| CHANGED |
bamba |
test_simple_batched_generate_with_padding |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
bamba |
test_simple_generate |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
cohere2_vision |
test_model_integration_generate_chat_template |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
deepseek_v2 |
test_deepseek_v2_lite |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
deepseek_v2 |
test_logits_eager |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
emu3 |
test_model_generation |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
emu3 |
test_model_generation_batched |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
emu3 |
test_model_generation_multi_image |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
flava |
test_inference |
AssertionError: string != |
AssertionError: string != |
| NEW |
gemma |
test_model_7b_fp16_static_cache |
AssertionError: Lists differ |
N/A |
| CHANGED |
gemma4 |
test_export_text_only |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
glm |
test_model_9b_fp16 |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
glm |
test_model_9b_sdpa |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
glm4_moe |
test_compile_static_cache |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
glm4_moe_lite |
test_compile_static_cache |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
llama4 |
test_model_17b_16e_batch |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
llama4 |
test_model_17b_16e_fp32 |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
llava |
test_pixtral |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
llava |
test_pixtral_4bit |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
mamba2 |
test_batched_equivalence_with_cache |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
mamba2 |
test_batched_equivalence_without_cache |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
mamba2 |
test_simple_generate |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
mimi |
test_integration |
AssertionError: np.False_ |
AssertionError |
| CHANGED |
mimi |
test_integration_longform |
AssertionError: np.False_ |
AssertionError |
| NEW |
minicpm3 |
test_eager_padding_matches_padding_free_with_position_ids |
AssertionError: Tensor-likes |
N/A |
| NEW |
minicpm3 |
test_sdpa_padding_matches_padding_free_with_position_ids |
AssertionError: Tensor-likes |
N/A |
| CHANGED |
modernvbert |
test_masked_lm_inference |
RepositoryNotFoundError |
RepositoryNotFoundError |
| NEW |
moshi |
test_eager_padding_matches_padding_free_with_position_ids |
AssertionError: Tensor-likes |
N/A |
| CHANGED |
moshi |
test_moshika_greedy_unconditional_fp16 |
AssertionError: np.False_ |
AssertionError |
| CHANGED |
moshi |
test_moshiko_greedy_unconditional_fp16 |
AssertionError: np.False_ |
AssertionError |
| CHANGED |
moshi |
test_moshiko_greedy_unconditional_fp32 |
AssertionError: np.False_ |
AssertionError |
| CHANGED |
muse_glimmer |
test_image_generation_matches_reference |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
muse_glimmer |
test_text_generation_matches_reference |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
olmo |
test_model_7b_logits |
OutOfMemoryError |
OutOfMemoryError |
| NEW |
olmo |
test_model_7b_twin_2t_logits |
OutOfMemoryError |
N/A |
| CHANGED |
olmo3 |
test_generate_beyond_sliding_window |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
olmo3 |
test_real_model_7b_greedy_generation |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
pi0 |
test_train_pi0_base_libero |
OutOfMemoryError |
OutOfMemoryError |
| CHANGED |
vision_encoder_decoder |
test_forward_pass |
RemoteEntryNotFoundError |
RemoteEntryNotFoundError |
| CHANGED |
vision_encoder_decoder |
test_generation |
RemoteEntryNotFoundError |
RemoteEntryNotFoundError |
| CHANGED |
wav2vec2_with_lm |
test_save_load_pretrained_additional_features |
FileExistsError |
FileExistsError |
| CHANGED |
wav2vec2_with_lm |
test_save_load_pretrained_default |
FileExistsError |
FileExistsError |
@ydshieh here is the diff of single GPU failures between 08/13 and 08/14 CI runs, covering all error types for
models_*keys.Status legend:
NEW— test not present in 08/13 single GPU failures (may exist in 08/13 multi GPU)CHANGED— test present in both days but trace content changedQuick analysis:
mimix2,moshix3. Numpy changed how it represents booleans (False->np.False_). Not a model regression. Prev error was genericAssertionErrorbefore thenp.False_repr appeared in the trace.minicpm3x2,moshix1. Appeared new in 08/14 single GPU.minicpm3::test_eager_paddingwas already in 08/13 multi GPU; the other two were not in 08/13 at all.gemma::test_model_7b_fp16_static_cache. Fixed by [Gemma] Update expected values for A10G #47976 (merged).flava::test_inference. Diff got closer to expected (0.155 -> 0.0005), barely still fails 4-decimal tolerance.modernvbert. Repo access issue, pre-existing.vision_encoder_decoder. Remote file missing, pre-existing.wav2vec2_with_lm. File conflict in test teardown, pre-existing.