Hugging Face's logo

Datasets:
hf-internal-testing
/
transformers_flash_attn_ci

Dataset card Files
xet
Community
hf-transformers-bot commited on
Commit
57abe81
·
verified ·
1 Parent(s): 1b7f08b

Upload 2025-11-13/ci_results_run_models_gpu/new_failures.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. 2025-11-13/ci_results_run_models_gpu/new_failures.json +710 -0
2025-11-13/ci_results_run_models_gpu/new_failures.json ADDED
@@ -0,0 +1,710 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audioflamingo3": {
3
+ "single-gpu": [
4
+ "tests/models/audioflamingo3/test_modeling_audioflamingo3.py::AudioFlamingo3ForConditionalGenerationModelTest::test_eager_matches_fa2_generate"
5
+ ],
6
+ "multi-gpu": [
7
+ "tests/models/audioflamingo3/test_modeling_audioflamingo3.py::AudioFlamingo3ForConditionalGenerationModelTest::test_eager_matches_fa2_generate"
8
+ ]
9
+ },
10
+ "bamba": {
11
+ "single-gpu": [
12
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
13
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
14
+ ],
15
+ "multi-gpu": [
16
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
17
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
18
+ ]
19
+ },
20
+ "bark": {
21
+ "single-gpu": [
22
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_eager_matches_fa2_generate",
23
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_flash_attention_2_continue_generate_with_position_ids",
24
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_eager_matches_fa2_generate",
25
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_flash_attention_2_continue_generate_with_position_ids"
26
+ ],
27
+ "multi-gpu": [
28
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_eager_matches_fa2_generate",
29
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_flash_attention_2_continue_generate_with_position_ids",
30
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_eager_matches_fa2_generate",
31
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_flash_attention_2_continue_generate_with_position_ids"
32
+ ]
33
+ },
34
+ "blt": {
35
+ "single-gpu": [
36
+ "tests/models/blt/test_modeling_blt.py::BltModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
37
+ ],
38
+ "multi-gpu": [
39
+ "tests/models/blt/test_modeling_blt.py::BltModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
40
+ ]
41
+ },
42
+ "chameleon": {
43
+ "single-gpu": [
44
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
45
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonVision2SeqModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
46
+ ],
47
+ "multi-gpu": [
48
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
49
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonVision2SeqModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
50
+ ]
51
+ },
52
+ "clap": {
53
+ "single-gpu": [
54
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence",
55
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence_right_padding",
56
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence",
57
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence_right_padding"
58
+ ],
59
+ "multi-gpu": [
60
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence",
61
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence_right_padding",
62
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence",
63
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence_right_padding"
64
+ ]
65
+ },
66
+ "colpali": {
67
+ "single-gpu": [
68
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence",
69
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence_right_padding"
70
+ ],
71
+ "multi-gpu": [
72
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence",
73
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence_right_padding"
74
+ ]
75
+ },
76
+ "deepseek_v2": {
77
+ "single-gpu": [
78
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
79
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break",
80
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_fp32_ln"
81
+ ],
82
+ "multi-gpu": [
83
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
84
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break",
85
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_fp32_ln"
86
+ ]
87
+ },
88
+ "deepseek_vl_hybrid": {
89
+ "single-gpu": [
90
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attention_2_continue_generate_with_position_ids",
91
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_fp32_ln",
92
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_from_config"
93
+ ],
94
+ "multi-gpu": [
95
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attention_2_continue_generate_with_position_ids",
96
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_fp32_ln",
97
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_from_config"
98
+ ]
99
+ },
100
+ "diffllama": {
101
+ "single-gpu": [
102
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_generate_padding_right",
103
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence",
104
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
105
+ ],
106
+ "multi-gpu": [
107
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_generate_padding_right",
108
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence",
109
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
110
+ ]
111
+ },
112
+ "donut": {
113
+ "single-gpu": [
114
+ "tests/models/donut/test_modeling_donut_swin.py::DonutSwinModelTest::test_flash_attn_2_inference_equivalence",
115
+ "tests/models/donut/test_modeling_donut_swin.py::DonutSwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
116
+ ],
117
+ "multi-gpu": [
118
+ "tests/models/donut/test_modeling_donut_swin.py::DonutSwinModelTest::test_flash_attn_2_inference_equivalence",
119
+ "tests/models/donut/test_modeling_donut_swin.py::DonutSwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
120
+ ]
121
+ },
122
+ "efficientloftr": {
123
+ "single-gpu": [
124
+ "tests/models/efficientloftr/test_modeling_efficientloftr.py::EfficientLoFTRModelTest::test_flash_attn_2_inference_equivalence",
125
+ "tests/models/efficientloftr/test_modeling_efficientloftr.py::EfficientLoFTRModelTest::test_flash_attn_2_inference_equivalence_right_padding"
126
+ ],
127
+ "multi-gpu": [
128
+ "tests/models/efficientloftr/test_modeling_efficientloftr.py::EfficientLoFTRModelTest::test_flash_attn_2_inference_equivalence",
129
+ "tests/models/efficientloftr/test_modeling_efficientloftr.py::EfficientLoFTRModelTest::test_flash_attn_2_inference_equivalence_right_padding"
130
+ ]
131
+ },
132
+ "emu3": {
133
+ "single-gpu": [
134
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence",
135
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
136
+ ],
137
+ "multi-gpu": [
138
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence",
139
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
140
+ ]
141
+ },
142
+ "exaone4": {
143
+ "single-gpu": [
144
+ "tests/models/exaone4/test_modeling_exaone4.py::Exaone4ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
145
+ ],
146
+ "multi-gpu": [
147
+ "tests/models/exaone4/test_modeling_exaone4.py::Exaone4ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
148
+ "tests/models/exaone4/test_modeling_exaone4.py::Exaone4IntegrationTest::test_model_generation_long_flash"
149
+ ]
150
+ },
151
+ "falcon": {
152
+ "single-gpu": [
153
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence",
154
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence_right_padding"
155
+ ],
156
+ "multi-gpu": [
157
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence",
158
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence_right_padding"
159
+ ]
160
+ },
161
+ "flex_olmo": {
162
+ "single-gpu": [
163
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
164
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs",
165
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
166
+ ],
167
+ "multi-gpu": [
168
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
169
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
170
+ ]
171
+ },
172
+ "gemma3n": {
173
+ "single-gpu": [
174
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_equivalence",
175
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence",
176
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
177
+ ],
178
+ "multi-gpu": [
179
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_equivalence",
180
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence",
181
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
182
+ ]
183
+ },
184
+ "glm4": {
185
+ "single-gpu": [],
186
+ "multi-gpu": [
187
+ "tests/models/glm4/test_modeling_glm4.py::Glm4ModelTest::test_flash_attn_2_equivalence"
188
+ ]
189
+ },
190
+ "glm4_moe": {
191
+ "single-gpu": [
192
+ "tests/models/glm4_moe/test_modeling_glm4_moe.py::Glm4MoeModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
193
+ ],
194
+ "multi-gpu": [
195
+ "tests/models/glm4_moe/test_modeling_glm4_moe.py::Glm4MoeModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
196
+ ]
197
+ },
198
+ "gpt_neo": {
199
+ "single-gpu": [
200
+ "tests/models/gpt_neo/test_modeling_gpt_neo.py::GPTNeoModelTest::test_eager_matches_fa2_generate"
201
+ ],
202
+ "multi-gpu": []
203
+ },
204
+ "gpt_oss": {
205
+ "single-gpu": [
206
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence",
207
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence_right_padding"
208
+ ],
209
+ "multi-gpu": [
210
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence",
211
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence_right_padding",
212
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
213
+ ]
214
+ },
215
+ "granitemoe": {
216
+ "single-gpu": [
217
+ "tests/models/granitemoe/test_modeling_granitemoe.py::GraniteMoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
218
+ ],
219
+ "multi-gpu": [
220
+ "tests/models/granitemoe/test_modeling_granitemoe.py::GraniteMoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
221
+ ]
222
+ },
223
+ "granitemoehybrid": {
224
+ "single-gpu": [
225
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
226
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding",
227
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence",
228
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence_right_padding"
229
+ ],
230
+ "multi-gpu": [
231
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
232
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding",
233
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence",
234
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence_right_padding"
235
+ ]
236
+ },
237
+ "granitemoeshared": {
238
+ "single-gpu": [
239
+ "tests/models/granitemoeshared/test_modeling_granitemoeshared.py::GraniteMoeSharedModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
240
+ ],
241
+ "multi-gpu": [
242
+ "tests/models/granitemoeshared/test_modeling_granitemoeshared.py::GraniteMoeSharedModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
243
+ ]
244
+ },
245
+ "grounding_dino": {
246
+ "single-gpu": [
247
+ "tests/models/grounding_dino/test_modeling_grounding_dino.py::GroundingDinoModelTest::test_flash_attn_2_inference_equivalence",
248
+ "tests/models/grounding_dino/test_modeling_grounding_dino.py::GroundingDinoModelTest::test_flash_attn_2_inference_equivalence_right_padding"
249
+ ],
250
+ "multi-gpu": [
251
+ "tests/models/grounding_dino/test_modeling_grounding_dino.py::GroundingDinoModelTest::test_flash_attn_2_inference_equivalence",
252
+ "tests/models/grounding_dino/test_modeling_grounding_dino.py::GroundingDinoModelTest::test_flash_attn_2_inference_equivalence_right_padding"
253
+ ]
254
+ },
255
+ "instructblip": {
256
+ "single-gpu": [
257
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
258
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
259
+ ],
260
+ "multi-gpu": [
261
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
262
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
263
+ ]
264
+ },
265
+ "instructblipvideo": {
266
+ "single-gpu": [
267
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
268
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
269
+ ],
270
+ "multi-gpu": [
271
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
272
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
273
+ ]
274
+ },
275
+ "janus": {
276
+ "single-gpu": [
277
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence",
278
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
279
+ ],
280
+ "multi-gpu": [
281
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence",
282
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
283
+ ]
284
+ },
285
+ "jetmoe": {
286
+ "single-gpu": [
287
+ "tests/models/jetmoe/test_modeling_jetmoe.py::JetMoeModelTest::test_flash_attn_2_fp32_ln"
288
+ ],
289
+ "multi-gpu": [
290
+ "tests/models/jetmoe/test_modeling_jetmoe.py::JetMoeModelTest::test_flash_attn_2_fp32_ln"
291
+ ]
292
+ },
293
+ "kosmos2": {
294
+ "single-gpu": [
295
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_eager_matches_fa2_generate",
296
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attention_2_continue_generate_with_position_ids",
297
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_can_dispatch_composite_models",
298
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_fp32_ln",
299
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_from_config",
300
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence",
301
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
302
+ ],
303
+ "multi-gpu": [
304
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_eager_matches_fa2_generate",
305
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attention_2_continue_generate_with_position_ids",
306
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_can_dispatch_composite_models",
307
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_fp32_ln",
308
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_from_config",
309
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence",
310
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
311
+ ]
312
+ },
313
+ "kosmos2_5": {
314
+ "single-gpu": [
315
+ "tests/models/kosmos2_5/test_modeling_kosmos2_5.py::Kosmos2_5ModelTest::test_flash_attn_2_can_dispatch_composite_models"
316
+ ],
317
+ "multi-gpu": [
318
+ "tests/models/kosmos2_5/test_modeling_kosmos2_5.py::Kosmos2_5ModelTest::test_flash_attn_2_can_dispatch_composite_models"
319
+ ]
320
+ },
321
+ "kyutai_speech_to_text": {
322
+ "single-gpu": [
323
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_eager_matches_fa2_generate",
324
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence",
325
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
326
+ ],
327
+ "multi-gpu": [
328
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_eager_matches_fa2_generate",
329
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence",
330
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
331
+ ]
332
+ },
333
+ "lfm2": {
334
+ "single-gpu": [
335
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
336
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
337
+ ],
338
+ "multi-gpu": [
339
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
340
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
341
+ ]
342
+ },
343
+ "lfm2_moe": {
344
+ "single-gpu": [
345
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
346
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
347
+ ],
348
+ "multi-gpu": [
349
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
350
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
351
+ ]
352
+ },
353
+ "lfm2_vl": {
354
+ "single-gpu": [
355
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence",
356
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence_right_padding"
357
+ ],
358
+ "multi-gpu": [
359
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence",
360
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence_right_padding"
361
+ ]
362
+ },
363
+ "llava_next": {
364
+ "single-gpu": [
365
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
366
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
367
+ ],
368
+ "multi-gpu": [
369
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
370
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
371
+ ]
372
+ },
373
+ "llava_next_video": {
374
+ "single-gpu": [
375
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
376
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
377
+ ],
378
+ "multi-gpu": [
379
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
380
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
381
+ ]
382
+ },
383
+ "llava_onevision": {
384
+ "single-gpu": [
385
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
386
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
387
+ ],
388
+ "multi-gpu": [
389
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
390
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
391
+ ]
392
+ },
393
+ "mask2former": {
394
+ "single-gpu": [
395
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence",
396
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
397
+ ],
398
+ "multi-gpu": [
399
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence",
400
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
401
+ ]
402
+ },
403
+ "maskformer": {
404
+ "single-gpu": [
405
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence",
406
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding",
407
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence",
408
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
409
+ ],
410
+ "multi-gpu": [
411
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence",
412
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding",
413
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence",
414
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
415
+ ]
416
+ },
417
+ "mixtral": {
418
+ "single-gpu": [],
419
+ "multi-gpu": [
420
+ "tests/models/mixtral/test_modeling_mixtral.py::MixtralModelTest::test_flash_attn_2_equivalence"
421
+ ]
422
+ },
423
+ "mllama": {
424
+ "single-gpu": [
425
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_eager_matches_fa2_generate",
426
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
427
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
428
+ ],
429
+ "multi-gpu": [
430
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_eager_matches_fa2_generate",
431
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
432
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
433
+ ]
434
+ },
435
+ "mm_grounding_dino": {
436
+ "single-gpu": [
437
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence",
438
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence_right_padding"
439
+ ],
440
+ "multi-gpu": [
441
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence",
442
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence_right_padding"
443
+ ]
444
+ },
445
+ "moshi": {
446
+ "single-gpu": [
447
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence",
448
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence_right_padding",
449
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_eager_matches_fa2_generate",
450
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_fp32_ln",
451
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_from_config",
452
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence",
453
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence_right_padding"
454
+ ],
455
+ "multi-gpu": [
456
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence",
457
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence_right_padding",
458
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_eager_matches_fa2_generate",
459
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_fp32_ln",
460
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_from_config",
461
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence",
462
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence_right_padding"
463
+ ]
464
+ },
465
+ "nemotron": {
466
+ "single-gpu": [
467
+ "tests/models/nemotron/test_modeling_nemotron.py::NemotronModelTest::test_flash_attn_2_equivalence"
468
+ ],
469
+ "multi-gpu": [
470
+ "tests/models/nemotron/test_modeling_nemotron.py::NemotronModelTest::test_flash_attn_2_equivalence"
471
+ ]
472
+ },
473
+ "olmo": {
474
+ "single-gpu": [
475
+ "tests/models/olmo/test_modeling_olmo.py::OlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
476
+ ],
477
+ "multi-gpu": [
478
+ "tests/models/olmo/test_modeling_olmo.py::OlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
479
+ ]
480
+ },
481
+ "olmo2": {
482
+ "single-gpu": [
483
+ "tests/models/olmo2/test_modeling_olmo2.py::Olmo2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
484
+ ],
485
+ "multi-gpu": [
486
+ "tests/models/olmo2/test_modeling_olmo2.py::Olmo2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
487
+ ]
488
+ },
489
+ "omdet_turbo": {
490
+ "single-gpu": [
491
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence",
492
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence_right_padding"
493
+ ],
494
+ "multi-gpu": [
495
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence",
496
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence_right_padding"
497
+ ]
498
+ },
499
+ "oneformer": {
500
+ "single-gpu": [
501
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence",
502
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
503
+ ],
504
+ "multi-gpu": [
505
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence",
506
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
507
+ ]
508
+ },
509
+ "paligemma": {
510
+ "single-gpu": [
511
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_from_config",
512
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
513
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
514
+ ],
515
+ "multi-gpu": [
516
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_from_config",
517
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
518
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
519
+ ]
520
+ },
521
+ "paligemma2": {
522
+ "single-gpu": [
523
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_from_config",
524
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
525
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
526
+ ],
527
+ "multi-gpu": [
528
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_from_config",
529
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
530
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
531
+ ]
532
+ },
533
+ "pegasus_x": {
534
+ "single-gpu": [
535
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence",
536
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence_right_padding",
537
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence",
538
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence_right_padding"
539
+ ],
540
+ "multi-gpu": [
541
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence",
542
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence_right_padding",
543
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence",
544
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence_right_padding"
545
+ ]
546
+ },
547
+ "perception_lm": {
548
+ "single-gpu": [
549
+ "tests/models/perception_lm/test_modeling_perception_lm.py::PerceptionLMForConditionalGenerationModelTest::test_flash_attention_2_continue_generate_with_position_ids"
550
+ ],
551
+ "multi-gpu": [
552
+ "tests/models/perception_lm/test_modeling_perception_lm.py::PerceptionLMForConditionalGenerationModelTest::test_flash_attention_2_continue_generate_with_position_ids"
553
+ ]
554
+ },
555
+ "phi": {
556
+ "single-gpu": [
557
+ "tests/models/phi/test_modeling_phi.py::PhiModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
558
+ ],
559
+ "multi-gpu": [
560
+ "tests/models/phi/test_modeling_phi.py::PhiModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
561
+ ]
562
+ },
563
+ "phimoe": {
564
+ "single-gpu": [
565
+ "tests/models/phimoe/test_modeling_phimoe.py::PhimoeModelTest::test_flash_attn_2_equivalence"
566
+ ],
567
+ "multi-gpu": [
568
+ "tests/models/phimoe/test_modeling_phimoe.py::PhimoeModelTest::test_flash_attn_2_equivalence"
569
+ ]
570
+ },
571
+ "pixtral": {
572
+ "single-gpu": [
573
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence",
574
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence_right_padding"
575
+ ],
576
+ "multi-gpu": [
577
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence",
578
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence_right_padding"
579
+ ]
580
+ },
581
+ "qwen2_5_vl": {
582
+ "single-gpu": [
583
+ "tests/models/qwen2_5_vl/test_modeling_qwen2_5_vl.py::Qwen2_5_VLIntegrationTest::test_small_model_integration_test_batch_wo_image_flashatt2"
584
+ ],
585
+ "multi-gpu": [
586
+ "tests/models/qwen2_5_vl/test_modeling_qwen2_5_vl.py::Qwen2_5_VLIntegrationTest::test_small_model_integration_test_batch_wo_image_flashatt2"
587
+ ]
588
+ },
589
+ "qwen3_omni_moe": {
590
+ "single-gpu": [
591
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniThinkerForConditionalGenerationModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
592
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniThinkerForConditionalGenerationModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs",
593
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniModelIntegrationTest::test_small_model_integration_test_batch_flashatt2"
594
+ ],
595
+ "multi-gpu": [
596
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniThinkerForConditionalGenerationModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
597
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniThinkerForConditionalGenerationModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs",
598
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniModelIntegrationTest::test_small_model_integration_test_batch_flashatt2"
599
+ ]
600
+ },
601
+ "qwen3_vl_moe": {
602
+ "single-gpu": [
603
+ "tests/models/qwen3_vl_moe/test_modeling_qwen3_vl_moe.py::Qwen3VLMoeIntegrationTest::test_small_model_integration_test_batch_flashatt2",
604
+ "tests/models/qwen3_vl_moe/test_modeling_qwen3_vl_moe.py::Qwen3VLMoeIntegrationTest::test_small_model_integration_test_batch_wo_image_flashatt2"
605
+ ],
606
+ "multi-gpu": [
607
+ "tests/models/qwen3_vl_moe/test_modeling_qwen3_vl_moe.py::Qwen3VLMoeIntegrationTest::test_small_model_integration_test_batch_flashatt2",
608
+ "tests/models/qwen3_vl_moe/test_modeling_qwen3_vl_moe.py::Qwen3VLMoeIntegrationTest::test_small_model_integration_test_batch_wo_image_flashatt2"
609
+ ]
610
+ },
611
+ "roberta": {
612
+ "single-gpu": [],
613
+ "multi-gpu": [
614
+ "tests/models/roberta/test_modeling_roberta.py::RobertaModelTest::test_flash_attn_2_inference_equivalence",
615
+ "tests/models/roberta/test_modeling_roberta.py::RobertaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
616
+ ]
617
+ },
618
+ "roberta_prelayernorm": {
619
+ "single-gpu": [],
620
+ "multi-gpu": [
621
+ "tests/models/roberta_prelayernorm/test_modeling_roberta_prelayernorm.py::RobertaPreLayerNormModelTest::test_flash_attn_2_inference_equivalence_right_padding"
622
+ ]
623
+ },
624
+ "sam2": {
625
+ "single-gpu": [
626
+ "tests/models/sam2/test_modeling_sam2.py::Sam2ModelTest::test_flash_attn_2_can_dispatch_composite_models"
627
+ ],
628
+ "multi-gpu": [
629
+ "tests/models/sam2/test_modeling_sam2.py::Sam2ModelTest::test_flash_attn_2_can_dispatch_composite_models"
630
+ ]
631
+ },
632
+ "smollm3": {
633
+ "single-gpu": [
634
+ "tests/models/smollm3/test_modeling_smollm3.py::SmolLM3IntegrationTest::test_model_3b_long_prompt"
635
+ ],
636
+ "multi-gpu": [
637
+ "tests/models/smollm3/test_modeling_smollm3.py::SmolLM3IntegrationTest::test_model_3b_long_prompt"
638
+ ]
639
+ },
640
+ "squeezebert": {
641
+ "single-gpu": [
642
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence",
643
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence_right_padding"
644
+ ],
645
+ "multi-gpu": [
646
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence",
647
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence_right_padding"
648
+ ]
649
+ },
650
+ "starcoder2": {
651
+ "single-gpu": [
652
+ "tests/models/starcoder2/test_modeling_starcoder2.py::Starcoder2IntegrationTest::test_starcoder2_batched_generation_fa2"
653
+ ],
654
+ "multi-gpu": [
655
+ "tests/models/starcoder2/test_modeling_starcoder2.py::Starcoder2IntegrationTest::test_starcoder2_batched_generation_fa2"
656
+ ]
657
+ },
658
+ "swin": {
659
+ "single-gpu": [
660
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence",
661
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
662
+ ],
663
+ "multi-gpu": [
664
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence",
665
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
666
+ ]
667
+ },
668
+ "swin2sr": {
669
+ "single-gpu": [
670
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence",
671
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence_right_padding"
672
+ ],
673
+ "multi-gpu": [
674
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence",
675
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence_right_padding"
676
+ ]
677
+ },
678
+ "swinv2": {
679
+ "single-gpu": [
680
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence",
681
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
682
+ ],
683
+ "multi-gpu": [
684
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence",
685
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
686
+ ]
687
+ },
688
+ "t5gemma": {
689
+ "single-gpu": [
690
+ "tests/models/t5gemma/test_modeling_t5gemma.py::T5GemmaModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
691
+ ],
692
+ "multi-gpu": [
693
+ "tests/models/t5gemma/test_modeling_t5gemma.py::T5GemmaModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
694
+ ]
695
+ },
696
+ "xmod": {
697
+ "single-gpu": [],
698
+ "multi-gpu": [
699
+ "tests/models/xmod/test_modeling_xmod.py::XmodModelTest::test_flash_attn_2_inference_equivalence_right_padding"
700
+ ]
701
+ },
702
+ "zamba": {
703
+ "single-gpu": [
704
+ "tests/models/zamba/test_modeling_zamba.py::ZambaModelTest::test_flash_attn_2_fp32_ln"
705
+ ],
706
+ "multi-gpu": [
707
+ "tests/models/zamba/test_modeling_zamba.py::ZambaModelTest::test_flash_attn_2_fp32_ln"
708
+ ]
709
+ }
710
+ }