Code2Lora Data Snapshots: Repository Embeddings and Test Assertions

A collection of static snapshots for the Code2LoRA direct-projection trainer. The dataset includes one row per repository and anchor commit, with each row containing a 2048-dimensional Qwen3-Embedding vector representing the repository state and a single assertion extracted from a held-out test file. It was created by code2lora and last updated on 2026-05-25.

Use Cases

• Training or fine-tuning repository embedding models based on the provided 2048-dimension vectors.
• Evaluating the performance of code understanding models on held-out test assertions.
• Analyzing the relationship between repository state and specific code functionality based on the extracted QnA rows.

Strengths

• Each commit row includes a 2048-dimensional embedding vector generated by Qwen3-Embedding.
• Data is structured with separate splits for training, in-repository validation, and in-repository test snapshots.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Likely extracted from software repositories for the Code2LoRA trainer.

Freshness

Last updated 2026-05-25 00:08:52; freshness should be verified.