deepjavalibrary / deepjavalibrary/djl

High heap memory usage observed with DJL's HuggingFaceTokenizer implementation

Open
#3,604 2 comments 0 reactions 0 assignees View on GitHub
bug
Dominant language
Java
Stars
4.9k
Forks
759
Avg merge
19h 26m
Merged PRs (30d)
17

Description

## Current Behavior
- High heap memory consumption detected when using DJL's [HuggingFaceTokenizer](https://github.com/deepjavalibrary/djl/blob/master/extensions/tokenizers/src/main/java/ai/djl/huggingface/tokenizers/HuggingFaceTokenizer.java)'s tokenize method.
- Issue persists despite using the tokenizer as a singleton object.

## Impact
- Increased heap memory utilization affecting application performance.
- Potential memory management concerns in production environment.

## Questions
- Is the tokenize operation inherently memory intensive?
- Are there existing plans or recommendations to optimize memory usage?

TT: P198626352

Contributor guide

Open the contributing guide

Research direction

Start with extensions/tokenizers/src/main/java/ai/djl/huggingface/tokenizers/HuggingFaceTokenizer.java, focusing on the tokenize method named in the report. Reproduce the high heap usage and profile the operation, comparing singleton and repeated use. Done means identifying the memory source and documenting or implementing a confirmed optimization or recommendation.

Written by the indexing model from the issue text.

Assessment

Tech stack
huggingface, java
Domain
machine-learning, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.