---
base_model: nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16
library_name: peft
license: other
tags:
  - lora
  - peft
  - adapter
  - adaption
---

# adaption_adventure_travel_assistant

## Model Training

A LORA adapter for `nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16`. This model was trained with SFT using [Adaption](https://adaptionlabs.ai)'s AutoScientist on the adventure_travel_assistant dataset.


![Training metrics](training-metrics.png)

### AutoScientist Config

```json
{
  "job_id": "a1440533-e5fa-414f-86c6-bb263cf70568",
  "training_experiment_id": "1f1428e7-5c64-4e8f-9b1d-87dd40d124ca",
  "original_model_name": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
  "trained_model_name": "adaption_adventure_travel_assistant",
  "training_method": "sft",
  "training_type": "lora",
  "data_format": "chat",
  "hyperparams": {
    "lora": "true",
    "lora_r": 8,
    "n_evals": 5,
    "n_epochs": 1,
    "batch_size": "max",
    "lora_alpha": 8,
    "lora_dropout": 0,
    "min_lr_ratio": 0.1,
    "warmup_ratio": 0.1,
    "weight_decay": 0,
    "learning_rate": 0.0001,
    "max_grad_norm": 2,
    "base_model_size": "120B",
    "train_on_inputs": "false",
    "training_method": "sft",
    "lr_scheduler_type": "cosine",
    "scheduler_num_cycles": 0.5,
    "lora_trainable_modules": "q_proj,v_proj"
  }
}
```

## Training Data

The model was trained on 8,724 rows of adapted data with the following domain distribution: travel (56%), geography (6%), code (4%), fitness-sports (4%), culture (3%), math (3%), academic-education (3%), corporate-business (3%), cooking (3%), history (3%), writing-editing-communication (3%), animal-nature (3%), transportation (2%), how-to (1%), science (0%), entertainment (0%), sports (0%), governance (0%), technology (0%), language (0%), medical (0%), religion (0%), product-advice (0%), personal-finance (0%), architecture-design (0%), games (0%), legal (0%), music (0%), career-workplace (0%), news (0%), marketing (0%), art (0%), social (0%), agriculture (0%), dating (0%), literature (0%), fashion-beauty (0%), parenting-family (0%), data-analysis-visualization (0%).

## Model Evaluation

The model was evaluated on an in-distribution held-out test set as well as a broader domain-specific test set to measure generalization.


![Win rates](win-rates.png)

| Domain | Win rate vs. base model |
| --- | --- |
| travel | 89% |

## How to use

```bash
pip install torch transformers peft
```

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

BASE = "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16"
ADAPTER = "<this-repo-id>"

device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float32 if device == "cpu" else torch.bfloat16

base = AutoModelForCausalLM.from_pretrained(BASE, dtype=dtype).to(device)
model = PeftModel.from_pretrained(base, ADAPTER)
# Optional: merge the LoRA weights into the base for faster inference
model = model.merge_and_unload()
model.eval()

tokenizer = AutoTokenizer.from_pretrained(BASE)
messages = [{"role": "user", "content": "Hello!"}]
text = tokenizer.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(device)

with torch.inference_mode():
    out = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
```