# 'No Log' for validation loss during training with Trainer
**URL:**
**Author:**  [@shoang](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fshoang)%5C
**Post date:** [May 16, 2023, 7:47pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fno-log-for-validation-loss-during-training-with-trainer%2F40094%2F1 "2023-05-16T19:47:22Z")
Hello,
I’m trying to fine-tune a Custom BertModel on a sequence classification task, but I’m having some issues getting the Trainer to log the validation loss. Specifically, the log looks like this:
Here is the code I’m using:
```auto
from datasets import load_dataset
import torch.nn as nn
from transformers import (
AutoModel,
AutoModelForSequenceClassification,
AutoTokenizer,
DataCollatorWithPadding,
Trainer,
TrainingArguments
)
class CustomTrainer(Trainer):
def compute_loss(self, model, inputs, return_outputs=False):
# print(inputs)
criterion = nn.CrossEntropyLoss()
outputs = model(**inputs)
loss = criterion(outputs, inputs['labels'])
return (loss, outputs) if return_outputs else loss
class CustomModel(nn.Module):
def __init__ (self, d_out):
super(CustomModel, self). __init__ ()
self.encoder = AutoModel.from_pretrained('bert-base-uncased')
d_in = self.encoder.config.hidden_size
self.fc1 = nn.Linear(d_in, d_in)
self.activation = nn.ReLU()
self.fc2 = nn.Linear(d_in, d_out)
def forward(self, input_ids, **kwargs):
x = self.encoder(input_ids=input_ids).pooler_output
x = self.fc1(x)
x = self.activation(x)
logits = self.fc2(x)
return logits
def tokenize_function(example):
return tokenizer(example["sentence1"], example["sentence2"], truncation=True)
raw_datasets = load_dataset("glue", "mrpc")
checkpoint = "bert-base-uncased"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
tokenized_datasets = raw_datasets.map(tokenize_function, batched=True)
data_collator = DataCollatorWithPadding(tokenizer=tokenizer)
training_args = TrainingArguments(
'test-trainer',
report_to='none',
save_strategy='no',
evaluation_strategy='steps',
eval_steps=50
)
# model = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)
model = CustomModel(2)
trainer = CustomTrainer(
model,
training_args,
train_dataset=tokenized_datasets["train"],
eval_dataset=tokenized_datasets["validation"],
data_collator=data_collator,
tokenizer=tokenizer,
)
trainer.train()
```
When I use a regular BertModelForSequenceClassification, it works fine. Is there something I’m missing?
---
**Author:**  [@tatiana-merz](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Ftatiana-merz)%5C
**Post date:** [June 29, 2023, 3:31pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fno-log-for-validation-loss-during-training-with-trainer%2F40094%2F2 "2023-06-29T15:31:15Z")
Could be due to the low amount of training data that you have. Try setting logging\_steps in TrainingArguments for 100 or less.
---
**Author:**  [@hfskb](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fhfskb)%5C
**Post date:** [May 15, 2024, 4:46am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fno-log-for-validation-loss-during-training-with-trainer%2F40094%2F3 "2024-05-15T04:46:13Z")
For LLM validation loss you might also need `label_names=[]` in the TrainingArguments and set `trainer.can_return_loss = True` to satisfy [transformers/src/transformers/trainer.py at v4.40.2 · huggingface/transformers · GitHub](https://github.com/huggingface/transformers/blob/v4.40.2/src/transformers/trainer.py#L3796) for some models.
---
**Author:**  [@nik-55](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fnik-55)%5C
**Post date:** [September 6, 2025, 1:30pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fno-log-for-validation-loss-during-training-with-trainer%2F40094%2F4 "2025-09-06T13:30:52Z")
I am facing issue with gemma 3 model for getting validation loss as “No Log”. After above changes, it worked.