Instructions to use kz919/llama3_3b_cautious_chinchilla_8152025 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kz919/llama3_3b_cautious_chinchilla_8152025 with Transformers:

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

pipe = pipeline("text-generation", model="kz919/llama3_3b_cautious_chinchilla_8152025")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("kz919/llama3_3b_cautious_chinchilla_8152025")
model = AutoModelForCausalLM.from_pretrained("kz919/llama3_3b_cautious_chinchilla_8152025")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use kz919/llama3_3b_cautious_chinchilla_8152025 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kz919/llama3_3b_cautious_chinchilla_8152025"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kz919/llama3_3b_cautious_chinchilla_8152025",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/kz919/llama3_3b_cautious_chinchilla_8152025

SGLang

How to use kz919/llama3_3b_cautious_chinchilla_8152025 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 "kz919/llama3_3b_cautious_chinchilla_8152025" \
    --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": "kz919/llama3_3b_cautious_chinchilla_8152025",
		"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 "kz919/llama3_3b_cautious_chinchilla_8152025" \
        --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": "kz919/llama3_3b_cautious_chinchilla_8152025",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use kz919/llama3_3b_cautious_chinchilla_8152025 with Docker Model Runner:
```
docker model run hf.co/kz919/llama3_3b_cautious_chinchilla_8152025
```

Model Details

This model is a 3B llama3 model pretrained from scratch with torchtitan on fineweb-edu with C_AdamW optimizer. 20x chinchilla rule for 60B tokens seen.

How to use

import torch
from transformers import pipeline


pipe = pipeline(
    "text-generation",
    model="kz919/llama3_3b_chinchilla_8142025",
)

print(pipe("The key to life is"))

Downstream Eval

ARC, Hellaswag, Lambda_OpenAI, OpenbookQA, PIQA

lm_eval --model hf --model_args pretrained=kz919/llama3_3b_chinchilla_8142025,dtype="bfloat16",add_bos_token=True --tasks lambada_openai,hellaswag,piqa,arc_easy,arc_challenge,openbookqa --device cuda:7 --batch_size 8

Tasks	Version	Filter	Metric		Value		Stderr
arc_challenge	1	none	acc	↑	0.2892	±	0.0133
		none	acc_norm	↑	0.2892	±	0.0133
arc_easy	1	none	acc	↑	0.6162	±	0.0100
		none	acc_norm	↑	0.5311	±	0.0102
hellaswag	1	none	acc	↑	0.3698	±	0.0048
		none	acc_norm	↑	0.4611	±	0.0050
lambada_openai	1	none	acc	↑	0.3670	±	0.0067
		none	perplexity	↓	34.2265	±	1.4167
openbookqa	1	none	acc	↑	0.2380	±	0.0191
		none	acc_norm	↑	0.3460	±	0.0213
piqa	1	none	acc	↑	0.6904	±	0.0108
		none	acc_norm	↑	0.6975	±	0.0107

MMLU

Groups	Version	Filter	Metric		Value		Stderr
mmlu	2	none	acc	↑	0.2453	±	0.0036
- humanities	2	none	acc	↑	0.2502	±	0.0063
- other	2	none	acc	↑	0.2620	±	0.0079
- social sciences	2	none	acc	↑	0.2320	±	0.0076
- stem	2	none	acc	↑	0.2347	±	0.0076

Downloads last month: 1

Safetensors

Model size

4B params

Tensor type

F32

kz919
/

llama3_3b_cautious_chinchilla_8152025

Model Details

How to use

Downstream Eval

ARC, Hellaswag, Lambda_OpenAI, OpenbookQA, PIQA

MMLU

Dataset used to train kz919/llama3_3b_cautious_chinchilla_8152025