Spaces:
Paused
Paused
File size: 9,407 Bytes
2f86d9b | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 | # Copyright 2026 DeepMind Technologies Limited.
#
# 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.
from gemma.diffusion.hackable_diffusion_adapter.hd import mask_helpers
import jax.numpy as jnp
import numpy as np
from absl.testing import absltest
class BuildPositionsFromMaskTest(absltest.TestCase):
"""Tests for build_positions_from_mask."""
def test_no_padding(self):
"""All-real tokens produce a simple arange."""
mask = jnp.array([[True, True, True, True]])
positions = mask_helpers.build_positions_from_mask(mask)
np.testing.assert_array_equal(positions, [[0, 1, 2, 3]])
def test_right_padding(self):
"""Pad tokens repeat the last real position."""
mask = jnp.array([[True, True, True, False, False]])
positions = mask_helpers.build_positions_from_mask(mask)
np.testing.assert_array_equal(positions, [[0, 1, 2, 2, 2]])
def test_all_padding(self):
"""All-pad input produces all zeros."""
mask = jnp.array([[False, False, False]])
positions = mask_helpers.build_positions_from_mask(mask)
np.testing.assert_array_equal(positions, [[0, 0, 0]])
def test_single_token(self):
mask = jnp.array([[True]])
positions = mask_helpers.build_positions_from_mask(mask)
np.testing.assert_array_equal(positions, [[0]])
def test_batched(self):
"""Different padding lengths within a batch."""
mask = jnp.array([
[True, True, True, True, False],
[True, True, False, False, False],
])
positions = mask_helpers.build_positions_from_mask(mask)
np.testing.assert_array_equal(
positions,
[
[0, 1, 2, 3, 3],
[0, 1, 1, 1, 1],
],
)
class MakeCausalPrefillMaskTest(absltest.TestCase):
"""Tests for make_causal_prefill_mask."""
def test_no_padding_no_cache_pad(self):
"""cache_length == seq_len, no right-padding on either axis."""
mask = jnp.array([[True, True, True]])
result = mask_helpers.make_causal_prefill_mask(mask, cache_length=3)
expected = jnp.array(
[[[True, False, False], [True, True, False], [True, True, True]]]
)
np.testing.assert_array_equal(result, expected)
def test_right_padded_tokens(self):
"""Pad tokens in the mask should zero out their columns."""
mask = jnp.array([[True, True, False]])
result = mask_helpers.make_causal_prefill_mask(mask, cache_length=3)
expected = jnp.array(
[[[True, False, False], [True, True, False], [True, True, False]]]
)
np.testing.assert_array_equal(result, expected)
def test_cache_padding(self):
"""cache_length > seq_len adds False columns on the right."""
mask = jnp.array([[True, True]])
result = mask_helpers.make_causal_prefill_mask(mask, cache_length=5)
self.assertEqual(result.shape, (1, 2, 5))
expected = jnp.array([[
[True, False, False, False, False],
[True, True, False, False, False],
]])
np.testing.assert_array_equal(result, expected)
def test_combined_token_and_cache_padding(self):
"""Both token-level pad and cache_length > seq_len."""
mask = jnp.array([[True, False]])
result = mask_helpers.make_causal_prefill_mask(mask, cache_length=4)
self.assertEqual(result.shape, (1, 2, 4))
expected = jnp.array(
[[[True, False, False, False], [True, False, False, False]]]
)
np.testing.assert_array_equal(result, expected)
def test_cache_length_equals_seq_len(self):
"""No extra padding when cache_length == seq_len."""
mask = jnp.array([[True, True, True]])
result = mask_helpers.make_causal_prefill_mask(mask, cache_length=3)
self.assertEqual(result.shape, (1, 3, 3))
def test_batched(self):
"""Different padding within a batch."""
mask = jnp.array([
[True, True, True],
[True, False, False],
])
result = mask_helpers.make_causal_prefill_mask(mask, cache_length=5)
self.assertEqual(result.shape, (2, 3, 5))
np.testing.assert_array_equal(
result[0],
[
[True, False, False, False, False],
[True, True, False, False, False],
[True, True, True, False, False],
],
)
np.testing.assert_array_equal(
result[1],
[
[True, False, False, False, False],
[True, False, False, False, False],
[True, False, False, False, False],
],
)
def test_output_dtype_is_bool(self):
mask = jnp.array([[True, True]])
result = mask_helpers.make_causal_prefill_mask(mask, cache_length=3)
self.assertEqual(result.dtype, jnp.bool_)
class SetCacheEndIndexTest(absltest.TestCase):
"""Tests for set_cache_end_index."""
def _make_fake_cache(self, batch_size, cache_length, num_layers=2):
cache = {}
for i in range(num_layers):
cache[f'layer_{i}'] = {
'k': jnp.zeros((batch_size, cache_length, 2, 4)),
'v': jnp.zeros((batch_size, cache_length, 2, 4)),
'end_index': jnp.zeros((batch_size,), dtype=jnp.int32),
'positions': jnp.zeros((batch_size, cache_length), dtype=jnp.int32),
}
return cache
def test_sets_end_index_all_layers(self):
cache = self._make_fake_cache(batch_size=2, cache_length=8)
end_index = jnp.array([3, 5], dtype=jnp.int32)
new_cache = mask_helpers.set_cache_end_index(cache, end_index)
for layer_name in ['layer_0', 'layer_1']:
np.testing.assert_array_equal(new_cache[layer_name]['end_index'], [3, 5])
def test_preserves_other_keys(self):
cache = self._make_fake_cache(batch_size=1, cache_length=4)
cache['layer_0']['k'] = jnp.ones_like(cache['layer_0']['k'])
end_index = jnp.array([2], dtype=jnp.int32)
new_cache = mask_helpers.set_cache_end_index(cache, end_index)
np.testing.assert_array_equal(
new_cache['layer_0']['k'], jnp.ones_like(cache['layer_0']['k'])
)
def test_scalar_broadcast(self):
cache = self._make_fake_cache(batch_size=3, cache_length=8)
end_index = jnp.array(7, dtype=jnp.int32)
new_cache = mask_helpers.set_cache_end_index(cache, end_index)
for layer_name in cache:
np.testing.assert_array_equal(
new_cache[layer_name]['end_index'], [7, 7, 7]
)
def test_does_not_mutate_original(self):
cache = self._make_fake_cache(batch_size=1, cache_length=4)
end_index = jnp.array([5], dtype=jnp.int32)
_ = mask_helpers.set_cache_end_index(cache, end_index)
np.testing.assert_array_equal(cache['layer_0']['end_index'], [0])
class MakeFullAttentionMaskTest(absltest.TestCase):
"""Tests for make_full_attention_mask."""
def test_basic(self):
input_mask = jnp.array([[True, True, False]])
result = mask_helpers.make_full_attention_mask(input_mask, cache_length=5)
expected = jnp.array([[True, True, False, True, True]])
np.testing.assert_array_equal(result, expected)
def test_no_padding(self):
input_mask = jnp.array([[True, True, True]])
result = mask_helpers.make_full_attention_mask(input_mask, cache_length=5)
expected = jnp.array([[True, True, True, True, True]])
np.testing.assert_array_equal(result, expected)
class CreateDecoderAttentionMaskTest(absltest.TestCase):
"""Tests for create_decoder_attention_mask."""
def test_shape(self):
batch_size = 2
prompt_len = 3
total_canvas_len = 4
canvas_size = 2
num_queries = 4
prompt_mask = jnp.ones((batch_size, prompt_len), dtype=jnp.bool_)
canvas_mask = jnp.ones((batch_size, total_canvas_len), dtype=jnp.bool_)
selected_canvas_idx = jnp.array([0, 1])
result = mask_helpers.create_decoder_attention_mask(
prompt_mask=prompt_mask,
canvas_mask=canvas_mask,
selected_canvas_idx=selected_canvas_idx,
prompt_len=prompt_len,
total_canvas_len=total_canvas_len,
canvas_size=canvas_size,
num_queries=num_queries,
)
self.assertEqual(
result.shape, (batch_size, num_queries, prompt_len + total_canvas_len)
)
class MakeCausalAttentionMaskRightPadTest(absltest.TestCase):
"""Tests for make_causal_attention_mask_right_pad."""
def test_no_cache(self):
"""Without cache, returns a simple lower-triangular mask."""
result = mask_helpers.make_causal_attention_mask_right_pad(
batch_size=1,
canvas_length=3,
cache_length=None,
num_valid_cache_tokens=None,
)
expected = jnp.array([
[[True, False, False], [True, True, False], [True, True, True]],
])
np.testing.assert_array_equal(result, expected)
def test_with_cache(self):
"""With cache, mask reveals valid entries + causal triangle at write pos."""
result = mask_helpers.make_causal_attention_mask_right_pad(
batch_size=1,
canvas_length=2,
cache_length=5,
num_valid_cache_tokens=jnp.array([3]),
)
self.assertEqual(result.shape, (1, 2, 5))
if __name__ == '__main__':
absltest.main()
|