Spaces:
Running
on
Zero
Running
on
Zero
File size: 4,686 Bytes
22a452a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 |
import gc
import unittest
import numpy as np
import torch
from transformers import AutoTokenizer, GemmaConfig, GemmaForCausalLM
from diffusers import (
AutoencoderKL,
FlowMatchEulerDiscreteScheduler,
LuminaNextDiT2DModel,
LuminaPipeline,
)
from diffusers.utils.testing_utils import (
backend_empty_cache,
numpy_cosine_similarity_distance,
require_torch_accelerator,
slow,
torch_device,
)
from ..test_pipelines_common import PipelineTesterMixin
class LuminaPipelineFastTests(unittest.TestCase, PipelineTesterMixin):
pipeline_class = LuminaPipeline
params = frozenset(
[
"prompt",
"height",
"width",
"guidance_scale",
"negative_prompt",
"prompt_embeds",
"negative_prompt_embeds",
]
)
batch_params = frozenset(["prompt", "negative_prompt"])
supports_dduf = False
test_layerwise_casting = True
test_group_offloading = True
def get_dummy_components(self):
torch.manual_seed(0)
transformer = LuminaNextDiT2DModel(
sample_size=4,
patch_size=2,
in_channels=4,
hidden_size=4,
num_layers=2,
num_attention_heads=1,
num_kv_heads=1,
multiple_of=16,
ffn_dim_multiplier=None,
norm_eps=1e-5,
learn_sigma=True,
qk_norm=True,
cross_attention_dim=8,
scaling_factor=1.0,
)
torch.manual_seed(0)
vae = AutoencoderKL()
scheduler = FlowMatchEulerDiscreteScheduler()
tokenizer = AutoTokenizer.from_pretrained("hf-internal-testing/dummy-gemma")
torch.manual_seed(0)
config = GemmaConfig(
head_dim=2,
hidden_size=8,
intermediate_size=37,
num_attention_heads=4,
num_hidden_layers=2,
num_key_value_heads=4,
)
text_encoder = GemmaForCausalLM(config)
components = {
"transformer": transformer.eval(),
"vae": vae.eval(),
"scheduler": scheduler,
"text_encoder": text_encoder.eval(),
"tokenizer": tokenizer,
}
return components
def get_dummy_inputs(self, device, seed=0):
if str(device).startswith("mps"):
generator = torch.manual_seed(seed)
else:
generator = torch.Generator(device="cpu").manual_seed(seed)
inputs = {
"prompt": "A painting of a squirrel eating a burger",
"generator": generator,
"num_inference_steps": 2,
"guidance_scale": 5.0,
"output_type": "np",
}
return inputs
@unittest.skip("xformers attention processor does not exist for Lumina")
def test_xformers_attention_forwardGenerator_pass(self):
pass
@slow
@require_torch_accelerator
class LuminaPipelineSlowTests(unittest.TestCase):
pipeline_class = LuminaPipeline
repo_id = "Alpha-VLLM/Lumina-Next-SFT-diffusers"
def setUp(self):
super().setUp()
gc.collect()
backend_empty_cache(torch_device)
def tearDown(self):
super().tearDown()
gc.collect()
backend_empty_cache(torch_device)
def get_inputs(self, device, seed=0):
if str(device).startswith("mps"):
generator = torch.manual_seed(seed)
else:
generator = torch.Generator(device="cpu").manual_seed(seed)
return {
"prompt": "A photo of a cat",
"num_inference_steps": 2,
"guidance_scale": 5.0,
"output_type": "np",
"generator": generator,
}
def test_lumina_inference(self):
pipe = self.pipeline_class.from_pretrained(self.repo_id, torch_dtype=torch.bfloat16)
pipe.enable_model_cpu_offload(device=torch_device)
inputs = self.get_inputs(torch_device)
image = pipe(**inputs).images[0]
image_slice = image[0, :10, :10]
expected_slice = np.array(
[
[0.17773438, 0.18554688, 0.22070312],
[0.046875, 0.06640625, 0.10351562],
[0.0, 0.0, 0.02148438],
[0.0, 0.0, 0.0],
[0.0, 0.0, 0.0],
[0.0, 0.0, 0.0],
[0.0, 0.0, 0.0],
[0.0, 0.0, 0.0],
[0.0, 0.0, 0.0],
[0.0, 0.0, 0.0],
],
dtype=np.float32,
)
max_diff = numpy_cosine_similarity_distance(expected_slice.flatten(), image_slice.flatten())
assert max_diff < 1e-4
|