---
base_model:
- google/gemma-2-2b-it
- VAGOsolutions/SauerkrautLM-gemma-2-2b-it
- stvlynn/Gemma-2-2b-Chinese-it
library_name: transformers
tags:
- mergekit
- merge
license: apache-2.0
---
# Gemma2-2B-it Merged Fine-Tuned Models for Chinese & German understanding

Lightweight language model based on Gemma2 2B created by merging multiple fine tuned Gemma2-2B-IT versions to test multilingual conversation capabilities in specialized low parameter language models.

## 🤏 Models Merged
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
This model was merged using the [Model Stock](https://arxiv.org/abs/2403.19522) merge method using [google/gemma-2-2b-it](https://huggingface.co/google/gemma-2-2b-it) as a base.

The following models were included in the merge:
* [VAGOsolutions/SauerkrautLM-gemma-2-2b-it](https://huggingface.co/VAGOsolutions/SauerkrautLM-gemma-2-2b-it)
* [stvlynn/Gemma-2-2b-Chinese-it](https://huggingface.co/stvlynn/Gemma-2-2b-Chinese-it)

## 🧩 Configuration

The following YAML configuration was used to produce this model:

```yaml
models:
  - model: google/gemma-2-2b-it
  - model: VAGOsolutions/SauerkrautLM-gemma-2-2b-it
  - model: stvlynn/Gemma-2-2b-Chinese-it
merge_method: model_stock
base_model: google/gemma-2-2b-it
dtype: bfloat16

```

### 💻 Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("AdamLucek/gemma2-2b-it-chinese-german")
model = AutoModelForCausalLM.from_pretrained(
    "AdamLucek/gemma2-2b-it-chinese-german",
    device_map="cuda",
    torch_dtype=torch.bfloat16
)

# Prepare the input text
input_text = "请解释一下量子力学中的叠加原理，并举例说明该原理在实际应用中的重要性和挑战。"
input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")

# Generate the output
outputs = model.generate(
    **input_ids,
    max_new_tokens=256,
    pad_token_id=tokenizer.eos_token_id
)

# Decode and print the generated text
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```