File size: 2,099 Bytes
7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 7935d06 d4bcc51 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 |
---
license: gemma
language:
- sl
- en
base_model:
- cjvt/GaMS-9B-Instruct
datasets:
- nvidia/Nemotron-Post-Training-Dataset-v1
pipeline_tag: text-generation
---
# Model Card for GaMS-9B-Instruct-Nemotron
**GaMS-9B-Instruct-Nemotron** is a variant of GaMS-9B-Instruct, further trained with supervised fine-tuning (SFT) on a curated subset of the `chat` part of [*nvidia/Nemotron-Post-Training-Dataset-v1*](https://huggingface.co/datasets/nvidia/Nemotron-Post-Training-Dataset-v1).
The training data included \~80k Slovenian instruction–response pairs (based on translations with additional modifications to adjust identity and context) and \~20k English examples.
This instruction tuned version of GaMS 9B was developed as part of a master thesis project by Timotej Petrič. Further details will follow.
## Acknowledgment
The model was developed within the [PoVeJMo](https://www.cjvt.si/povejmo/en/project/) research program (Adaptive Natural Language Processing with Large Language Models), particularly within the research project titled SloLLaMai -- Open-access computationally efficient models for Slovenian. The program is funded within the Recovery and Resilience Plan by the Slovenian Research and Innovation Agency (ARIS) and NextGenerationEU. The authors also acknowledge the financial support from the Slovenian Research and Innovation Agency (research core funding No. P6-0411 -- Language Resources and Technologies for Slovene).
## Usage
The model can be run through `pipeline` API using the following code:
```python
from transformers import pipeline
model_id = "GaMS-Beta/GaMS-9B-Instruct-Nemotron"
pline = pipeline(
"text-generation",
model=model_id,
device_map="cuda" # replace with "mps" to run on a Mac device
)
# Example of response generation
message = [{"role": "user", "content": f"Pozdravljen. Kdo si?"}]
response = pline(message, max_new_tokens=512)
print("Slovene translation:", response[0]["generated_text"][-1]["content"])
```
|