codet5p-110m-embedding

by Salesforce

Unified code understanding and generation with T5 architecture

102Kdl/month

68likes

110Mparams

HuggingFace Use in Pipeline

Identifiers

Model ID

Salesforce/codet5p-110m-embedding

Feature URI

mixpeek://document_extractor@v1/salesforce_codet5p_v1

Overview

CodeT5+ is a family of encoder-decoder code LLMs that support both understanding and generation tasks. The 110M embedding variant is optimized for producing high-quality code embeddings for retrieval.

On Mixpeek, CodeT5+ provides an alternative to CodeBERT for code embedding extraction, with support for more programming languages and stronger performance on code search tasks.

Architecture

T5-based encoder-decoder. The 110M embedding variant uses only the encoder, trained with contrastive learning on code-text pairs. Supports 10+ programming languages.

Mixpeek SDK Integration

import { Mixpeek } from "mixpeek";

const mx = new Mixpeek({ apiKey: "API_KEY" });

await mx.collections.ingest({
  collection_id: "my-collection",
  source: { url: "https://example.com/codebase.zip" },
  feature_extractors: [{
    name: "code_extraction",
    version: "v1",
    params: {
      model_id: "Salesforce/codet5p-110m-embedding"
    }
  }]
});