TransMLA: Multi-Head Latent Attention Is All You Need
active 2025-03-05 → 2026-03-04 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 365 days from 2025-03-05 to 2026-03-04. Pushes: 32 total, peak 4 in a day. Pull requests: 0 total, peak 0 in a day. Issues: 24 total, peak 8 in a day. Comments: 30 total, peak 6 in a day. Stars: 100 total, peak 4 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 |
|---|---|---|---|---|
| Stanleytowne | 36 | 15 | 0 | 10 |
| fxmeng | 14 | 10 | 0 | 3 |
| BILLYisavailable | 5 | 5 | 0 | 0 |
| ruiqizhu-ricky | 4 | 0 | 0 | 2 |
| JSYRD | 4 | 0 | 0 | 4 |
| mingyangHao | 3 | 0 | 0 | 1 |
| ItGirls | 2 | 0 | 0 | 1 |
| wym42 | 2 | 0 | 0 | 1 |
| XiaojuanTang | 2 | 2 | 0 | 0 |
| nikita-yatchenko | 2 | 0 | 0 | 2 |
| xueliangyang-oeuler | 1 | 0 | 0 | 1 |
| huuyu | 1 | 0 | 0 | 1 |
| NielsRogge | 1 | 0 | 0 | 0 |
| EricLingRui | 1 | 0 | 0 | 1 |
| GiorgioZacharo | 1 | 0 | 0 | 1 |
| NonvolatileMemory | 1 | 0 | 0 | 1 |
| JiaxiangChen | 1 | 0 | 0 | 0 |
| libai-lab | 1 | 0 | 0 | 0 |
| zimei11 | 1 | 0 | 0 | 0 |
| Firefan | 1 | 0 | 0 | 0 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#38fxmeng2025-10-16 02:30Support Qwen3, e.g Qwen3-32B(GQA) can be tranasformed to MLA
- Issue comment#38xueliangyang-oeuler2025-10-16 02:12Support Qwen3, e.g Qwen3-32B(GQA) can be tranasformed to MLA
- Issue comment#37fxmeng2025-10-16 01:17Question: Applicability of TransMLA to DLM
- Issue comment#38fxmeng2025-10-16 01:04Support Qwen3, e.g Qwen3-32B(GQA) can be tranasformed to MLA
- Issue comment#38GiorgioZacharo2025-10-15 15:37Support Qwen3, e.g Qwen3-32B(GQA) can be tranasformed to MLA
- Issue#28Stanleytowne2025-09-23 11:58论文中kv cache压缩如何复现
- Issue comment#34Stanleytowne2025-09-23 11:57No performance difference with vLLM: qwen2.5-7b vs qwen2.5-7b-MLA
- Issue#29Stanleytowne2025-09-23 05:21SpeedUp Test and KV Cache is not actually Used?
- Issue comment#15Stanleytowne2025-09-23 05:16torchtune related code open source
- Issue#15Stanleytowne2025-09-23 05:16torchtune related code open source
- Issue#13Stanleytowne2025-09-23 05:16转换qwen-7b模型时，vllm推理报‘Cannot use FlashAttention-2 backend for head size 512.’错误
- Issue#12Stanleytowne2025-09-23 05:16Inference results are not consistent
- Issue#10Stanleytowne2025-09-23 05:10rank r
- Issue#5Stanleytowne2025-09-23 05:10vLLM/TensorRT
- Issue#2Stanleytowne2025-09-23 05:09rope compatible
- Issue comment#32Stanleytowne2025-09-23 05:07Can't run the converted model with vllm.LLM
- Issue comment#28Stanleytowne2025-09-23 05:06论文中kv cache压缩如何复现
- Issue comment#35Stanleytowne2025-09-23 04:49Error with running qwen-3b-instruct converted
- Issue comment#36Stanleytowne2025-09-23 04:35Question: Details of KV cache compression ratios and settings for all models
- Issue comment#34nikita-yatchenko2025-09-07 15:16No performance difference with vLLM: qwen2.5-7b vs qwen2.5-7b-MLA
- Issue comment#24nikita-yatchenko2025-09-04 15:38qwen2-7b and llama2-7b unable to deploy
- Issue comment#32ItGirls2025-09-04 04:06Can't run the converted model with vllm.LLM
- Issue comment#33JSYRD2025-09-04 01:32Qwen 转换后丢失中文能力
- Issue#33wym422025-09-03 03:47Qwen 转换后丢失中文能力
- Issue comment#32wym422025-09-03 03:07Can't run the converted model with vllm.LLM
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 100 stars here means stars gained during the window, not the repo's star count.