File size: 4,095 Bytes
ac2243f | 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 | # 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 sys
import unittest
import torch
from transformers import AutoTokenizer, T5EncoderModel
from diffusers import AutoencoderKLMochi, FlowMatchEulerDiscreteScheduler, MochiPipeline, MochiTransformer3DModel
from ..testing_utils import (
floats_tensor,
require_peft_backend,
skip_mps,
)
sys.path.append(".")
from .utils import PeftLoraLoaderMixinTests # noqa: E402
@require_peft_backend
@skip_mps
class MochiLoRATests(unittest.TestCase, PeftLoraLoaderMixinTests):
pipeline_class = MochiPipeline
scheduler_cls = FlowMatchEulerDiscreteScheduler
scheduler_kwargs = {}
transformer_kwargs = {
"patch_size": 2,
"num_attention_heads": 2,
"attention_head_dim": 8,
"num_layers": 2,
"pooled_projection_dim": 16,
"in_channels": 12,
"out_channels": None,
"qk_norm": "rms_norm",
"text_embed_dim": 32,
"time_embed_dim": 4,
"activation_fn": "swiglu",
"max_sequence_length": 16,
}
transformer_cls = MochiTransformer3DModel
vae_kwargs = {
"latent_channels": 12,
"out_channels": 3,
"encoder_block_out_channels": (32, 32, 32, 32),
"decoder_block_out_channels": (32, 32, 32, 32),
"layers_per_block": (1, 1, 1, 1, 1),
}
vae_cls = AutoencoderKLMochi
tokenizer_cls, tokenizer_id = AutoTokenizer, "hf-internal-testing/tiny-random-t5"
text_encoder_cls, text_encoder_id = T5EncoderModel, "hf-internal-testing/tiny-random-t5"
text_encoder_target_modules = ["q", "k", "v", "o"]
supports_text_encoder_loras = False
@property
def output_shape(self):
return (1, 7, 16, 16, 3)
def get_dummy_inputs(self, with_generator=True):
batch_size = 1
sequence_length = 16
num_channels = 4
num_frames = 7
num_latent_frames = 3
sizes = (2, 2)
generator = torch.manual_seed(0)
noise = floats_tensor((batch_size, num_latent_frames, num_channels) + sizes)
input_ids = torch.randint(1, sequence_length, size=(batch_size, sequence_length), generator=generator)
pipeline_inputs = {
"prompt": "dance monkey",
"num_frames": num_frames,
"num_inference_steps": 4,
"guidance_scale": 6.0,
# Cannot reduce because convolution kernel becomes bigger than sample
"height": 16,
"width": 16,
"max_sequence_length": sequence_length,
"output_type": "np",
}
if with_generator:
pipeline_inputs.update({"generator": generator})
return noise, input_ids, pipeline_inputs
def test_simple_inference_with_text_lora_denoiser_fused_multi(self):
super().test_simple_inference_with_text_lora_denoiser_fused_multi(expected_atol=9e-3)
def test_simple_inference_with_text_denoiser_lora_unfused(self):
super().test_simple_inference_with_text_denoiser_lora_unfused(expected_atol=9e-3)
@unittest.skip("Not supported in Mochi.")
def test_simple_inference_with_text_denoiser_block_scale(self):
pass
@unittest.skip("Not supported in Mochi.")
def test_simple_inference_with_text_denoiser_block_scale_for_all_dict_options(self):
pass
@unittest.skip("Not supported in Mochi.")
def test_modify_padding_mode(self):
pass
@unittest.skip("Not supported in CogVideoX.")
def test_simple_inference_with_text_denoiser_multi_adapter_block_lora(self):
pass
|