Instructions to use BatboldSamdan/babybata2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BatboldSamdan/babybata2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="BatboldSamdan/babybata2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("BatboldSamdan/babybata2")
model = AutoModelForCausalLM.from_pretrained("BatboldSamdan/babybata2", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

PEFT
How to use BatboldSamdan/babybata2 with PEFT:
```
Task type is invalid.
```
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use BatboldSamdan/babybata2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "BatboldSamdan/babybata2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BatboldSamdan/babybata2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/BatboldSamdan/babybata2

SGLang

How to use BatboldSamdan/babybata2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "BatboldSamdan/babybata2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BatboldSamdan/babybata2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "BatboldSamdan/babybata2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BatboldSamdan/babybata2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use BatboldSamdan/babybata2 with Docker Model Runner:
```
docker model run hf.co/BatboldSamdan/babybata2
```

babybata2 / README.md

BatboldSamdan

Upload README.md with huggingface_hub

c962b43 verified 25 days ago

preview code

Raw History Blame Contribute Delete

2.56 kB

metadata

license: apache-2.0
language:
  - en
base_model: Qwen/Qwen2.5-0.5B-Instruct
library_name: transformers
pipeline_tag: text-generation
tags:
  - lora
  - peft
  - chatbot
  - portfolio

BabyBata 2.0 🍼

A small assistant that answers questions about Batbold Samdan's portfolio: his projects, skills, education and contact info. An AI/ML learning project.

Base model: Qwen2.5-0.5B-Instruct
Method: LoRA fine-tuning (r=16, all attention + MLP projections, 1.7% of weights trained), merged into the base model, saved in float16
Data: ~630 generated conversations: topic Q&A, "does he know X?", "I don't know" cases, off-topic refusals, multi-turn follow-ups
Training: 3 epochs on a free Colab T4 (~10 min), loss on assistant replies only
Code: github.com/batbold9182/babybata

Important: it needs the profile

BabyBata is trained to answer from a fact sheet in the system prompt (grounding), not from memory. Use it with the same system prompt it was trained with: the rules + profile.txt from the repo (bot.py builds it).

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "BatboldSamdan/babybata2"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo)

system = open("babybata2/profile.txt").read()   # in practice: bot.system_prompt() from the repo
messages = [{"role": "system", "content": system},
            {"role": "user", "content": "what projects has he built?"}]
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt", return_dict=True)
out = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))

Evaluation

Custom fact check on questions not in the training data (incl. honesty checks like "does he know Rust?" → must say it doesn't know):

Model	Score
Qwen2.5-0.5B-Instruct + profile, no fine-tuning	18/27
BabyBata 2.0 v1	26/27 (4 test questions had leaked into training dialogues, so slightly optimistic)

Limitations

Only knows what's in the profile; it's a small model and can still be wrong. Verify important details on the portfolio site.
Known v1 issues: sometimes loses track in follow-up questions; once claimed a project had a live demo when it doesn't.
All training data is public information from the portfolio site.