Gemma3 (#36658)
* Fix converter * [Broken] Adds Gemma 3 to Hugging Face Transformers * Consolidating Config and Processor params across impls * Sorting out configuration parameters. Adds qk_norm before RoPE. Still not sure if RoPE is right. * Additional plumbing for CausalLM and ConditionalGeneration variants * incomplete draft of Orbax conversion script * More complete checkpoint conversion * Supporting Gemma 3 1B checkpoints * Updating RoPE for multiple frequencies * Adjustments to rotary embedder * Proof of life for text-only operation * Updating the conversion script to handle multimodal projection weights * Fixing tet-only conversions * Cleaner conversion script with multimodal support and a simpler processor * Additional refatcors to the Gemma3Processor * Simplified Processor to work over text representations * Updated conversion script to join text and vision embeddings at converion time * Logging for debugging * Update src/transformers/models/gemma2/modeling_gemma2.py Co-authored-by: Joshua Lochner <admin@xenova.com> * Removed extraneous Config params * Switching to fast tokenizer for checkpoint conversions * isolating siglip for performance tetsing * Minor changes for debugging tests against baselines * Adding average pooling for soft tokens * Updating processor code to enable simpler embedding interleaving for arbitrary number of images in prompts * Updating conversion script for ShieldGemma 2 conversion compatibility * Allow disable_compile to be provided as a kwarg * Refresh from modular * Updated conversion script and corrected sliding window * Fix type mismatch in cache_position (#4) * Fix dtype (#5) * Fix type mismatch in cache_position * Actually fix in the modular file Co-authored-by: Aritra Roy Gosthipaty <aritra.born2fly@gmail.com> --------- Co-authored-by: Aritra Roy Gosthipaty <aritra.born2fly@gmail.com> * fixes for embedding table overflow and missing image_soft_token_mask from Gemma3Processor * Adding 2D pooling for image embeddings * Revert "Adding 2D pooling for image embeddings" This reverts commit 65350cf531296f050b2078a5b8e46f61642b2648. * Gemma3 average pooling changed from 1D to 2D * Major refactor to Gemma3MultimodalInputProjection * Updating Gemm 3 Auto* registrations * Add option to save Gemma 3 chat template with tokenizer during weights conversion * Removing unused imports * Moving out-of-vocab handling from Gemma3Processor to Gemma3ForConditionalGeneration * Removing duplicate config property * Removing final logit softcapping and 1-indexing of position ids * Fixing image processor config and none --> None typo * Fixing sliding window size for 1B * Updating image_mean and image_std in Image Processor * Attention masking changed to lower triangular * Moving image special tokens to conversion script * Mirror image processor defaults from conversion script into Gemma3ProcessorKwargs * Remove special token variables from symbol space * Moving image soft token mask computation from Gemma3Processor to Gemma3ForConditionalGeneration * tie lm_head and embedding weights Co-authored-by: Matthew Douglas <38992547+matthewdouglas@users.noreply.github.com> * Correct tied weights in Gemma3CausalLM * iterative bidirectional attention * resolving merge conflicts * Reverting to Gemma 2 HybridCache with sldiing window support and a sliding_window_pattern of 6 * Correcting RoPE scaling * clean up first pass, dummy model geenration works * final clean up before fixing tests * causal lm test works, so fine * Fix conversion * Update src/transformers/models/gemma3/processing_gemma3.py * model tests are happy * processor tests are happy * image processing tests added * fixup * Fix pre-processing in conversion * Inputs merging * Do not normalize vision embeddings * Apply Ryan's (and team) changes to attention * token type ids + mask * template * move embed scale, add rope scale, fix tests * Add chat template to tokenizer * Use prefix for causal model loading * use existing code for sliding mask from gemma2 * self.embed_tokens already normalizes * Correcting Gemma3TextConfig parameters in conversion script * typo, modular overwrites my fixes * enable device map for text model * Conversion updates * ultra nit: no einsums * update image token * copy deepcopy config + some docs * add some test, still WIP * Refactoring --include_chat_tempalte logic in converter * Update src/transformers/models/gemma3/modular_gemma3.py Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com> * Add eos tokens for instruct models * dump so i can work on dgx * Removing add_bos by default * dump * add fast im proc * docs for PaS + fixup * another fixup * one more fixup * fix tests * Inverting prior BOS change * ultra nit * Reverting to Tokenizer saved with add_bos_token=True and chat template starting with BOS * resize embeds, remove sqrt, add slow test outputs * FA2 but quality is meh * nit * skip FA2, no idea what happened * last bit for green CI * please, green CI for docs * T_T * Fix for Gemma3 logits * Support both options for system prompt * Update src/transformers/models/gemma3/image_processing_gemma3_fast.py Co-authored-by: Pedro Cuenca <pedro@huggingface.co> * Update docs/source/en/model_doc/gemma3.md Co-authored-by: Pedro Cuenca <pedro@huggingface.co> * Update docs/source/en/model_doc/gemma3.md Co-authored-by: Pedro Cuenca <pedro@huggingface.co> * Update docs/source/en/model_doc/gemma3.md Co-authored-by: Pedro Cuenca <pedro@huggingface.co> * Update docs/source/en/model_doc/gemma3.md Co-authored-by: Pedro Cuenca <pedro@huggingface.co> * Update docs/source/en/model_doc/gemma3.md Co-authored-by: Pedro Cuenca <pedro@huggingface.co> * Docs updates now that assets are live * Style fixes --------- Co-authored-by: Joshua Lochner <admin@xenova.com> Co-authored-by: Pedro Cuenca <pedro@huggingface.co> Co-authored-by: Aritra Roy Gosthipaty <aritra.born2fly@gmail.com> Co-authored-by: Mayank Chaturvedi <imayank@google.com> Co-authored-by: Matthew Douglas <38992547+matthewdouglas@users.noreply.github.com> Co-authored-by: raushan <raushan@huggingface.co> Co-authored-by: Raushan Turganbay <raushan.turganbay@alumni.nu.edu.kz> Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com> Co-authored-by: Lysandre <hi@lysand.re>
This commit is contained in:
Ryan Mullins
229
tests/models/gemma3/test_image_processing_gemma3.py
Normal file
229
tests/models/gemma3/test_image_processing_gemma3.py
Normal file
@@ -0,0 +1,229 @@
|
||||
# coding=utf-8
|
||||
# Copyright 2025 HuggingFace Inc.
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
import unittest
|
||||
|
||||
import numpy as np
|
||||
|
||||
from transformers.image_utils import IMAGENET_STANDARD_MEAN, IMAGENET_STANDARD_STD
|
||||
from transformers.testing_utils import require_torch, require_vision
|
||||
from transformers.utils import is_torch_available, is_torchvision_available, is_vision_available
|
||||
|
||||
from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
|
||||
|
||||
|
||||
if is_torch_available():
|
||||
import torch
|
||||
|
||||
if is_vision_available():
|
||||
from PIL import Image
|
||||
|
||||
from transformers import Gemma3ImageProcessor
|
||||
|
||||
if is_torchvision_available():
|
||||
from transformers import Gemma3ImageProcessorFast
|
||||
|
||||
|
||||
class Gemma3ImageProcessingTester:
|
||||
def __init__(
|
||||
self,
|
||||
parent,
|
||||
batch_size=7,
|
||||
num_channels=3,
|
||||
image_size=18,
|
||||
min_resolution=30,
|
||||
max_resolution=400,
|
||||
do_resize=True,
|
||||
size=None,
|
||||
do_normalize=True,
|
||||
image_mean=IMAGENET_STANDARD_MEAN,
|
||||
image_std=IMAGENET_STANDARD_STD,
|
||||
do_convert_rgb=True,
|
||||
do_pan_and_scan=True,
|
||||
pan_and_scan_min_crop_size=10,
|
||||
pan_and_scan_max_num_crops=2,
|
||||
pan_and_scan_min_ratio_to_activate=1.2,
|
||||
):
|
||||
super().__init__()
|
||||
size = size if size is not None else {"height": 18, "width": 18}
|
||||
self.parent = parent
|
||||
self.batch_size = batch_size
|
||||
self.num_channels = num_channels
|
||||
self.image_size = image_size
|
||||
self.min_resolution = min_resolution
|
||||
self.max_resolution = max_resolution
|
||||
self.do_resize = do_resize
|
||||
self.size = size
|
||||
self.do_normalize = do_normalize
|
||||
self.image_mean = image_mean
|
||||
self.image_std = image_std
|
||||
self.do_convert_rgb = do_convert_rgb
|
||||
self.do_pan_and_scan = do_pan_and_scan
|
||||
self.pan_and_scan_min_crop_size = pan_and_scan_min_crop_size
|
||||
self.pan_and_scan_max_num_crops = pan_and_scan_max_num_crops
|
||||
self.pan_and_scan_min_ratio_to_activate = pan_and_scan_min_ratio_to_activate
|
||||
|
||||
def prepare_image_processor_dict(self):
|
||||
return {
|
||||
"do_resize": self.do_resize,
|
||||
"size": self.size,
|
||||
"do_normalize": self.do_normalize,
|
||||
"image_mean": self.image_mean,
|
||||
"image_std": self.image_std,
|
||||
"do_convert_rgb": self.do_convert_rgb,
|
||||
"do_pan_and_scan": self.do_pan_and_scan,
|
||||
"pan_and_scan_min_crop_size": self.pan_and_scan_min_crop_size,
|
||||
"pan_and_scan_max_num_crops": self.pan_and_scan_max_num_crops,
|
||||
"pan_and_scan_min_ratio_to_activate": self.pan_and_scan_min_ratio_to_activate,
|
||||
}
|
||||
|
||||
def expected_output_image_shape(self, images):
|
||||
return self.num_channels, self.size["height"], self.size["width"]
|
||||
|
||||
# Copied from tests.models.clip.test_image_processing_clip.CLIPImageProcessingTester.prepare_image_inputs
|
||||
def prepare_image_inputs(self, equal_resolution=False, numpify=False, torchify=False):
|
||||
return prepare_image_inputs(
|
||||
batch_size=self.batch_size,
|
||||
num_channels=self.num_channels,
|
||||
min_resolution=self.min_resolution,
|
||||
max_resolution=self.max_resolution,
|
||||
equal_resolution=equal_resolution,
|
||||
numpify=numpify,
|
||||
torchify=torchify,
|
||||
)
|
||||
|
||||
|
||||
@require_torch
|
||||
@require_vision
|
||||
class Gemma3ImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
|
||||
image_processing_class = Gemma3ImageProcessor if is_vision_available() else None
|
||||
fast_image_processing_class = Gemma3ImageProcessorFast if is_torchvision_available() else None
|
||||
|
||||
# Copied from tests.models.clip.test_image_processing_clip.CLIPImageProcessingTest.setUp with CLIP->Gemma3
|
||||
def setUp(self):
|
||||
super().setUp()
|
||||
self.image_processor_tester = Gemma3ImageProcessingTester(self)
|
||||
|
||||
@property
|
||||
# Copied from tests.models.clip.test_image_processing_clip.CLIPImageProcessingTest.image_processor_dict
|
||||
def image_processor_dict(self):
|
||||
return self.image_processor_tester.prepare_image_processor_dict()
|
||||
|
||||
def test_image_processor_properties(self):
|
||||
for image_processing_class in self.image_processor_list:
|
||||
image_processing = image_processing_class(**self.image_processor_dict)
|
||||
self.assertTrue(hasattr(image_processing, "do_resize"))
|
||||
self.assertTrue(hasattr(image_processing, "size"))
|
||||
self.assertTrue(hasattr(image_processing, "do_normalize"))
|
||||
self.assertTrue(hasattr(image_processing, "image_mean"))
|
||||
self.assertTrue(hasattr(image_processing, "image_std"))
|
||||
self.assertTrue(hasattr(image_processing, "do_convert_rgb"))
|
||||
self.assertTrue(hasattr(image_processing, "do_pan_and_scan"))
|
||||
self.assertTrue(hasattr(image_processing, "pan_and_scan_min_crop_size"))
|
||||
self.assertTrue(hasattr(image_processing, "pan_and_scan_max_num_crops"))
|
||||
self.assertTrue(hasattr(image_processing, "pan_and_scan_min_ratio_to_activate"))
|
||||
|
||||
def test_image_processor_from_dict_with_kwargs(self):
|
||||
for image_processing_class in self.image_processor_list:
|
||||
image_processor = image_processing_class.from_dict(self.image_processor_dict)
|
||||
self.assertEqual(image_processor.size, {"height": 18, "width": 18})
|
||||
|
||||
image_processor = image_processing_class.from_dict(self.image_processor_dict, size=84)
|
||||
self.assertEqual(image_processor.size, {"height": 84, "width": 84})
|
||||
|
||||
def test_pan_and_scan(self):
|
||||
"""
|
||||
Enables Pan and Scan path by choosing the correct input image resolution. If you are changing
|
||||
image processor attributes for PaS, please update this test.
|
||||
"""
|
||||
for image_processing_class in self.image_processor_list:
|
||||
# Initialize image_processing
|
||||
image_processing = image_processing_class(**self.image_processor_dict)
|
||||
# create random numpy tensors
|
||||
"""This function prepares a list of PIL images"""
|
||||
image_inputs = [np.random.randint(255, size=(3, 300, 600), dtype=np.uint8)] * 3
|
||||
image_inputs = [Image.fromarray(np.moveaxis(x, 0, -1)) for x in image_inputs]
|
||||
|
||||
# Test not batched input, 3 images because we have base image + 2 crops
|
||||
encoded_images = image_processing(image_inputs[0], return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (3, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
# Test batched, 9 images because we have base image + 2 crops per each item
|
||||
encoded_images = image_processing(image_inputs, return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (9, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
def test_call_pil(self):
|
||||
for image_processing_class in self.image_processor_list:
|
||||
# Initialize image_processing
|
||||
image_processing = image_processing_class(**self.image_processor_dict)
|
||||
# create random PIL images
|
||||
image_inputs = self.image_processor_tester.prepare_image_inputs(equal_resolution=True)
|
||||
for image in image_inputs:
|
||||
self.assertIsInstance(image, Image.Image)
|
||||
|
||||
# Test not batched input
|
||||
encoded_images = image_processing(image_inputs[0], return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (1, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
# Test batched
|
||||
encoded_images = image_processing(image_inputs, return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (7, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
def test_call_numpy(self):
|
||||
for image_processing_class in self.image_processor_list:
|
||||
# Initialize image_processing
|
||||
image_processing = image_processing_class(**self.image_processor_dict)
|
||||
# create random numpy tensors
|
||||
image_inputs = self.image_processor_tester.prepare_image_inputs(equal_resolution=True, numpify=True)
|
||||
for image in image_inputs:
|
||||
self.assertIsInstance(image, np.ndarray)
|
||||
|
||||
# Test not batched input
|
||||
encoded_images = image_processing(image_inputs[0], return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (1, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
# Test batched
|
||||
encoded_images = image_processing(image_inputs, return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (7, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
def test_call_pytorch(self):
|
||||
for image_processing_class in self.image_processor_list:
|
||||
# Initialize image_processing
|
||||
image_processing = image_processing_class(**self.image_processor_dict)
|
||||
# create random PyTorch tensors
|
||||
image_inputs = self.image_processor_tester.prepare_image_inputs(equal_resolution=True, torchify=True)
|
||||
|
||||
for image in image_inputs:
|
||||
self.assertIsInstance(image, torch.Tensor)
|
||||
|
||||
# Test not batched input
|
||||
encoded_images = image_processing(image_inputs[0], return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (1, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
# Test batched
|
||||
encoded_images = image_processing(image_inputs, return_tensors="pt").pixel_values
|
||||
expected_output_image_shape = (7, 3, 18, 18)
|
||||
self.assertEqual(tuple(encoded_images.shape), expected_output_image_shape)
|
||||
|
||||
@unittest.skip("Gemma3 doesn't work with 4 channels due to pan and scan method")
|
||||
def test_call_numpy_4_channels(self):
|
||||
pass
|
||||
Reference in New Issue
Block a user