pipeline_tag: text-generation
inference: true
widget:
- text: |-
public class HelloWorld {
public static void main(String[] args) {
example_title: Hello world
group: Java
license: bigcode-openrail-m
datasets:
- bigcode/starcoderdata
metrics:
- code_eval
library_name: transformers
tags:
- code
model-index:
- name: NT-Java-1.1B
results:
- task:
type: text-generation
dataset:
type: nuprl/MultiPL-E
name: MultiPL-HumanEval (Java)
metrics:
- name: pass@1
type: pass@1
value: 18.3
verified: false
extra_gated_prompt: >-
## Model License Agreement
Please read the BigCode [OpenRAIL-M
license](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement)
agreement before accepting it.
extra_gated_fields:
I accept the above license agreement, and will use the Model complying with the set of use restrictions and sharing requirements: checkbox
duplicated_from: bigcode-data/starcoderbase-1b
NT-Java-1.1B
Table of Contents
Model Summary
The Narrow Transformer (NT) model NT-Java-1.1B is an open-source specialized code model built by extending pre-training on StarCoderBase-1B, designed for coding tasks in Java programming. The model is a decoder-only transformer with Multi-Query Attention and with a context length of 8192 tokens. The model was trained with Java subset of the StarCoderData dataset, which is ~22B tokens.
- Repository: bigcode/Megatron-LM
- Paper:
- Language(s): Java
Use
Intended use
Large code models require specialized hardware like GPUs for inference, highlighting the need for research into building small code models that can be deployed on developer desktops. This model addresses the gap by focusing on the development of a small Java code model and introducing a quantized version of NT-Java-1.1B, which performs comparably to open 1.1B models on MultiPL-E Java code benchmarks, making it ideal for desktop deployment.
Feel free to share your generations in the Community tab!
Generation
# pip install -q transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
checkpoint = "infosys/NT-Java-1.1B"
device = "cuda" # for GPU usage or "cpu" for CPU usage
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
inputs = tokenizer.encode("public class HelloWorld {\n public static void main(String[] args) {", return_tensors="pt").to(device)
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))
Quantized Versions through bitsandbytes
- Using 8-bit precision (int8)
# pip install bitsandbytes accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
# to use 4bit use `load_in_4bit=True` instead
quantization_config = BitsAndBytesConfig(load_in_8bit=True)
checkpoint = "infosys/NT-Java-1.1B"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint, quantization_config=quantization_config)
inputs = tokenizer.encode("public class HelloWorld {\n public static void main(String[] args) {", return_tensors="pt").to("cuda")
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))
Attribution & Other Requirements
The pretraining dataset of the model was filtered for permissive licenses only. Nevertheless, the model can generate source code verbatim from the dataset. The code's license might require attribution and/or other specific requirements that must be respected. We provide a search index that let's you search through the pretraining data to identify where generated code came from and apply the proper attribution to your code.
Benefits
Large code models require specialized hardware like GPUs for inference, highlighting the need for research into building small code models that can be deployed on developer desktops. This model addresses the gap by focusing on the development of a small Java code model and introducing a quantized version (in different forms like GGML, GGUF) of NT-Java-1.1B, which performs comparably to open 1.1B models on MultiPL-E Java code benchmarks, making it ideal for desktop deployment.
Limitations
The model, NT-Java-1.1B, has been trained on publicly available datasets and comes without any safety guarantees. Due to this, like all Language Models, its outputs cannot be reliably predicted and sometimes the generated code is not guaranteed to work as intended. It can also be inefficient and may contain bugs or exploits. Therefore, it's crucial for users and developers to conduct thorough safety testing and implement filtering mechanisms tailored to their needs.
Training
Model
- Architecture: GPT-2 model with Multi-Query Attention and Fill-in-the-Middle objective
- •Fine-training steps: 50k
- Pretraining tokens: 22 Billion
- Precision: bfloat16
Hardware
- GPUs: 6 NVIDIA A100 80GB
- Training time: 4 days
Software
- Orchestration: Megatron-LM
- Neural networks: PyTorch
License
The model checkpoint and vocabulary file are licensed under the BigCode OpenRAIL-M v1 . Under the license, you must evaluate if your use case does not violate the use-case restriction under Attachment A of the License. Any modification of the model (finetuning or extended pre training) for further downstream task needs to be released under BigCode OpenRAIL-M v1.
Citation
@article{li2023starcoder,
title={NARROW TRANSFORMER: STARCODER-BASED JAVA-LM FOR DESKTOP},
author={Kamalkumar Rathinasamy and Balaji A J and Rajab Ali Mondal and Ankush Kumar and Harshini K and Gagan Gayari and Sreenivasa Raghavan Karumboor Seshadri},
year={2024},
eprint={2305.06161},
archivePrefix={arXiv},
primaryClass={cs.CL}
}