Update all references to canonical models (#29001)

* Script & Manual edition

* Update
This commit is contained in:
Lysandre Debut
2024-02-16 08:16:58 +01:00
committed by GitHub
parent 1e402b957d
commit f497f564bb
561 changed files with 2682 additions and 2687 deletions

View File

@@ -89,8 +89,8 @@ class StreamerTester(unittest.TestCase):
# Tests that we can pass `decode_kwargs` to the streamer to control how the tokens are decoded. Must be tested
# with actual models -- the dummy models' tokenizers are not aligned with their models, and
# `skip_special_tokens=True` has no effect on them
tokenizer = AutoTokenizer.from_pretrained("distilgpt2")
model = AutoModelForCausalLM.from_pretrained("distilgpt2").to(torch_device)
tokenizer = AutoTokenizer.from_pretrained("distilbert/distilgpt2")
model = AutoModelForCausalLM.from_pretrained("distilbert/distilgpt2").to(torch_device)
model.config.eos_token_id = -1
input_ids = torch.ones((1, 5), device=torch_device).long() * model.config.bos_token_id