Instructions to use KiteFishAI/Nano-Math-1.5Bv2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use KiteFishAI/Nano-Math-1.5Bv2 with Transformers:

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

pipe = pipeline("text-generation", model="KiteFishAI/Nano-Math-1.5Bv2")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("KiteFishAI/Nano-Math-1.5Bv2")
model = AutoModelForCausalLM.from_pretrained("KiteFishAI/Nano-Math-1.5Bv2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use KiteFishAI/Nano-Math-1.5Bv2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "KiteFishAI/Nano-Math-1.5Bv2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KiteFishAI/Nano-Math-1.5Bv2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/KiteFishAI/Nano-Math-1.5Bv2

SGLang

How to use KiteFishAI/Nano-Math-1.5Bv2 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 "KiteFishAI/Nano-Math-1.5Bv2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KiteFishAI/Nano-Math-1.5Bv2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "KiteFishAI/Nano-Math-1.5Bv2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KiteFishAI/Nano-Math-1.5Bv2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use KiteFishAI/Nano-Math-1.5Bv2 with Docker Model Runner:
```
docker model run hf.co/KiteFishAI/Nano-Math-1.5Bv2
```

anuj0456 commited on Feb 20

Commit

c0b74e4

verified ·

1 Parent(s): 7c070a4

Update README.md

Browse files

Files changed (1) hide show

README.md +0 -12

README.md CHANGED Viewed

@@ -19,8 +19,6 @@ library_name: transformers
 This is a **base scientific language model** (not instruction-tuned).
----
 ## Overview
 KiteFish-A1-1.5B explores what it takes to train a domain-specialized scientific language model directly from structured LaTeX archives.
@@ -35,8 +33,6 @@ KiteFish-A1-1.5B explores what it takes to train a domain-specialized scientific
 The focus of this project is *scientific language modeling robustness*, not benchmark optimization.
----
 ## Model Architecture
 - 24 Transformer layers
@@ -57,8 +53,6 @@ The focus of this project is *scientific language modeling robustness*, not benc
 **Validation Perplexity:** ~4.2 (held-out scientific corpus)
----
 ## Intended Use
 KiteFish-A1-1.5B is suitable for:
@@ -76,8 +70,6 @@ It is **not optimized for:**
 - General conversational AI
 - Benchmark leaderboard performance
----
 ## Performance Notes
 This model was trained under moderate compute constraints and without instruction tuning or alignment stages.
@@ -92,8 +84,6 @@ Observed characteristics:
 Performance improves significantly with supervised fine-tuning (SFT), LoRA adaptation, or domain-specific instruction tuning.
----
 ## Limitations
 - Not instruction-tuned
@@ -105,8 +95,6 @@ Performance improves significantly with supervised fine-tuning (SFT), LoRA adapt
 This release is intended primarily for research and experimentation.
----
 ## Example Usage
 ```python

 This is a **base scientific language model** (not instruction-tuned).
 ## Overview
 KiteFish-A1-1.5B explores what it takes to train a domain-specialized scientific language model directly from structured LaTeX archives.
 The focus of this project is *scientific language modeling robustness*, not benchmark optimization.
 ## Model Architecture
 - 24 Transformer layers
 **Validation Perplexity:** ~4.2 (held-out scientific corpus)
 ## Intended Use
 KiteFish-A1-1.5B is suitable for:
 - General conversational AI
 - Benchmark leaderboard performance
 ## Performance Notes
 This model was trained under moderate compute constraints and without instruction tuning or alignment stages.
 Performance improves significantly with supervised fine-tuning (SFT), LoRA adaptation, or domain-specific instruction tuning.
 ## Limitations
 - Not instruction-tuned
 This release is intended primarily for research and experimentation.
 ## Example Usage
 ```python