min": 253.0, + "pixel_values": { + "max": 2.1459, + "mean": 0.18241, + "min": -1.79226, "shape": [ - 1 + 1, + 98, + 56, + 3 ], - "std": 0.0 + "std": 0.87886 + } + }, + "wide_48x96": { + "image_sizes": { + "max": 98.0, + "mean": 77.0, + "min": 56.0, + "shape": [ + 1, + 2 + ], + "std": 21.0 }, "pixel_values": { - "max": 1.0, - "mean": 0.45044, - "min": 0.0, + "max": 2.1459, + "mean": 0.18237, + "min": -1.79226, "shape": [ 1, - 2520, - 768 + 56, + 98, + 3 ], - "std": 0.26779 + "std": 0.87719 } } }, - "Gemma4UnifiedImageProcessor": { + "MobileNetV2ImageProcessor": { "large_300x500": { - "image_position_ids": { - "max": 20.0, - "mean": 6.875, - "min": -1.0, + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 280, - 2 + 224, + 224, + 3 ], - "std": 5.77143 - }, - "num_soft_tokens_per_image": { - "max": 252.0, - "mean": 252.0, - "min": 252.0, + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ - 1 + 1, + 224, + 224, + 3 ], - "std": 0.0 - }, + "std": 1.04603 + } + }, + "tall_96x48": { "pixel_values": { - "max": 1.0, - "mean": 0.44821, - "min": 0.0, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 280, - 6912 + 224, + 224, + 3 ], - "std": 0.268 + "std": 1.05487 } }, - "odd_223x225": { - "image_position_ids": { - "max": 15.0, - "mean": 6.77143, - "min": -1.0, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 280, - 2 + 224, + 224, + 3 ], - "std": 5.00905 - }, - "num_soft_tokens_per_image": { - "max": 256.0, - "mean": 256.0, - "min": 256.0, + "std": 1.05204 + } + } + }, + "MobileNetV3ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.96529, + "mean": -0.00404, + "min": -0.98002, "shape": [ - 1 + 1, + 224, + 224, + 3 ], - "std": 0.0 - }, + "std": 0.27224 + } + }, + "odd_223x225": { "pixel_values": { - "max": 1.0, - "mean": 0.45492, - "min": 0.0, + "max": 1.20094, + "mean": -0.00492, + "min": -1.23322, "shape": [ 1, - 280, - 6912 + 224, + 224, + 3 ], - "std": 0.26392 + "std": 0.46871 } }, "tall_96x48": { - "image_position_ids": { - "max": 22.0, - "mean": 7.13214, - "min": -1.0, + "pixel_values": { + "max": 1.24145, + "mean": -0.00304, + "min": -1.36308, "shape": [ 1, - 280, - 2 + 224, + 224, + 3 ], - "std": 6.29175 - }, - "num_soft_tokens_per_image": { - "max": 253.0, - "mean": 253.0, - "min": 253.0, + "std": 0.47369 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27045, + "mean": -0.00303, + "min": -1.26931, "shape": [ - 1 + 1, + 224, + 224, + 3 ], - "std": 0.0 - }, + "std": 0.47238 + } + } + }, + "MobileNetV4ImageProcessor": { + "large_300x500": { "pixel_values": { - "max": 1.0, - "mean": 0.45048, - "min": 0.0, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 280, - 6912 + 224, + 224, + 3 ], - "std": 0.26824 + "std": 0.61972 } }, - "wide_48x96": { - "image_position_ids": { - "max": 22.0, - "mean": 7.13214, - "min": -1.0, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 280, - 2 + 224, + 224, + 3 ], - "std": 6.29175 - }, - "num_soft_tokens_per_image": { - "max": 253.0, - "mean": 253.0, - "min": 253.0, + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ - 1 + 1, + 224, + 224, + 3 ], - "std": 0.0 - }, + "std": 1.05487 + } + }, + "wide_48x96": { "pixel_values": { - "max": 1.0, - "mean": 0.45044, - "min": 0.0, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 280, - 6912 + 224, + 224, + 3 ], - "std": 0.26779 + "std": 1.05204 } } }, - "Glm4vImageProcessor": { + "MobileViTImageProcessor": { "large_300x500": { - "image_grid_thw": { - "max": 36.0, - "mean": 19.66667, - "min": 1.0, + "pixel_values": { + "max": 0.98039, + "mean": 0.49863, + "min": 0.00392, "shape": [ 1, + 256, + 256, 3 ], - "std": 14.38363 - }, + "std": 0.18565 + } + }, + "odd_223x225": { "pixel_values": { - "max": 2.1459, - "mean": 0.18024, - "min": -1.79226, + "max": 0.98824, + "mean": 0.4975, + "min": 0.0, "shape": [ - 792, - 1176 + 1, + 256, + 256, + 3 ], - "std": 0.88065 + "std": 0.19257 } }, - "odd_223x225": { - "image_grid_thw": { - "max": 16.0, - "mean": 11.0, - "min": 1.0, + "tall_96x48": { + "pixel_values": { + "max": 0.98039, + "mean": 0.49889, + "min": 0.00784, "shape": [ 1, + 256, + 256, 3 ], - "std": 7.07107 - }, + "std": 0.19272 + } + }, + "wide_48x96": { "pixel_values": { - "max": 2.1459, - "mean": 0.17867, - "min": -1.79226, + "max": 0.97647, + "mean": 0.49557, + "min": 0.01569, + "shape": [ + 1, + 256, + 256, + 3 + ], + "std": 0.19092 + } + } + }, + "MobileViTV2ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.98039, + "mean": 0.49863, + "min": 0.00392, "shape": [ + 1, 256, - 1176 + 256, + 3 ], - "std": 0.8767 + "std": 0.18565 } }, - "tall_96x48": { - "image_grid_thw": { - "max": 12.0, - "mean": 6.33333, - "min": 1.0, + "odd_223x225": { + "pixel_values": { + "max": 0.98824, + "mean": 0.4975, + "min": 0.0, "shape": [ 1, + 256, + 256, 3 ], - "std": 4.49691 - }, + "std": 0.19257 + } + }, + "tall_96x48": { "pixel_values": { - "max": 2.1459, - "mean": 0.18258, - "min": -1.79226, + "max": 0.98039, + "mean": 0.49889, + "min": 0.00784, "shape": [ - 72, - 1176 + 1, + 256, + 256, + 3 ], - "std": 0.8836 + "std": 0.19272 } }, "wide_48x96": { - "image_grid_thw": { - "max": 12.0, - "mean": 6.33333, - "min": 1.0, + "pixel_values": { + "max": 0.97647, + "mean": 0.49557, + "min": 0.01569, "shape": [ 1, + 256, + 256, 3 ], - "std": 4.49691 - }, - "pixel_values": { - "max": 2.1459, - "mean": 0.18233, - "min": -1.79226, - "shape": [ - 72, - 1176 - ], - "std": 0.88152 + "std": 0.19092 } } }, - "GroundingDinoImageProcessor": { + "NextViTImageProcessor": { "large_300x500": { - "pixel_mask": { - "max": 1.0, - "mean": 1.0, - "min": 1.0, - "shape": [ - 1, - 800, - 1333 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.62257, - "mean": 0.21811, - "min": -2.10078, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 800, - 1333, + 224, + 224, 3 ], - "std": 0.86341 + "std": 0.61972 } }, "odd_223x225": { - "pixel_mask": { - "max": 1.0, - "mean": 1.0, - "min": 1.0, - "shape": [ - 1, - 800, - 807 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.62257, - "mean": 0.2156, - "min": -2.1179, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 800, - 807, + 224, + 224, 3 ], - "std": 0.86402 + "std": 1.04603 } }, "tall_96x48": { - "pixel_mask": { - "max": 1.0, - "mean": 1.0, - "min": 1.0, - "shape": [ - 1, - 1333, - 666 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.62257, - "mean": 0.21961, - "min": -2.1179, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 1333, - 666, + 224, + 224, 3 ], - "std": 0.86992 + "std": 1.05487 } }, "wide_48x96": { - "pixel_mask": { - "max": 1.0, - "mean": 1.0, - "min": 1.0, - "shape": [ - 1, - 666, - 1333 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.60514, - "mean": 0.21961, - "min": -2.10078, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 666, - 1333, + 224, + 224, 3 ], - "std": 0.86695 + "std": 1.05204 } } }, - "InternVLImageProcessor": { + "OneFormerImageProcessor": { "large_300x500": { - "num_patches": { - "max": 7.0, - "mean": 7.0, - "min": 7.0, - "shape": [ - 1 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.64, - "mean": 0.21774, + "max": 2.60677, + "mean": -0.6647, "min": -2.1179, "shape": [ - 7, - 448, - 448, + 1, + 512, + 512, 3 ], - "std": 1.04048 + "std": 1.27298 } }, "odd_223x225": { - "num_patches": { - "max": 1.0, - "mean": 1.0, - "min": 1.0, - "shape": [ - 1 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.64, - "mean": 0.21573, + "max": 2.61896, + "mean": 0.19412, "min": -2.1179, "shape": [ 1, - 448, - 448, + 512, + 512, 3 ], - "std": 1.04644 + "std": 0.88675 } }, "tall_96x48": { - "num_patches": { - "max": 3.0, - "mean": 3.0, - "min": 3.0, - "shape": [ - 1 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.64, - "mean": 0.21988, + "max": 2.58792, + "mean": -0.88321, "min": -2.1179, "shape": [ - 3, - 448, - 448, + 1, + 512, + 512, 3 ], - "std": 1.0509 + "std": 1.26595 } }, "wide_48x96": { - "num_patches": { - "max": 3.0, - "mean": 3.0, - "min": 3.0, - "shape": [ - 1 - ], - "std": 0.0 - }, "pixel_values": { - "max": 2.64, - "mean": 0.21974, + "max": 2.57913, + "mean": -0.8832, "min": -2.1179, "shape": [ - 3, - 448, - 448, + 1, + 512, + 512, 3 ], - "std": 1.04848 + "std": 1.26493 } } }, - "JanusImageProcessor": { + "OwlViTImageProcessor": { "large_300x500": { "pixel_values": { - "max": 1.11482, - "mean": -0.00399, - "min": -1.15907, + "max": 2.63783, + "mean": 0.18019, + "min": -2.28343, "shape": [ 1, - 384, - 384, + 768, + 768, 3 ], - "std": 0.27936 + "std": 0.88346 } }, "odd_223x225": { "pixel_values": { - "max": 1.23231, - "mean": -0.0049, - "min": -1.24259, + "max": 2.60321, + "mean": 0.17857, + "min": -2.31688, "shape": [ 1, - 384, - 384, + 768, + 768, 3 ], - "std": 0.46848 + "std": 0.88212 } }, - "tall_96x48": { - "pixel_values": { - "max": 1.24756, - "mean": -0.00348, - "min": -1.37134, + "tall_96x48": { + "pixel_values": { + "max": 2.53206, + "mean": 0.18215, + "min": -2.28856, "shape": [ 1, - 384, - 384, + 768, + 768, 3 ], - "std": 0.33484 + "std": 0.88677 } }, "wide_48x96": { "pixel_values": { - "max": 1.27612, - "mean": -0.00348, - "min": -1.28159, + "max": 2.68594, + "mean": 0.18216, + "min": -2.23357, "shape": [ 1, - 384, - 384, + 768, + 768, 3 ], - "std": 0.33391 + "std": 0.88441 } } }, - "KimiK25ImageProcessor": { + "Owlv2ImageProcessor": { "large_300x500": { - "image_grid_thw": { - "max": 36.0, - "mean": 19.66667, - "min": 1.0, + "pixel_values": { + "max": 2.56833, + "mean": -0.56184, + "min": -2.34171, "shape": [ 1, + 960, + 960, 3 ], - "std": 14.38363 - }, - "pixel_values": { - "max": 0.99216, - "mean": -0.0376, - "min": -1.0, - "shape": [ - 792, - 3, - 14, - 14 - ], - "std": 0.59538 + "std": 1.14068 } }, "odd_223x225": { - "image_grid_thw": { - "max": 18.0, - "mean": 11.66667, - "min": 1.0, + "pixel_values": { + "max": 2.62367, + "mean": 0.16211, + "min": -2.34614, "shape": [ 1, + 960, + 960, 3 ], - "std": 7.58654 - }, - "pixel_values": { - "max": 0.99216, - "mean": -0.1155, - "min": -1.0, - "shape": [ - 288, - 3, - 14, - 14 - ], - "std": 0.62716 + "std": 0.89456 } }, "tall_96x48": { - "image_grid_thw": { - "max": 8.0, - "mean": 4.33333, - "min": 1.0, + "pixel_values": { + "max": 2.52769, + "mean": -0.74635, + "min": -2.29293, "shape": [ 1, + 960, + 960, 3 ], - "std": 2.86744 - }, - "pixel_values": { - "max": 0.99216, - "mean": -0.26755, - "min": -1.0, - "shape": [ - 32, - 3, - 14, - 14 - ], - "std": 0.66262 + "std": 1.12304 } }, "wide_48x96": { - "image_grid_thw": { - "max": 8.0, - "mean": 4.33333, - "min": 1.0, + "pixel_values": { + "max": 2.67747, + "mean": -0.74634, + "min": -2.23153, "shape": [ 1, + 960, + 960, 3 ], - "std": 2.86744 - }, - "pixel_values": { - "max": 0.99216, - "mean": -0.26755, - "min": -1.0, - "shape": [ - 32, - 3, - 14, - 14 - ], - "std": 0.66262 + "std": 1.12234 } } }, - "LocateAnythingImageProcessor": { + "PiTImageProcessor": { "large_300x500": { - "image_grid_hws": { - "max": 36.0, - "mean": 29.0, - "min": 22.0, - "shape": [ - 1, - 2 - ], - "std": 7.0 - }, "pixel_values": { - "max": 1.0, - "mean": -0.004, - "min": -1.0, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ - 792, - 3, - 14, - 14 + 1, + 224, + 224, + 3 ], - "std": 0.46899 + "std": 0.61972 } }, "odd_223x225": { - "image_grid_hws": { - "max": 18.0, - "mean": 17.0, - "min": 16.0, - "shape": [ - 1, - 2 - ], - "std": 1.0 - }, "pixel_values": { - "max": 1.0, - "mean": -0.00483, - "min": -1.0, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ - 288, - 3, - 14, - 14 + 1, + 224, + 224, + 3 ], - "std": 0.46811 + "std": 1.04603 } }, "tall_96x48": { - "image_grid_hws": { - "max": 8.0, - "mean": 6.0, - "min": 4.0, - "shape": [ - 1, - 2 - ], - "std": 2.0 - }, "pixel_values": { - "max": 1.0, - "mean": -0.00286, - "min": -1.0, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ - 32, - 3, - 14, - 14 + 1, + 224, + 224, + 3 ], - "std": 0.47032 + "std": 1.05487 } }, "wide_48x96": { - "image_grid_hws": { - "max": 8.0, - "mean": 6.0, - "min": 4.0, - "shape": [ - 1, - 2 - ], - "std": 2.0 - }, "pixel_values": { - "max": 1.0, - "mean": -0.00292, - "min": -1.0, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ - 32, - 3, - 14, - 14 + 1, + 224, + 224, + 3 ], - "std": 0.46883 + "std": 1.05204 } } }, - "Mask2FormerImageProcessor": { + "PoolFormerImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.61525, - "mean": -0.66613, - "min": -2.1179, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 384, - 384, + 224, + 224, 3 ], - "std": 1.27301 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.61651, - "mean": 0.19259, - "min": -2.1179, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 384, - 384, + 224, + 224, 3 ], - "std": 0.88845 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.52889, - "mean": -0.88322, - "min": -2.1179, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 384, - 384, + 224, + 224, 3 ], - "std": 1.26184 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.55503, - "mean": -0.88322, - "min": -2.1179, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 384, - 384, + 224, + 224, 3 ], - "std": 1.26078 + "std": 1.05204 } } }, - "MaskFormerImageProcessor": { + "PvtImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.60677, - "mean": -0.6647, - "min": -2.1179, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ - 1, - 512, - 512, + 1, + 224, + 224, 3 ], - "std": 1.27298 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.61896, - "mean": 0.19412, - "min": -2.1179, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 512, - 512, + 224, + 224, 3 ], - "std": 0.88675 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.58792, - "mean": -0.88321, - "min": -2.1179, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 512, - 512, + 224, + 224, 3 ], - "std": 1.26595 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.57913, - "mean": -0.8832, - "min": -2.1179, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 512, - 512, + 224, + 224, 3 ], - "std": 1.26493 + "std": 1.05204 } } }, - "MetaClip2ImageProcessor": { + "PvtV2ImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.0748, - "mean": 0.18018, - "min": -1.66088, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, 224, 224, 3 ], - "std": 0.51898 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.1459, - "mean": 0.17867, - "min": -1.79226, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, 224, 224, 3 ], - "std": 0.8767 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.1459, - "mean": 0.18242, - "min": -1.79226, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, 224, 224, 3 ], - "std": 0.88356 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.1459, - "mean": 0.18227, - "min": -1.79226, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, 224, 224, 3 ], - "std": 0.88139 + "std": 1.05204 } } }, - "MiniMaxM3VLImageProcessor": { + "Qwen2VLImageProcessor": { "large_300x500": { "image_grid_thw": { "max": 36.0, @@ -2209,655 +4311,525 @@ } } }, - "Mistral3ImageProcessor": { + "RFDETRImageProcessor": { "large_300x500": { - "image_sizes": { - "max": 504.0, - "mean": 406.0, - "min": 308.0, - "shape": [ - 1, - 2 - ], - "std": 98.0 - }, "pixel_values": { - "max": 2.1459, - "mean": 0.18024, - "min": -1.79226, + "max": 2.60693, + "mean": 0.21768, + "min": -2.11403, "shape": [ 1, - 308, - 504, + 560, + 560, 3 ], - "std": 0.88065 + "std": 0.86464 } }, "odd_223x225": { - "image_sizes": { - "max": 238.0, - "mean": 231.0, - "min": 224.0, - "shape": [ - 1, - 2 - ], - "std": 7.0 - }, "pixel_values": { - "max": 2.1459, - "mean": 0.17866, - "min": -1.79226, + "max": 2.61792, + "mean": 0.21558, + "min": -2.11309, "shape": [ 1, - 224, - 238, + 560, + 560, 3 ], - "std": 0.87866 + "std": 0.8639 } }, "tall_96x48": { - "image_sizes": { - "max": 98.0, - "mean": 77.0, - "min": 56.0, - "shape": [ - 1, - 2 - ], - "std": 21.0 - }, "pixel_values": { - "max": 2.1459, - "mean": 0.18241, - "min": -1.79226, + "max": 2.62133, + "mean": 0.21955, + "min": -2.11423, "shape": [ 1, - 98, - 56, + 560, + 560, 3 ], - "std": 0.87886 + "std": 0.87004 } }, "wide_48x96": { - "image_sizes": { - "max": 98.0, - "mean": 77.0, - "min": 56.0, - "shape": [ - 1, - 2 - ], - "std": 21.0 - }, "pixel_values": { - "max": 2.1459, - "mean": 0.18237, - "min": -1.79226, + "max": 2.60663, + "mean": 0.21956, + "min": -2.10763, "shape": [ 1, - 56, - 98, + 560, + 560, 3 ], - "std": 0.87719 + "std": 0.86707 } } }, - "MobileViTImageProcessor": { + "RTDETRImageProcessor": { "large_300x500": { "pixel_values": { - "max": 0.98039, - "mean": 0.49863, - "min": 0.00392, + "max": 0.99528, + "mean": 0.49798, + "min": 0.00313, "shape": [ 1, - 256, - 256, + 640, + 640, 3 ], - "std": 0.18565 + "std": 0.1927 } }, "odd_223x225": { "pixel_values": { - "max": 0.98824, - "mean": 0.4975, - "min": 0.0, + "max": 0.99564, + "mean": 0.49754, + "min": 0.00016, "shape": [ 1, - 256, - 256, + 640, + 640, 3 ], - "std": 0.19257 + "std": 0.19273 } }, "tall_96x48": { "pixel_values": { - "max": 0.98039, - "mean": 0.49889, - "min": 0.00784, + "max": 0.99247, + "mean": 0.49847, + "min": 0.00204, "shape": [ 1, - 256, - 256, + 640, + 640, 3 ], - "std": 0.19272 + "std": 0.19463 } }, "wide_48x96": { "pixel_values": { - "max": 0.97647, - "mean": 0.49557, - "min": 0.01569, - "shape": [ - 1, - 256, - 256, - 3 - ], - "std": 0.19092 - } - } - }, - "MobileViTV2ImageProcessor": { - "large_300x500": { - "pixel_values": { - "max": 0.98039, - "mean": 0.49863, - "min": 0.00392, - "shape": [ - 1, - 256, - 256, - 3 - ], - "std": 0.18565 - } - }, - "odd_223x225": { - "pixel_values": { - "max": 0.98824, - "mean": 0.4975, + "max": 0.99063, + "mean": 0.49848, "min": 0.0, "shape": [ 1, - 256, - 256, - 3 - ], - "std": 0.19257 - } - }, - "tall_96x48": { - "pixel_values": { - "max": 0.98039, - "mean": 0.49889, - "min": 0.00784, - "shape": [ - 1, - 256, - 256, - 3 - ], - "std": 0.19272 - } - }, - "wide_48x96": { - "pixel_values": { - "max": 0.97647, - "mean": 0.49557, - "min": 0.01569, - "shape": [ - 1, - 256, - 256, + 640, + 640, 3 ], - "std": 0.19092 + "std": 0.19393 } } }, - "OneFormerImageProcessor": { - "large_300x500": { - "pixel_values": { - "max": 2.60677, - "mean": -0.6647, - "min": -2.1179, + "RTDETRV2ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.99528, + "mean": 0.49798, + "min": 0.00313, "shape": [ 1, - 512, - 512, + 640, + 640, 3 ], - "std": 1.27298 + "std": 0.1927 } }, "odd_223x225": { "pixel_values": { - "max": 2.61896, - "mean": 0.19412, - "min": -2.1179, + "max": 0.99564, + "mean": 0.49754, + "min": 0.00016, "shape": [ 1, - 512, - 512, + 640, + 640, 3 ], - "std": 0.88675 + "std": 0.19273 } }, "tall_96x48": { "pixel_values": { - "max": 2.58792, - "mean": -0.88321, - "min": -2.1179, + "max": 0.99247, + "mean": 0.49847, + "min": 0.00204, "shape": [ 1, - 512, - 512, + 640, + 640, 3 ], - "std": 1.26595 + "std": 0.19463 } }, "wide_48x96": { "pixel_values": { - "max": 2.57913, - "mean": -0.8832, - "min": -2.1179, + "max": 0.99063, + "mean": 0.49848, + "min": 0.0, "shape": [ 1, - 512, - 512, + 640, + 640, 3 ], - "std": 1.26493 + "std": 0.19393 } } }, - "OwlViTImageProcessor": { + "RegNetImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.63783, - "mean": 0.18019, - "min": -2.28343, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.88346 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.60321, - "mean": 0.17857, - "min": -2.31688, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.88212 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.53206, - "mean": 0.18215, - "min": -2.28856, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.88677 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.68594, - "mean": 0.18216, - "min": -2.23357, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 768, - 768, + 224, + 224, 3 ], - "std": 0.88441 + "std": 1.05204 } } }, - "Owlv2ImageProcessor": { + "Res2NetImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.56833, - "mean": -0.56184, - "min": -2.34171, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 960, - 960, + 224, + 224, 3 ], - "std": 1.14068 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.62367, - "mean": 0.16211, - "min": -2.34614, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 960, - 960, + 224, + 224, 3 ], - "std": 0.89456 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.52769, - "mean": -0.74635, - "min": -2.29293, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 960, - 960, + 224, + 224, 3 ], - "std": 1.12304 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.67747, - "mean": -0.74634, - "min": -2.23153, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 960, - 960, + 224, + 224, 3 ], - "std": 1.12234 + "std": 1.05204 } } }, - "Qwen2VLImageProcessor": { + "ResMLPImageProcessor": { "large_300x500": { - "image_grid_thw": { - "max": 36.0, - "mean": 19.66667, - "min": 1.0, + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, + 224, + 224, 3 ], - "std": 14.38363 - }, - "pixel_values": { - "max": 2.1459, - "mean": 0.18024, - "min": -1.79226, - "shape": [ - 792, - 1176 - ], - "std": 0.88065 + "std": 0.61972 } }, "odd_223x225": { - "image_grid_thw": { - "max": 16.0, - "mean": 11.0, - "min": 1.0, + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, + 224, + 224, 3 ], - "std": 7.07107 - }, - "pixel_values": { - "max": 2.1459, - "mean": 0.17867, - "min": -1.79226, - "shape": [ - 256, - 1176 - ], - "std": 0.8767 + "std": 1.04603 } }, "tall_96x48": { - "image_grid_thw": { - "max": 6.0, - "mean": 3.66667, - "min": 1.0, + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, + 224, + 224, 3 ], - "std": 2.0548 - }, - "pixel_values": { - "max": 2.1459, - "mean": 0.18243, - "min": -1.79226, - "shape": [ - 24, - 1176 - ], - "std": 0.82552 + "std": 1.05487 } }, "wide_48x96": { - "image_grid_thw": { - "max": 6.0, - "mean": 3.66667, - "min": 1.0, + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, + 224, + 224, 3 ], - "std": 2.0548 - }, - "pixel_values": { - "max": 2.1459, - "mean": 0.18232, - "min": -1.79226, - "shape": [ - 24, - 1176 - ], - "std": 0.82587 + "std": 1.05204 } } }, - "RFDETRImageProcessor": { + "ResNeXtImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.60693, - "mean": 0.21768, - "min": -2.11403, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 560, - 560, + 224, + 224, 3 ], - "std": 0.86464 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.61792, - "mean": 0.21558, - "min": -2.11309, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 560, - 560, + 224, + 224, 3 ], - "std": 0.8639 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.62133, - "mean": 0.21955, - "min": -2.11423, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 560, - 560, + 224, + 224, 3 ], - "std": 0.87004 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.60663, - "mean": 0.21956, - "min": -2.10763, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 560, - 560, + 224, + 224, 3 ], - "std": 0.86707 + "std": 1.05204 } } }, - "RTDETRImageProcessor": { + "ResNetImageProcessor": { "large_300x500": { "pixel_values": { - "max": 0.99528, - "mean": 0.49798, - "min": 0.00313, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.1927 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 0.99564, - "mean": 0.49754, - "min": 0.00016, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.19273 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 0.99247, - "mean": 0.49847, - "min": 0.00204, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.19463 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 0.99063, - "mean": 0.49848, - "min": 0.0, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.19393 + "std": 1.05204 } } }, - "RTDETRV2ImageProcessor": { + "ResNetV2ImageProcessor": { "large_300x500": { "pixel_values": { - "max": 0.99528, - "mean": 0.49798, - "min": 0.00313, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.1927 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 0.99564, - "mean": 0.49754, - "min": 0.00016, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.19273 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 0.99247, - "mean": 0.49847, - "min": 0.00204, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.19463 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 0.99063, - "mean": 0.49848, - "min": 0.0, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 640, - 640, + 224, + 224, 3 ], - "std": 0.19393 + "std": 1.05204 } } }, @@ -3277,183 +5249,415 @@ "mean": 768.0, "min": 512.0, "shape": [ - 2 + 2 + ], + "std": 256.0 + } + } + }, + "SENetImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "SegFormerImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.61077, + "mean": 0.21756, + "min": -2.11253, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 0.86476 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 2.61913, + "mean": 0.21549, + "min": -2.11083, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 0.86406 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 2.59833, + "mean": 0.21957, + "min": -2.09623, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 0.86939 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 2.60194, + "mean": 0.2196, + "min": -2.08591, + "shape": [ + 1, + 512, + 512, + 3 + ], + "std": 0.8664 + } + } + }, + "SigLIP2ImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.96529, + "mean": -0.00404, + "min": -0.98002, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.27224 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.20094, + "mean": -0.00492, + "min": -1.23322, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.46871 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24145, + "mean": -0.00304, + "min": -1.36308, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47369 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27045, + "mean": -0.00303, + "min": -1.26931, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47238 + } + } + }, + "SigLIPImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.96529, + "mean": -0.00404, + "min": -0.98002, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.27224 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.20094, + "mean": -0.00492, + "min": -1.23322, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.46871 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24145, + "mean": -0.00304, + "min": -1.36308, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47369 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27045, + "mean": -0.00303, + "min": -1.26931, + "shape": [ + 1, + 224, + 224, + 3 ], - "std": 256.0 + "std": 0.47238 } } }, - "SegFormerImageProcessor": { + "SwinImageProcessor": { "large_300x500": { "pixel_values": { - "max": 2.61077, - "mean": 0.21756, - "min": -2.11253, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, - 512, - 512, + 224, + 224, 3 ], - "std": 0.86476 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 2.61913, - "mean": 0.21549, - "min": -2.11083, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, - 512, - 512, + 224, + 224, 3 ], - "std": 0.86406 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 2.59833, - "mean": 0.21957, - "min": -2.09623, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, - 512, - 512, + 224, + 224, 3 ], - "std": 0.86939 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 2.60194, - "mean": 0.2196, - "min": -2.08591, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, - 512, - 512, + 224, + 224, 3 ], - "std": 0.8664 + "std": 1.05204 } } }, - "SigLIP2ImageProcessor": { + "SwinV2ImageProcessor": { "large_300x500": { "pixel_values": { - "max": 0.96529, - "mean": -0.00404, - "min": -0.98002, + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, "shape": [ 1, 224, 224, 3 ], - "std": 0.27224 + "std": 0.61972 } }, "odd_223x225": { "pixel_values": { - "max": 1.20094, - "mean": -0.00492, - "min": -1.23322, + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, "shape": [ 1, 224, 224, 3 ], - "std": 0.46871 + "std": 1.04603 } }, "tall_96x48": { "pixel_values": { - "max": 1.24145, - "mean": -0.00304, - "min": -1.36308, + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, "shape": [ 1, 224, 224, 3 ], - "std": 0.47369 + "std": 1.05487 } }, "wide_48x96": { "pixel_values": { - "max": 1.27045, - "mean": -0.00303, - "min": -1.26931, + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, "shape": [ 1, 224, 224, 3 ], - "std": 0.47238 + "std": 1.05204 } } }, - "SigLIPImageProcessor": { + "TableTransformerImageProcessor": { "large_300x500": { "pixel_values": { - "max": 0.96529, - "mean": -0.00404, - "min": -0.98002, + "max": 2.60739, + "mean": 0.21759, + "min": -2.10553, "shape": [ 1, - 224, - 224, + 800, + 800, 3 ], - "std": 0.27224 + "std": 0.86184 } }, "odd_223x225": { "pixel_values": { - "max": 1.20094, - "mean": -0.00492, - "min": -1.23322, + "max": 2.62243, + "mean": 0.21562, + "min": -2.1155, "shape": [ 1, - 224, - 224, + 800, + 800, 3 ], - "std": 0.46871 + "std": 0.86389 } }, "tall_96x48": { "pixel_values": { - "max": 1.24145, - "mean": -0.00304, - "min": -1.36308, + "max": 2.61699, + "mean": 0.21958, + "min": -2.10523, "shape": [ 1, - 224, - 224, + 800, + 800, 3 ], - "std": 0.47369 + "std": 0.87022 } }, "wide_48x96": { "pixel_values": { - "max": 1.27045, - "mean": -0.00303, - "min": -1.26931, + "max": 2.62048, + "mean": 0.21955, + "min": -2.11123, "shape": [ 1, - 224, - 224, + 800, + 800, 3 ], - "std": 0.47238 + "std": 0.86719 } } }, @@ -3572,5 +5776,179 @@ "std": 0.19385 } } + }, + "VGGImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 2.56287, + "mean": 0.2176, + "min": -1.96277, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.61972 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 3.08654, + "mean": 0.21557, + "min": -2.62713, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.04603 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 3.0413, + "mean": 0.2196, + "min": -2.62093, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05487 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 3.241, + "mean": 0.21962, + "min": -2.61223, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 1.05204 + } + } + }, + "ViTImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 0.96529, + "mean": -0.00404, + "min": -0.98002, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.27224 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.20094, + "mean": -0.00492, + "min": -1.23322, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.46871 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24145, + "mean": -0.00304, + "min": -1.36308, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47369 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.27045, + "mean": -0.00303, + "min": -1.26931, + "shape": [ + 1, + 224, + 224, + 3 + ], + "std": 0.47238 + } + } + }, + "XceptionImageProcessor": { + "large_300x500": { + "pixel_values": { + "max": 1.11938, + "mean": -0.00404, + "min": -1.11777, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.36307 + } + }, + "odd_223x225": { + "pixel_values": { + "max": 1.25057, + "mean": -0.00492, + "min": -1.20156, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47012 + } + }, + "tall_96x48": { + "pixel_values": { + "max": 1.24844, + "mean": -0.00304, + "min": -1.37676, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47368 + } + }, + "wide_48x96": { + "pixel_values": { + "max": 1.26429, + "mean": -0.00304, + "min": -1.27812, + "shape": [ + 1, + 299, + 299, + 3 + ], + "std": 0.47237 + } + } } } From 1b43624067799d0c06c760337cb7485b6fb11274 Mon Sep 17 00:00:00 2001 From: IMvision12 Date: Sun, 30 Aug 2026 17:24:42 -0700 Subject: [PATCH 4/5] fix --- tests/integration/test_data_formats.py | 28 +++++++++++++++++++ .../efficientdet/efficientdet_layers.py | 23 +++++++++++++-- .../models/efficientdet/efficientdet_model.py | 11 +++----- 3 files changed, 52 insertions(+), 10 deletions(-) diff --git a/tests/integration/test_data_formats.py b/tests/integration/test_data_formats.py index 343c790d..8833e067 100644 --- a/tests/integration/test_data_formats.py +++ b/tests/integration/test_data_formats.py @@ -98,6 +98,30 @@ def _transpose_input(input_data, data_format): return input_data +def _randomize_images(input_data, seed=0): + """Replace all-ones image tensors with fixed-seed random pixels for parity. + + ``create_test_input`` feeds ones. A near-uniform image gives normalization + layers (GroupNorm / LayerNorm) ~0 variance, so ``1/sqrt(var + eps)`` amplifies + the tiny cross-layout float noise (~1e-7) into a false parity failure -- seen + on OneFormer under jax, where its pixel-decoder GroupNorm blew a 1e-7 conv + difference up to rel ~0.5. Real pixels give the layers real variance, so the + check reflects the layout math, not noise amplification. Non-image inputs + (token ids, grids, masks) are left untouched. + """ + rng = np.random.RandomState(seed) + + def rnd(v): + return ops.convert_to_tensor(rng.standard_normal(v.shape).astype("float32")) + + if isinstance(input_data, dict): + return { + k: (rnd(v) if _is_channels_last_image(k, v) else v) + for k, v in input_data.items() + } + return rnd(input_data) if _is_channels_last_image(None, input_data) else input_data + + @pytest.mark.data_format @pytest.mark.parametrize("model_name", MODEL_IDS) def test_channels_last(model_name): @@ -257,6 +281,10 @@ def test_channels_first_matches_channels_last(model_name): input_data = create_test_input(config, model=model_cl) else: input_data = create_test_input(config) + # Use real (random) pixels, not the ones from create_test_input: a uniform + # image gives normalization layers ~0 variance, which amplifies float noise + # into a false parity failure (see _randomize_images). + input_data = _randomize_images(input_data) if not _has_transposable_image(input_data): pytest.skip( diff --git a/zeromodels/models/efficientdet/efficientdet_layers.py b/zeromodels/models/efficientdet/efficientdet_layers.py index 33244f30..27d0192c 100644 --- a/zeromodels/models/efficientdet/efficientdet_layers.py +++ b/zeromodels/models/efficientdet/efficientdet_layers.py @@ -109,6 +109,14 @@ def call(self, feat, target_height, target_width, training=False): ) return feat + def compute_output_spec(self, feat, target_height, target_width, training=False): + batch = feat.shape[0] + if self.data_format == "channels_first": + shape = (batch, self.target_channels, target_height, target_width) + else: + shape = (batch, target_height, target_width, self.target_channels) + return keras.KerasTensor(shape=shape, dtype=self.compute_dtype) + def get_config(self): config = super().get_config() config.update( @@ -335,6 +343,18 @@ def call(self, feats, level_sizes, training=False): break return feats + def compute_output_spec(self, feats, level_sizes, training=False): + batch = feats[0].shape[0] + specs = [] + for lvl in range(self.max_level - self.min_level + 1): + th, tw = level_sizes[lvl] + if self.data_format == "channels_first": + shape = (batch, self.fpn_num_filters, th, tw) + else: + shape = (batch, th, tw, self.fpn_num_filters) + specs.append(keras.KerasTensor(shape=shape, dtype=self.compute_dtype)) + return specs + @keras.saving.register_keras_serializable(package="zeromodels") class PredictionHead(layers.Layer): @@ -546,9 +566,6 @@ def build(self, input_shape): self.built = True def call(self, box_outputs): - # Anchors are a fixed function of the config, so they are baked into the graph - # as a constant rather than stored as a weight. This keeps EfficientDetDetect's - # weight set identical to EfficientDetModel's, so both load one hosted file. return decode_box_outputs( box_outputs, ops.cast(self.anchors, box_outputs.dtype) ) diff --git a/zeromodels/models/efficientdet/efficientdet_model.py b/zeromodels/models/efficientdet/efficientdet_model.py index 90cec73d..2f67f94d 100644 --- a/zeromodels/models/efficientdet/efficientdet_model.py +++ b/zeromodels/models/efficientdet/efficientdet_model.py @@ -190,6 +190,9 @@ def __init__( ) for attr in CONFIG_ATTRS: setattr(self, attr, locals()[attr]) + self( + keras.ops.zeros((1, *input_shape), dtype=self.compute_dtype), training=False + ) def get_config(self): config = super().get_config() @@ -229,9 +232,7 @@ class and box outputs, applies sigmoid to the class logits, and decodes the box BASE_WEIGHT_CONFIG = None HF_MODEL_TYPE = "efficientdet" config_class = EfficientDetConfig - # EfficientDetDetect shares its weights with EfficientDetModel (identical backbone, - # BiFPN and heads; decoding adds no weights). Hosted repos declare the canonical - # EfficientDetModel, and this class loads that same file by copying weights out. + HUB_REPO_SIBLINGS = frozenset({"EfficientDetModel"}) CHECKPOINT_SOURCE = CheckpointSource("EfficientDetModel") @@ -285,10 +286,6 @@ def __init__( class_outputs = base.output["class_outputs"] box_outputs = base.output["box_outputs"] - # Flatten each level's head output to (B, H*W*anchors, last). Under - # channels_first the head output is (B, anchors*last, H, W); transpose it to - # channels_last first so the anchor ordering (position-major) matches the - # anchor grid regardless of data format -> identical outputs either way. data_format = get_data_format() def flatten_levels(tensors, last, name): From dd5e34043eb8fbfecee5669a481306385bbd999e Mon Sep 17 00:00:00 2001 From: IMvision12 Date: Sun, 30 Aug 2026 17:44:11 -0700 Subject: [PATCH 5/5] torch 2 shards --- .github/workflows/test_code.yml | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/.github/workflows/test_code.yml b/.github/workflows/test_code.yml index ee98e3be..95a6eb39 100644 --- a/.github/workflows/test_code.yml +++ b/.github/workflows/test_code.yml @@ -29,7 +29,8 @@ jobs: # the per-job time cap. Stacks with the per-model build-once reuse. matrix: include: - - { backend: torch, shard: "1/1" } + - { backend: torch, shard: "1/2" } + - { backend: torch, shard: "2/2" } - { backend: tensorflow, shard: "1/3" } - { backend: tensorflow, shard: "2/3" } - { backend: tensorflow, shard: "3/3" }