[tests] fix blip2 edge case #40699
[tests] fix blip2 edge case #40699
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
|
run-slow: blip_2 |
|
This comment contains run-slow, running the specified jobs: models: ['models/blip_2'] |
| "I will remove it now.\n\n" | ||
| "See https://github.com/pypa/pip/issues/5466 for details.\n" | ||
| ).format(stale_egg_info) | ||
| f"Warning: {stale_egg_info} exists.\n\n" |
There was a problem hiding this comment.
make fixup corrected this 👀
| pass | ||
|
|
||
| @parameterized.expand(TEST_EAGER_MATCHES_SDPA_INFERENCE_PARAMETERIZATION) | ||
| @unittest.skip("Won't fix: Blip2 + T5 backbone needs custom input preparation for this test") |
There was a problem hiding this comment.
can be fixed, but I don't think it's worth it 👀
There was a problem hiding this comment.
agreed, let's assume that t5 passes the test so we are fine :)
|
run-slow: blip_2 |
|
This comment contains run-slow, running the specified jobs: models: ['models/blip_2'] |
|
|
||
| if model.config.get_text_config(decoder=True).is_encoder_decoder: | ||
| if model.config.is_encoder_decoder: | ||
| self.assertTrue(output_generate.sequences.shape[1] == self.max_new_tokens + 1) |
There was a problem hiding this comment.
cool thanks, can you check if Florence2 model works with these changes? It is same as blip and uses Bart as backbone
| pass | ||
|
|
||
| @parameterized.expand(TEST_EAGER_MATCHES_SDPA_INFERENCE_PARAMETERIZATION) | ||
| @unittest.skip("Won't fix: Blip2 + T5 backbone needs custom input preparation for this test") |
There was a problem hiding this comment.
agreed, let's assume that t5 passes the test so we are fine :)
| [0, 3, 7, 152, 2515, 11389, 3523, 1], | ||
| "san francisco", # TODO: check if this is ok | ||
| ), | ||
| ("cuda", None): ([0, 3, 7, 152, 2515, 11389, 3523, 1], "san francisco"), |
There was a problem hiding this comment.
i believe the CUDA expectation incorrect from the beginning
|
run-slow: florence2 |
|
This comment contains run-slow, running the specified jobs: models: ['models/florence2'] |
|
@zucchini-nlp florence2 slow tests were mostly green (there were 3 FA2 failures, which are unrelated to this PR) ✅ In florence2, |
|
[For maintainers] Suggested jobs to run (before merge) run-slow: blip_2 |
What does this PR do?
(Carved from #40553, which is becoming messy)
Our tests have the
model.config.get_text_config(decoder=True).is_encoder_decoderpattern, which doesn't make sense -- we're pulling the decoder if it exists, and then checking if it is encoder-decoder.The pattern exists because of
blip2, which wasn't settingis_encoder_decodercorrectly -- if its LLM is an encoder-decoder model, then it is also an encoder-decoder model.✅ blip 2 slow tests are passing