SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

VLMs: enable generation tests (#33533)

Browse Source 
* add tests

* fix whisper

* update

* nit

* add qwen2-vl

* more updates!

* better this way

* fix this one

* fix more tests

* fix final tests, hope so

* fix led

* Update tests/generation/test_utils.py

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* pr comments

* not pass pixels and extra for low-mem tests, very flaky because of visio tower

---------

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
This commit is contained in:
Raushan Turganbay
2024-09-19 12:04:24 +02:00
committed by GitHub
parent e40bb4845e
commit d7975a5874
22 changed files with 500 additions and 207 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
tests/models/paligemma/test_modeling_paligemma.py
View File

@@ -35,6 +35,7 @@ from transformers.testing_utils import (
torch_device,
)
from ...generation.test_utils import GenerationTesterMixin
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
@@ -82,7 +83,7 @@ class PaliGemmaVisionText2TextModelTester:
"initializer_range": 0.02,
"num_labels": 3,
"num_choices": 4,
"pad_token_id": 0,
"pad_token_id": 1,
},
is_training=True,
vision_config={
@@ -115,6 +116,7 @@ class PaliGemmaVisionText2TextModelTester:
self.vision_config = vision_config
self.seq_length = seq_length
self.projection_dim = projection_dim
self.pad_token_id = text_config["pad_token_id"]
self.num_hidden_layers = text_config["num_hidden_layers"]
self.vocab_size = text_config["vocab_size"]
@@ -160,7 +162,7 @@ class PaliGemmaVisionText2TextModelTester:
attention_mask = input_ids.ne(1).to(torch_device)
# set the 16 first tokens to be image, and ensure that no other tokens are image tokens
# do not change this unless you modified image size or patch size
input_ids = torch.where(input_ids == config.image_token_index, 2, input_ids)
input_ids[input_ids == config.image_token_index] = self.pad_token_id
input_ids[:, :16] = config.image_token_index
inputs_dict = {
"pixel_values": pixel_values,
@@ -173,12 +175,13 @@ class PaliGemmaVisionText2TextModelTester:
@require_torch
class PaliGemmaForConditionalGenerationModelTest(ModelTesterMixin, unittest.TestCase):
class PaliGemmaForConditionalGenerationModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
"""
Model tester for `PaliGemmaForConditionalGeneration`.
"""
all_model_classes = (PaliGemmaForConditionalGeneration,) if is_torch_available() else ()
all_generative_model_classes = (PaliGemmaForConditionalGeneration,) if is_torch_available() else ()
fx_compatible = False
test_pruning = False
test_torchscript = False
@@ -305,6 +308,12 @@ class PaliGemmaForConditionalGenerationModelTest(ModelTesterMixin, unittest.Test
def test_save_load_low_cpu_mem_usage_no_safetensors(self):
pass
@unittest.skip(
reason="VLMs doen't accept inputs embeds and pixel values at the same time. So if the test passed for bacbone LM, it passes for VLM also"
)
def test_generate_from_inputs_embeds_with_static_cache(self):
pass
@slow
@require_torch