-
I've maked dataset for SFT training. After SFT training (with Unsloth) i decide to increase results with ORPO training. But ORPO training don't doing anything, I have the same results after ORPO train…
-
### System Info
transformers: 4.39.3
python: 3.10.12
### Who can help?
_No response_
### Information
- [X] The official example scripts
- [ ] My own modified scripts
### Tasks
-…
-
I'm using a Mistral model and want to only train on responses. `train_on_responses_only` is supposed to only mask the user prompt, however, the following code masks **both** the user and assistant mes…
-
### Bug description
Hi there! I have previously created my first `LightningDataModule`. More specifically, a `NonGeoDataModule` which inherits from there (see [torchgeo-fork](https://github.com/Mat…
-
### Describe the bug
I used the code below to train an EfficientAD model for binary anomaly detection task.
```python
# Initialize the datamodule, model and engine
datamodule = Folder(
name…
-
### Feature request
This request proposes one of three changes (see **Motivation** for background, and **Your contribution** more thoughts on possible solutions) in order to allow saving of a certa…
-
### System Info
transformers==4.37.2 python 3
### Who can help?
_No response_
### Information
- [ ] The official example scripts
- [X] My own modified scripts
### Tasks
- [ ] An officially supp…
-
-
```
from trl import SFTTrainer
from transformers import TrainingArguments, DataCollatorForSeq2Seq
from unsloth import is_bfloat16_supported
trainer = SFTTrainer(
model = model,
tokeniz…
-
Trying to finetune Mistral Small
```
from unsloth.chat_templates import train_on_responses_only
trainer = train_on_responses_only(
trainer,
instruction_part = "[INST]",
response_part…