This repo contains the code for "VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks" [ICLR25]
active 2024-10-07 → 2026-06-24 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 626 days from 2024-10-07 to 2026-06-24. Pushes: 194 total, peak 11 in a day. Pull requests: 26 total, peak 3 in a day. Issues: 233 total, peak 10 in a day. Comments: 360 total, peak 14 in a day. Stars: 380 total, peak 10 in a day.
- Pushes
- Pull requests
- Issues
- Comments
- Stars
Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.
Top contributors
Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| XMHZZ2018 | 238 | 88 | 2 | 97 |
| memray | 111 | 36 | 2 | 58 |
| wenhuchen | 62 | 49 | 0 | 10 |
| MINGYISU | 22 | 3 | 7 | 6 |
| ErnestZYJ | 15 | 0 | 0 | 15 |
| tianyuzong | 14 | 0 | 0 | 11 |
| magicgh | 13 | 9 | 4 | 0 |
| wswaq | 11 | 0 | 0 | 8 |
| VincentVanNF | 10 | 0 | 0 | 6 |
| Adenialzz | 10 | 0 | 0 | 6 |
| weiyao-Wang | 9 | 0 | 0 | 7 |
| lijiaoyang | 9 | 0 | 0 | 3 |
| B-201 | 9 | 0 | 0 | 3 |
| lcxrocks | 8 | 0 | 0 | 5 |
| xiankgx | 7 | 0 | 0 | 5 |
| PixarHero | 7 | 0 | 0 | 4 |
| haon-chen | 7 | 0 | 0 | 0 |
| SnowNation101 | 7 | 0 | 0 | 4 |
| Wuyiche | 7 | 0 | 0 | 6 |
| kimwongyuda | 7 | 0 | 0 | 3 |
Recent activity
Latest issues, pull requests and releases
- Issue#200sanshi95232026-04-03 02:18评估支持使用vllm框架部署的模型吗
- Pull request#197MINGYISU2026-03-23 19:38
- Issue comment#183Shengnan-Zhu2026-03-13 04:33why not directly use the original Qwen implementation on huggingface?
- Issue comment#194SkyFishMoon2026-03-12 21:31VD-Vidore-V2 evaluation metrics do not match the paper.
- Issue#193jianghuyihei2026-03-05 12:20How to support eval Qwen2.5/3-omni
- Issue comment#113Shengnan-Zhu2026-02-27 03:59Question about evaluation results and lora target modules
- Issue#167MINGYISU2026-02-09 18:54find a bug in ViDoSeek-page
- Issue#191KeepMovingXX2026-01-19 03:06Does this code support training baseline models (GME, LamRA) on private datasets?
- Issue comment#190sahel-sh2026-01-16 17:25add utils for converting data to jsonl format required for pyserini
- Issue#189simplew20112026-01-16 09:48how to eval a openapi model
- Issue comment#188sahel-sh2026-01-16 07:06esg_reports_human_labeled_v2: discrepancy between test corpus size and the number of uploaded images
- Issue#188sahel-sh2026-01-16 06:06esg_reports_human_labeled_v2: discrepancy between test corpus size and the number of uploaded images
- Pull request#186memray2026-01-12 03:10
- Pull request#186memray2026-01-12 03:10
- Issue#185XMHZZ20182026-01-12 03:00Question about how the Overall score is computed in ViDoRe v2 (report_score_v2.py)
- Issue#164memray2026-01-12 02:55Why rewrite _inner_training_loop method of MMEBTrainer?
- Issue comment#143memray2026-01-12 02:53datasets.table.CastError: Couldn't cast
- Issue#177XMHZZ20182026-01-12 02:30WandB training logs
- Issue comment#177XMHZZ20182026-01-12 02:30WandB training logs
- Issue#179XMHZZ20182026-01-12 02:20Clarification on Vidore Test Set: Are answer annotations used during training/evaluation?
- Issue#179XMHZZ20182026-01-12 02:20Clarification on Vidore Test Set: Are answer annotations used during training/evaluation?
- Issue comment#162memray2026-01-12 02:10Train with mmeb in image_only mode, raise error about picke
- Issue#172XMHZZ20182026-01-12 02:08For videos eval
- Issue#180XMHZZ20182026-01-12 02:06Can this model run on Huawei Ascend graphics hardware?
- Issue#162memray2026-01-12 02:05Train with mmeb in image_only mode, raise error about picke
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 380 stars here means stars gained during the window, not the repo's star count.