IdeaCredIdeaCred

iamtatsuki05/MIREI

79

MIREI is a research workspace that builds encoder/decoder text-embedding models under matched conditions, tracks shared training pipelines, and benchmarks their performance differences.

What's novel

MIREI is a research workspace that builds encoder/decoder text-embedding models under matched conditions, tracks shared training pipelines, and benchmarks their performance differences.

Code Analysis

8 files read · 3 rounds

A research workspace that converts decoder-only LLMs (Llama, Mistral, Qwen2) into bidirectional sentence-embedding encoders, with full training pipelines (CLM/MLM/MNTP pre-training, sentence-transformer fine-tuning), dataset cleaning/splitting, tokenizer training, and benchmarking infrastructure for

Strengths

Genuinely non-trivial implementation of bidirectional attention masking for three different LLM architectures, sophisticated document-boundary-aware sequence packing with FlashAttention varlen support, and a rich analytical metrics suite (CKA, effective rank, kNN overlap, attention entropy) that goes well beyond standard evaluation. The packing module in particular shows deep understanding of how different attention implementations handle packed sequences.

Weaknesses

Significant code duplication: the SentenceEncoderMixin (~200 lines) and bidirectional mask logic are copy-pasted verbatim across all three model files (Llama, Mistral, Qwen2) with comments acknowledging the duplication. The package name 'constract_llm' is a typo of 'construct_llm', and the test suite (28 files) appears thin relative to the 90 source files, suggesting limited edge-case coverage for the complex packing and attention mask logic.

Score Breakdown

Innovation
6 (25%)
Craft
65 (35%)
Traction
9 (15%)
Scope
88 (25%)

Signal breakdown

Innovation

Not Fork+1
Code Novelty+1
Concept Novelty+2

Craft

Ci+5
Tests+8
Polish+1
Releases+0
Has License+0
Code Quality+17
Readme Quality+12
Recent Activity+7
Structure Quality+5
Commit Consistency+5
Has Dependency Mgmt+5

Traction

Forks+0
Stars+6
Hn Points+0
Watchers+3
Early Traction+0
Devto Reactions+0
Community Contribs+0

Scope

Commits+8
Languages+8
Subsystems+13
Bloat Penalty+0
Completeness+6
Contributors+5
Authored Files+15
Readme Code Match+3
Architecture Depth+7
Implementation Depth+8

Evidence

Commits

140

Contributors

1

Files

215

Active weeks

9

TestsCI/CDREADMELicenseContributing

Repository

Language

Python

Stars

1

Forks

0

License