Hugging Face's logo

Datasets:
hf-internal-testing
/
transformers_flash_attn_ci

Dataset card Files
xet
Community
hf-transformers-bot commited on
Commit
39fc2b5
·
verified ·
1 Parent(s): ce40c2f

Upload 2025-11-09/ci_results_run_models_gpu/new_failures.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. 2025-11-09/ci_results_run_models_gpu/new_failures.json +694 -0
2025-11-09/ci_results_run_models_gpu/new_failures.json ADDED
@@ -0,0 +1,694 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bamba": {
3
+ "single-gpu": [
4
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
5
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
6
+ ],
7
+ "multi-gpu": [
8
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
9
+ "tests/models/bamba/test_modeling_bamba.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
10
+ ]
11
+ },
12
+ "bark": {
13
+ "single-gpu": [
14
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_eager_matches_fa2_generate",
15
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_flash_attention_2_continue_generate_with_position_ids",
16
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_eager_matches_fa2_generate",
17
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_flash_attention_2_continue_generate_with_position_ids"
18
+ ],
19
+ "multi-gpu": [
20
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_eager_matches_fa2_generate",
21
+ "tests/models/bark/test_modeling_bark.py::BarkSemanticModelTest::test_flash_attention_2_continue_generate_with_position_ids",
22
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_eager_matches_fa2_generate",
23
+ "tests/models/bark/test_modeling_bark.py::BarkCoarseModelTest::test_flash_attention_2_continue_generate_with_position_ids"
24
+ ]
25
+ },
26
+ "bert": {
27
+ "single-gpu": [],
28
+ "multi-gpu": [
29
+ "tests/models/bert/test_modeling_bert.py::BertModelTest::test_flash_attn_2_inference_equivalence"
30
+ ]
31
+ },
32
+ "blt": {
33
+ "single-gpu": [
34
+ "tests/models/blt/test_modeling_blt.py::BltModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
35
+ ],
36
+ "multi-gpu": [
37
+ "tests/models/blt/test_modeling_blt.py::BltModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
38
+ ]
39
+ },
40
+ "chameleon": {
41
+ "single-gpu": [
42
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
43
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonVision2SeqModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
44
+ ],
45
+ "multi-gpu": [
46
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
47
+ "tests/models/chameleon/test_modeling_chameleon.py::ChameleonVision2SeqModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
48
+ ]
49
+ },
50
+ "clap": {
51
+ "single-gpu": [
52
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence",
53
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence_right_padding",
54
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence",
55
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence_right_padding"
56
+ ],
57
+ "multi-gpu": [
58
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence",
59
+ "tests/models/clap/test_modeling_clap.py::ClapAudioModelTest::test_flash_attn_2_inference_equivalence_right_padding",
60
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence",
61
+ "tests/models/clap/test_modeling_clap.py::ClapModelTest::test_flash_attn_2_inference_equivalence_right_padding"
62
+ ]
63
+ },
64
+ "colpali": {
65
+ "single-gpu": [
66
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence",
67
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence_right_padding"
68
+ ],
69
+ "multi-gpu": [
70
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence",
71
+ "tests/models/colpali/test_modeling_colpali.py::ColPaliForRetrievalModelTest::test_flash_attn_2_inference_equivalence_right_padding"
72
+ ]
73
+ },
74
+ "data2vec": {
75
+ "single-gpu": [
76
+ "tests/models/data2vec/test_modeling_data2vec_text.py::Data2VecTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
77
+ ],
78
+ "multi-gpu": []
79
+ },
80
+ "deepseek_v2": {
81
+ "single-gpu": [
82
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
83
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break",
84
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_fp32_ln"
85
+ ],
86
+ "multi-gpu": [
87
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
88
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break",
89
+ "tests/models/deepseek_v2/test_modeling_deepseek_v2.py::DeepseekV2ModelTest::test_flash_attn_2_fp32_ln"
90
+ ]
91
+ },
92
+ "deepseek_vl_hybrid": {
93
+ "single-gpu": [
94
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attention_2_continue_generate_with_position_ids",
95
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_fp32_ln",
96
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_from_config"
97
+ ],
98
+ "multi-gpu": [
99
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attention_2_continue_generate_with_position_ids",
100
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_fp32_ln",
101
+ "tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridModelTest::test_flash_attn_2_from_config"
102
+ ]
103
+ },
104
+ "diffllama": {
105
+ "single-gpu": [
106
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_generate_padding_right",
107
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence",
108
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
109
+ ],
110
+ "multi-gpu": [
111
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_generate_padding_right",
112
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence",
113
+ "tests/models/diffllama/test_modeling_diffllama.py::DiffLlamaModelTest::test_flash_attn_2_inference_equivalence_right_padding"
114
+ ]
115
+ },
116
+ "donut": {
117
+ "single-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
+ "multi-gpu": [
122
+ "tests/models/donut/test_modeling_donut_swin.py::DonutSwinModelTest::test_flash_attn_2_inference_equivalence",
123
+ "tests/models/donut/test_modeling_donut_swin.py::DonutSwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
124
+ ]
125
+ },
126
+ "efficientloftr": {
127
+ "single-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
+ "multi-gpu": [
132
+ "tests/models/efficientloftr/test_modeling_efficientloftr.py::EfficientLoFTRModelTest::test_flash_attn_2_inference_equivalence",
133
+ "tests/models/efficientloftr/test_modeling_efficientloftr.py::EfficientLoFTRModelTest::test_flash_attn_2_inference_equivalence_right_padding"
134
+ ]
135
+ },
136
+ "electra": {
137
+ "single-gpu": [],
138
+ "multi-gpu": [
139
+ "tests/models/electra/test_modeling_electra.py::ElectraModelTest::test_flash_attn_2_inference_equivalence"
140
+ ]
141
+ },
142
+ "emu3": {
143
+ "single-gpu": [
144
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence",
145
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
146
+ ],
147
+ "multi-gpu": [
148
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence",
149
+ "tests/models/emu3/test_modeling_emu3.py::Emu3Vision2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
150
+ ]
151
+ },
152
+ "exaone4": {
153
+ "single-gpu": [
154
+ "tests/models/exaone4/test_modeling_exaone4.py::Exaone4ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
155
+ ],
156
+ "multi-gpu": [
157
+ "tests/models/exaone4/test_modeling_exaone4.py::Exaone4ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
158
+ "tests/models/exaone4/test_modeling_exaone4.py::Exaone4IntegrationTest::test_model_generation_long_flash"
159
+ ]
160
+ },
161
+ "falcon": {
162
+ "single-gpu": [
163
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence",
164
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence_right_padding"
165
+ ],
166
+ "multi-gpu": [
167
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence",
168
+ "tests/models/falcon/test_modeling_falcon.py::FalconModelTest::test_flash_attn_2_inference_equivalence_right_padding"
169
+ ]
170
+ },
171
+ "flex_olmo": {
172
+ "single-gpu": [
173
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
174
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
175
+ ],
176
+ "multi-gpu": [
177
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
178
+ "tests/models/flex_olmo/test_modeling_flex_olmo.py::FlexOlmoModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
179
+ ]
180
+ },
181
+ "gemma3n": {
182
+ "single-gpu": [
183
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_equivalence",
184
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence",
185
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
186
+ ],
187
+ "multi-gpu": [
188
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_equivalence",
189
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence",
190
+ "tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
191
+ ]
192
+ },
193
+ "glm4_moe": {
194
+ "single-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
+ "multi-gpu": [
198
+ "tests/models/glm4_moe/test_modeling_glm4_moe.py::Glm4MoeModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
199
+ ]
200
+ },
201
+ "gpt2": {
202
+ "single-gpu": [
203
+ "tests/models/gpt2/test_modeling_gpt2.py::GPT2ModelLanguageGenerationTest::test_flash_attn_2_generate_padding_left"
204
+ ],
205
+ "multi-gpu": [
206
+ "tests/models/gpt2/test_modeling_gpt2.py::GPT2ModelLanguageGenerationTest::test_flash_attn_2_generate_padding_left"
207
+ ]
208
+ },
209
+ "gpt_oss": {
210
+ "single-gpu": [
211
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence",
212
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence_right_padding"
213
+ ],
214
+ "multi-gpu": [
215
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence",
216
+ "tests/models/gpt_oss/test_modeling_gpt_oss.py::GptOssModelTest::test_flash_attn_2_inference_equivalence_right_padding"
217
+ ]
218
+ },
219
+ "granitemoe": {
220
+ "single-gpu": [
221
+ "tests/models/granitemoe/test_modeling_granitemoe.py::GraniteMoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
222
+ ],
223
+ "multi-gpu": [
224
+ "tests/models/granitemoe/test_modeling_granitemoe.py::GraniteMoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
225
+ ]
226
+ },
227
+ "granitemoehybrid": {
228
+ "single-gpu": [
229
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
230
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding",
231
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence",
232
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence_right_padding"
233
+ ],
234
+ "multi-gpu": [
235
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence",
236
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::BambaModelTest::test_flash_attn_2_inference_equivalence_right_padding",
237
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence",
238
+ "tests/models/granitemoehybrid/test_modeling_granitemoehybrid.py::GraniteMoeHybridModelTest::test_flash_attn_2_inference_equivalence_right_padding"
239
+ ]
240
+ },
241
+ "granitemoeshared": {
242
+ "single-gpu": [
243
+ "tests/models/granitemoeshared/test_modeling_granitemoeshared.py::GraniteMoeSharedModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
244
+ ],
245
+ "multi-gpu": [
246
+ "tests/models/granitemoeshared/test_modeling_granitemoeshared.py::GraniteMoeSharedModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
247
+ ]
248
+ },
249
+ "grounding_dino": {
250
+ "single-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
+ "multi-gpu": [
255
+ "tests/models/grounding_dino/test_modeling_grounding_dino.py::GroundingDinoModelTest::test_flash_attn_2_inference_equivalence",
256
+ "tests/models/grounding_dino/test_modeling_grounding_dino.py::GroundingDinoModelTest::test_flash_attn_2_inference_equivalence_right_padding"
257
+ ]
258
+ },
259
+ "hunyuan_v1_moe": {
260
+ "single-gpu": [
261
+ "tests/models/hunyuan_v1_moe/test_modeling_hunyuan_v1_moe.py::HunYuanMoEV1ModelTest::test_flash_attn_2_equivalence"
262
+ ],
263
+ "multi-gpu": []
264
+ },
265
+ "instructblip": {
266
+ "single-gpu": [
267
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
268
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
269
+ ],
270
+ "multi-gpu": [
271
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
272
+ "tests/models/instructblip/test_modeling_instructblip.py::InstructBlipForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
273
+ ]
274
+ },
275
+ "instructblipvideo": {
276
+ "single-gpu": [
277
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
278
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
279
+ ],
280
+ "multi-gpu": [
281
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_fp32_ln",
282
+ "tests/models/instructblipvideo/test_modeling_instructblipvideo.py::InstructBlipVideoForConditionalGenerationDecoderOnlyTest::test_flash_attn_2_from_config"
283
+ ]
284
+ },
285
+ "janus": {
286
+ "single-gpu": [
287
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence",
288
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
289
+ ],
290
+ "multi-gpu": [
291
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence",
292
+ "tests/models/janus/test_modeling_janus.py::JanusVisionText2TextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
293
+ ]
294
+ },
295
+ "jetmoe": {
296
+ "single-gpu": [
297
+ "tests/models/jetmoe/test_modeling_jetmoe.py::JetMoeModelTest::test_flash_attn_2_fp32_ln"
298
+ ],
299
+ "multi-gpu": [
300
+ "tests/models/jetmoe/test_modeling_jetmoe.py::JetMoeModelTest::test_flash_attn_2_fp32_ln"
301
+ ]
302
+ },
303
+ "kosmos2": {
304
+ "single-gpu": [
305
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_eager_matches_fa2_generate",
306
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attention_2_continue_generate_with_position_ids",
307
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_can_dispatch_composite_models",
308
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_fp32_ln",
309
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_from_config",
310
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence",
311
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
312
+ ],
313
+ "multi-gpu": [
314
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_eager_matches_fa2_generate",
315
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attention_2_continue_generate_with_position_ids",
316
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_can_dispatch_composite_models",
317
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_fp32_ln",
318
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_from_config",
319
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence",
320
+ "tests/models/kosmos2/test_modeling_kosmos2.py::Kosmos2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
321
+ ]
322
+ },
323
+ "kosmos2_5": {
324
+ "single-gpu": [
325
+ "tests/models/kosmos2_5/test_modeling_kosmos2_5.py::Kosmos2_5ModelTest::test_flash_attn_2_can_dispatch_composite_models"
326
+ ],
327
+ "multi-gpu": [
328
+ "tests/models/kosmos2_5/test_modeling_kosmos2_5.py::Kosmos2_5ModelTest::test_flash_attn_2_can_dispatch_composite_models"
329
+ ]
330
+ },
331
+ "kyutai_speech_to_text": {
332
+ "single-gpu": [
333
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_eager_matches_fa2_generate",
334
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence",
335
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
336
+ ],
337
+ "multi-gpu": [
338
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_eager_matches_fa2_generate",
339
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence",
340
+ "tests/models/kyutai_speech_to_text/test_modeling_kyutai_speech_to_text.py::KyutaiSpeechToTextModelTest::test_flash_attn_2_inference_equivalence_right_padding"
341
+ ]
342
+ },
343
+ "lfm2": {
344
+ "single-gpu": [
345
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
346
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
347
+ ],
348
+ "multi-gpu": [
349
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
350
+ "tests/models/lfm2/test_modeling_lfm2.py::Lfm2ModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
351
+ ]
352
+ },
353
+ "lfm2_moe": {
354
+ "single-gpu": [
355
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
356
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
357
+ ],
358
+ "multi-gpu": [
359
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
360
+ "tests/models/lfm2_moe/test_modeling_lfm2_moe.py::Lfm2MoeModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
361
+ ]
362
+ },
363
+ "lfm2_vl": {
364
+ "single-gpu": [
365
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence",
366
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence_right_padding"
367
+ ],
368
+ "multi-gpu": [
369
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence",
370
+ "tests/models/lfm2_vl/test_modeling_lfm2_vl.py::Lfm2VlModelTest::test_flash_attn_2_inference_equivalence_right_padding"
371
+ ]
372
+ },
373
+ "llava_next": {
374
+ "single-gpu": [
375
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
376
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
377
+ ],
378
+ "multi-gpu": [
379
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
380
+ "tests/models/llava_next/test_modeling_llava_next.py::LlavaNextForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
381
+ ]
382
+ },
383
+ "llava_next_video": {
384
+ "single-gpu": [
385
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
386
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
387
+ ],
388
+ "multi-gpu": [
389
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
390
+ "tests/models/llava_next_video/test_modeling_llava_next_video.py::LlavaNextVideoForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
391
+ ]
392
+ },
393
+ "llava_onevision": {
394
+ "single-gpu": [
395
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
396
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
397
+ ],
398
+ "multi-gpu": [
399
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
400
+ "tests/models/llava_onevision/test_modeling_llava_onevision.py::LlavaOnevisionForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
401
+ ]
402
+ },
403
+ "mask2former": {
404
+ "single-gpu": [
405
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence",
406
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
407
+ ],
408
+ "multi-gpu": [
409
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence",
410
+ "tests/models/mask2former/test_modeling_mask2former.py::Mask2FormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
411
+ ]
412
+ },
413
+ "maskformer": {
414
+ "single-gpu": [
415
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence",
416
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding",
417
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence",
418
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
419
+ ],
420
+ "multi-gpu": [
421
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence",
422
+ "tests/models/maskformer/test_modeling_maskformer.py::MaskFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding",
423
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence",
424
+ "tests/models/maskformer/test_modeling_maskformer_swin.py::MaskFormerSwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
425
+ ]
426
+ },
427
+ "mixtral": {
428
+ "single-gpu": [
429
+ "tests/models/mixtral/test_modeling_mixtral.py::MixtralModelTest::test_flash_attn_2_equivalence"
430
+ ],
431
+ "multi-gpu": [
432
+ "tests/models/mixtral/test_modeling_mixtral.py::MixtralModelTest::test_flash_attn_2_equivalence"
433
+ ]
434
+ },
435
+ "mllama": {
436
+ "single-gpu": [
437
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_eager_matches_fa2_generate",
438
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
439
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
440
+ ],
441
+ "multi-gpu": [
442
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_eager_matches_fa2_generate",
443
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
444
+ "tests/models/mllama/test_modeling_mllama.py::MllamaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
445
+ ]
446
+ },
447
+ "mm_grounding_dino": {
448
+ "single-gpu": [
449
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence",
450
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence_right_padding"
451
+ ],
452
+ "multi-gpu": [
453
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence",
454
+ "tests/models/mm_grounding_dino/test_modeling_mm_grounding_dino.py::MMGroundingDinoModelTest::test_flash_attn_2_inference_equivalence_right_padding"
455
+ ]
456
+ },
457
+ "moshi": {
458
+ "single-gpu": [
459
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence",
460
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence_right_padding",
461
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_eager_matches_fa2_generate",
462
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_fp32_ln",
463
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_from_config",
464
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence",
465
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence_right_padding"
466
+ ],
467
+ "multi-gpu": [
468
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence",
469
+ "tests/models/moshi/test_modeling_moshi.py::MoshiDecoderTest::test_flash_attn_2_inference_equivalence_right_padding",
470
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_eager_matches_fa2_generate",
471
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_fp32_ln",
472
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_from_config",
473
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence",
474
+ "tests/models/moshi/test_modeling_moshi.py::MoshiTest::test_flash_attn_2_inference_equivalence_right_padding"
475
+ ]
476
+ },
477
+ "nemotron": {
478
+ "single-gpu": [
479
+ "tests/models/nemotron/test_modeling_nemotron.py::NemotronModelTest::test_flash_attn_2_equivalence"
480
+ ],
481
+ "multi-gpu": [
482
+ "tests/models/nemotron/test_modeling_nemotron.py::NemotronModelTest::test_flash_attn_2_equivalence"
483
+ ]
484
+ },
485
+ "olmo": {
486
+ "single-gpu": [
487
+ "tests/models/olmo/test_modeling_olmo.py::OlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
488
+ ],
489
+ "multi-gpu": [
490
+ "tests/models/olmo/test_modeling_olmo.py::OlmoModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
491
+ ]
492
+ },
493
+ "olmo2": {
494
+ "single-gpu": [
495
+ "tests/models/olmo2/test_modeling_olmo2.py::Olmo2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
496
+ ],
497
+ "multi-gpu": [
498
+ "tests/models/olmo2/test_modeling_olmo2.py::Olmo2ModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
499
+ ]
500
+ },
501
+ "omdet_turbo": {
502
+ "single-gpu": [
503
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence",
504
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence_right_padding"
505
+ ],
506
+ "multi-gpu": [
507
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence",
508
+ "tests/models/omdet_turbo/test_modeling_omdet_turbo.py::OmDetTurboModelTest::test_flash_attn_2_inference_equivalence_right_padding"
509
+ ]
510
+ },
511
+ "oneformer": {
512
+ "single-gpu": [
513
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence",
514
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
515
+ ],
516
+ "multi-gpu": [
517
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence",
518
+ "tests/models/oneformer/test_modeling_oneformer.py::OneFormerModelTest::test_flash_attn_2_inference_equivalence_right_padding"
519
+ ]
520
+ },
521
+ "paligemma": {
522
+ "single-gpu": [
523
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_from_config",
524
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
525
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
526
+ ],
527
+ "multi-gpu": [
528
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_from_config",
529
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
530
+ "tests/models/paligemma/test_modeling_paligemma.py::PaliGemmaForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
531
+ ]
532
+ },
533
+ "paligemma2": {
534
+ "single-gpu": [
535
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_from_config",
536
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
537
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
538
+ ],
539
+ "multi-gpu": [
540
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_from_config",
541
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence",
542
+ "tests/models/paligemma2/test_modeling_paligemma2.py::PaliGemma2ForConditionalGenerationModelTest::test_flash_attn_2_inference_equivalence_right_padding"
543
+ ]
544
+ },
545
+ "pegasus_x": {
546
+ "single-gpu": [
547
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence",
548
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence_right_padding",
549
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence",
550
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence_right_padding"
551
+ ],
552
+ "multi-gpu": [
553
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence",
554
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXModelTest::test_flash_attn_2_inference_equivalence_right_padding",
555
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence",
556
+ "tests/models/pegasus_x/test_modeling_pegasus_x.py::PegasusXStandaloneDecoderModelTest::test_flash_attn_2_inference_equivalence_right_padding"
557
+ ]
558
+ },
559
+ "perception_lm": {
560
+ "single-gpu": [
561
+ "tests/models/perception_lm/test_modeling_perception_lm.py::PerceptionLMForConditionalGenerationModelTest::test_flash_attention_2_continue_generate_with_position_ids"
562
+ ],
563
+ "multi-gpu": [
564
+ "tests/models/perception_lm/test_modeling_perception_lm.py::PerceptionLMForConditionalGenerationModelTest::test_flash_attention_2_continue_generate_with_position_ids"
565
+ ]
566
+ },
567
+ "phi": {
568
+ "single-gpu": [
569
+ "tests/models/phi/test_modeling_phi.py::PhiModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
570
+ ],
571
+ "multi-gpu": [
572
+ "tests/models/phi/test_modeling_phi.py::PhiModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids"
573
+ ]
574
+ },
575
+ "phimoe": {
576
+ "single-gpu": [
577
+ "tests/models/phimoe/test_modeling_phimoe.py::PhimoeModelTest::test_flash_attn_2_equivalence"
578
+ ],
579
+ "multi-gpu": [
580
+ "tests/models/phimoe/test_modeling_phimoe.py::PhimoeModelTest::test_flash_attn_2_equivalence",
581
+ "tests/models/phimoe/test_modeling_phimoe.py::PhimoeModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids_and_fa_kwargs"
582
+ ]
583
+ },
584
+ "pixtral": {
585
+ "single-gpu": [
586
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence",
587
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence_right_padding"
588
+ ],
589
+ "multi-gpu": [
590
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence",
591
+ "tests/models/pixtral/test_modeling_pixtral.py::PixtralVisionModelModelTest::test_flash_attn_2_inference_equivalence_right_padding"
592
+ ]
593
+ },
594
+ "qwen2_5_vl": {
595
+ "single-gpu": [
596
+ "tests/models/qwen2_5_vl/test_modeling_qwen2_5_vl.py::Qwen2_5_VLIntegrationTest::test_small_model_integration_test_batch_wo_image_flashatt2"
597
+ ],
598
+ "multi-gpu": [
599
+ "tests/models/qwen2_5_vl/test_modeling_qwen2_5_vl.py::Qwen2_5_VLIntegrationTest::test_small_model_integration_test_batch_wo_image_flashatt2"
600
+ ]
601
+ },
602
+ "qwen3_omni_moe": {
603
+ "single-gpu": [
604
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniThinkerForConditionalGenerationModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
605
+ "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",
606
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniModelIntegrationTest::test_small_model_integration_test_batch_flashatt2"
607
+ ],
608
+ "multi-gpu": [
609
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniThinkerForConditionalGenerationModelTest::test_flash_attention_2_padding_matches_padding_free_with_position_ids",
610
+ "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",
611
+ "tests/models/qwen3_omni_moe/test_modeling_qwen3_omni_moe.py::Qwen2_5OmniModelIntegrationTest::test_small_model_integration_test_batch_flashatt2"
612
+ ]
613
+ },
614
+ "sam2": {
615
+ "single-gpu": [
616
+ "tests/models/sam2/test_modeling_sam2.py::Sam2ModelTest::test_flash_attn_2_can_dispatch_composite_models"
617
+ ],
618
+ "multi-gpu": [
619
+ "tests/models/sam2/test_modeling_sam2.py::Sam2ModelTest::test_flash_attn_2_can_dispatch_composite_models"
620
+ ]
621
+ },
622
+ "smollm3": {
623
+ "single-gpu": [
624
+ "tests/models/smollm3/test_modeling_smollm3.py::SmolLM3IntegrationTest::test_model_3b_long_prompt"
625
+ ],
626
+ "multi-gpu": [
627
+ "tests/models/smollm3/test_modeling_smollm3.py::SmolLM3IntegrationTest::test_model_3b_long_prompt"
628
+ ]
629
+ },
630
+ "squeezebert": {
631
+ "single-gpu": [
632
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence",
633
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence_right_padding"
634
+ ],
635
+ "multi-gpu": [
636
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence",
637
+ "tests/models/squeezebert/test_modeling_squeezebert.py::SqueezeBertModelTest::test_flash_attn_2_inference_equivalence_right_padding"
638
+ ]
639
+ },
640
+ "starcoder2": {
641
+ "single-gpu": [
642
+ "tests/models/starcoder2/test_modeling_starcoder2.py::Starcoder2IntegrationTest::test_starcoder2_batched_generation_fa2"
643
+ ],
644
+ "multi-gpu": [
645
+ "tests/models/starcoder2/test_modeling_starcoder2.py::Starcoder2IntegrationTest::test_starcoder2_batched_generation_fa2"
646
+ ]
647
+ },
648
+ "swin": {
649
+ "single-gpu": [
650
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence",
651
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
652
+ ],
653
+ "multi-gpu": [
654
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence",
655
+ "tests/models/swin/test_modeling_swin.py::SwinModelTest::test_flash_attn_2_inference_equivalence_right_padding"
656
+ ]
657
+ },
658
+ "swin2sr": {
659
+ "single-gpu": [
660
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence",
661
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence_right_padding"
662
+ ],
663
+ "multi-gpu": [
664
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence",
665
+ "tests/models/swin2sr/test_modeling_swin2sr.py::Swin2SRModelTest::test_flash_attn_2_inference_equivalence_right_padding"
666
+ ]
667
+ },
668
+ "swinv2": {
669
+ "single-gpu": [
670
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence",
671
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
672
+ ],
673
+ "multi-gpu": [
674
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence",
675
+ "tests/models/swinv2/test_modeling_swinv2.py::Swinv2ModelTest::test_flash_attn_2_inference_equivalence_right_padding"
676
+ ]
677
+ },
678
+ "t5gemma": {
679
+ "single-gpu": [
680
+ "tests/models/t5gemma/test_modeling_t5gemma.py::T5GemmaModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
681
+ ],
682
+ "multi-gpu": [
683
+ "tests/models/t5gemma/test_modeling_t5gemma.py::T5GemmaModelTest::test_flash_attn_2_can_compile_with_attention_mask_None_without_graph_break"
684
+ ]
685
+ },
686
+ "zamba": {
687
+ "single-gpu": [
688
+ "tests/models/zamba/test_modeling_zamba.py::ZambaModelTest::test_flash_attn_2_fp32_ln"
689
+ ],
690
+ "multi-gpu": [
691
+ "tests/models/zamba/test_modeling_zamba.py::ZambaModelTest::test_flash_attn_2_fp32_ln"
692
+ ]
693
+ }
694
+ }