Skip to content

Fix tests trainer again #42933

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
SunMarc wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Fix tests trainer again #42933

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
25962f3
fix
SunMarc Dec 17, 2025
ff70cf8
be careful on checking generate
SunMarc Dec 17, 2025
b68e7d5
fix tied
SunMarc Dec 17, 2025
e3e5d39
fix
SunMarc Dec 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion src/transformers/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2349,7 +2349,8 @@ def _inner_training_loop(
if self.is_fsdp_enabled:
self.model = self.model_wrapped = model
# Fix `got mixed torch.Tensor and DTensor` error in model.generate() for FSDP2 with LoRA
dist.fsdp.register_fsdp_forward_method(self.model, "generate")
if hasattr(self.model, "generate"):
dist.fsdp.register_fsdp_forward_method(self.model, "generate")

# for the rest of this function `model` is the outside model, whether it was wrapped or not
if model is not self.model:
Expand Down
6 changes: 4 additions & 2 deletions tests/deepspeed/test_deepspeed.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -280,6 +280,8 @@ def test_init_zero3_missing_params(self):
from transformers.models.gpt2.modeling_gpt2 import GPT2PreTrainedModel

class TinyGPT2WithUninitializedWeights(GPT2PreTrainedModel):
_tied_weights_keys = {"lm_head.weight": "transformer.wte.weight"}

def __init__(self, config):
super().__init__(config)
self.transformer = AutoModel.from_pretrained(GPT2_TINY, config=config)
Expand Down Expand Up @@ -1064,10 +1066,10 @@ def _add_eos_to_examples(example):
return example

def _convert_to_features(example_batch):
input_encodings = tokenizer.batch_encode_plus(
input_encodings = tokenizer(
example_batch["input_text"], padding="max_length", max_length=512, truncation=True
)
target_encodings = tokenizer.batch_encode_plus(
target_encodings = tokenizer(
example_batch["target_text"], padding="max_length", max_length=16, truncation=True
)

Expand Down
15 changes: 4 additions & 11 deletions tests/trainer/test_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1541,9 +1541,7 @@ def test_multiple_peft_adapters(self):
tiny_model = get_peft_model(tiny_model, peft_config, "adapter1")
tiny_model.add_adapter("adapter2", peft_config)

max_len_single_sentence = self.model_max_length - self.num_special_tokens_to_add(pair=False)

train_dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, max_len_single_sentence)
train_dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, 100)

tokenizer.pad_token = tokenizer.eos_token

Expand Down Expand Up @@ -3733,9 +3731,7 @@ def test_trainer_eval_multiple(self):
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(MODEL_ID)

max_len_single_sentence = self.model_max_length - self.num_special_tokens_to_add(pair=False)

dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, max_len_single_sentence)
dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, 100)
with tempfile.TemporaryDirectory() as tmp_dir:
training_args = TrainingArguments(
output_dir=tmp_dir,
Expand All @@ -3759,8 +3755,7 @@ def test_trainer_eval_multiple(self):
def test_trainer_eval_lm(self):
MODEL_ID = "distilbert/distilroberta-base"
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
max_len_single_sentence = self.model_max_length - self.num_special_tokens_to_add(pair=False)
dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, max_len_single_sentence)
dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, 100)
self.assertEqual(len(dataset), 31)

def test_training_iterable_dataset(self):
Expand Down Expand Up @@ -4942,10 +4937,8 @@ def test_trainer_works_without_model_config(self):

tokenizer = AutoTokenizer.from_pretrained("hf-internal-testing/tiny-random-LlamaForCausalLM")
model = BasicTextGenerationModel(vocab_size=tokenizer.vocab_size, hidden_size=32)
# Note that this class does not have a config attribute
max_len_single_sentence = tokenizer.model_max_length - tokenizer.num_special_tokens_to_add(pair=False)

train_dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, max_len_single_sentence)
train_dataset = get_dataset(PATH_SAMPLE_TEXT, tokenizer, 100)

with tempfile.TemporaryDirectory() as tmpdir:
training_args = TrainingArguments(
Expand Down