Fix model integration ci (#26322)

* fix wav2vec2

* nit

* stash

* one more file to update

* fix byt5

* vocab size is 256, don't change that!

* use other revision

* test persimon in smaller size

* style

* tests

* nits

* update add tokens from pretrained

* test tokenization

* nits

* potential fnet fix?

* more nits

* nits

* correct test

* assert close

* udpate

* ouch

* fix it

* some more nits

* FINALLU

* use `adept` checkpoints

* more adept checkpoints

* that was invlved!
This commit is contained in:
Arthur
2023-10-02 13:55:46 +02:00
committed by GitHub
parent 6824461f2a
commit 63864e057f
9 changed files with 38 additions and 18 deletions

View File

@@ -386,11 +386,13 @@ class PersimmonIntegrationTest(unittest.TestCase):
@slow
def test_model_8b_chat_logits(self):
input_ids = [1, 306, 4658, 278, 6593, 310, 2834, 338]
model = PersimmonForCausalLM.from_pretrained("ArthurZ/persimmon-8b-chat", device_map="auto")
model = PersimmonForCausalLM.from_pretrained(
"adept/persimmon-8b-chat", device_map="auto", torch_dtype=torch.float16
)
out = model(torch.tensor([input_ids])).logits
EXPECTED_MEAN = torch.tensor(
[[-11.2879, -11.2628, -11.2498, -11.2534, -11.2676, -11.2638, -11.2501, -11.2431]], dtype=torch.float32
[[-11.2879, -11.2628, -11.2498, -11.2534, -11.2676, -11.2638, -11.2501, -11.2431]], dtype=torch.float16
)
torch.testing.assert_close(out.cpu().mean(-1), EXPECTED_MEAN, atol=1e-4, rtol=1e-4)
# fmt: off
@@ -403,9 +405,11 @@ class PersimmonIntegrationTest(unittest.TestCase):
def test_model_8b_chat_greedy_generation(self):
EXPECTED_TEXT_COMPLETION = """human: Simply put, the theory of relativity states that?\n\nadept: The theory of relativity states that the laws of physics are the same for all observers, regardless of their relative motion."""
prompt = "human: Simply put, the theory of relativity states that?\n\nadept:"
tokenizer = AutoTokenizer.from_pretrained("ArthurZ/persimmon-8b-chat", use_fast=False)
tokenizer = AutoTokenizer.from_pretrained("adept/persimmon-8b-chat", use_fast=False)
input_ids = tokenizer.encode(prompt, return_tensors="pt").to(torch_device)
model = PersimmonForCausalLM.from_pretrained("ArthurZ/persimmon-8b-chat").to(torch_device)
model = PersimmonForCausalLM.from_pretrained("adept/persimmon-8b-chat", torch_dtype=torch.float16).to(
torch_device
)
# greedy generation outputs
generated_ids = model.generate(input_ids, max_new_tokens=64)