体验尖端 LLM 推理优化的灵活框架
Python · active 2025-04-02 → 2026-08-08 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 494 days from 2025-04-02 to 2026-08-08. Pushes: 411 total, peak 12 in a day. Pull requests: 253 total, peak 14 in a day. Issues: 601 total, peak 252 in a day. Comments: 1,440 total, peak 132 in a day. Stars: 1,733 total, peak 30 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 |
|---|---|---|---|---|
| ErvinXie | 469 | 35 | 11 | 161 |
| gemini-code-assist[bot] | 353 | 0 | 0 | 280 |
| KMSorSMS | 141 | 21 | 26 | 49 |
| ovowei | 121 | 67 | 27 | 16 |
| Atream | 119 | 30 | 37 | 44 |
| qiyuxinlin | 98 | 18 | 24 | 45 |
| magikRUKKOLA | 82 | 0 | 0 | 81 |
| SkqLiao | 79 | 44 | 23 | 11 |
| JimmyPeilinLi | 77 | 44 | 21 | 6 |
| github-actions[bot] | 58 | 57 | 1 | 0 |
| createthis | 50 | 0 | 1 | 44 |
| ouqingliang | 48 | 34 | 10 | 4 |
| Azure-Tang | 33 | 4 | 3 | 19 |
| 344303947 | 33 | 0 | 0 | 25 |
| yyj6666667 | 33 | 16 | 6 | 7 |
| aubreyli | 32 | 4 | 7 | 16 |
| mrhaoxx | 23 | 18 | 5 | 0 |
| mrgaolei | 22 | 0 | 0 | 9 |
| trilog-inc | 19 | 0 | 0 | 13 |
| chenht2022 | 18 | 4 | 4 | 6 |
Recent activity
Latest issues, pull requests and releases
- Pull request#2071VectorPeak2026-07-06 07:43
- Issue#2045yyj66666672026-06-18 03:10apache-tvm-ffi 0.1.12 has a conflict with tilelang
- Pull request#2048Anai-Guo2026-06-18 03:10
- Pull request#2044Dayuxiaoshui2026-06-16 02:57
- Issue comment#1999harrychk2026-06-12 15:54[Help] Possible to run deepseek4-flash on ampere (3090) + threadripper 3970x (no AMX/AVX512) ?
- Issue comment#2023yyj66666672026-06-06 14:39完全使用hybird prefill时只加载冷门专家至内存，节约内存使用量
- Pull request#2031jdai02026-06-02 06:17
- Issue comment#1978goldengolden79812026-06-02 00:55[Bug] deepseek-v4-flash with sm80 ValueError("type fp8e4nv not supported in this architecture. The supported fp8 dtypes are ('fp8e4b15', 'fp8e5')")
- Issue comment#1978linchpinlin2026-06-01 01:07[Bug] deepseek-v4-flash with sm80 ValueError("type fp8e4nv not supported in this architecture. The supported fp8 dtypes are ('fp8e4b15', 'fp8e5')")
- Issue comment#1608KMSorSMS2026-05-29 00:56🚨FAQs | 常见问题🚨
- Pull request#2021hermannklie2026-05-22 10:09
- Issue comment#2018yyj66666672026-05-22 05:07[fix] Add runtime AMX BF16 check
- Issue comment#1969shrould88882026-05-17 14:29v0.6.1 fails to launch DeepSeek v3.2 (and other models) with CUDA out of memory
- Issue comment#1969vladie2026-05-17 11:42v0.6.1 fails to launch DeepSeek v3.2 (and other models) with CUDA out of memory
- Issue comment#1969shrould88882026-05-16 15:19v0.6.1 fails to launch DeepSeek v3.2 (and other models) with CUDA out of memory
- Issue comment#1829shrould88882026-05-16 09:29how to convert Kimi-K2.5 base model is in INT4 format to BF16 before SFT
- Issue comment#2007usrlocalben2026-05-16 03:42Speculative/EAGLE - Supported?
- Issue#2008usrlocalben2026-05-16 01:37Ambiguous startup warning wrt. sglang / kt-kernel / moe integration
- Issue comment#2001yyj66666672026-05-14 14:31DeepSeek-V4-Flash outputs all-zero tokens on single RTX PRO 4000 Blackwell (SM_120, 24GB)
- Issue#2001lizhucui02-arch2026-05-13 04:37DeepSeek-V4-Flash outputs all-zero tokens on single RTX PRO 4000 Blackwell (SM_120, 24GB)
- Issue comment#1985yyj66666672026-05-11 08:37Common Problem / Fix: Running deepseek v4 on L40 *4, AssertionError occurs in swa_radix_cache.py:565
- Issue comment#1969vladie2026-05-11 07:07v0.6.1 fails to launch DeepSeek v3.2 (and other models) with CUDA out of memory
- Issue comment#1992yyj66666672026-05-11 06:35Minimax2.5 TypeError: non-default argument 'quantization_config' follows default argument
- Issue comment#1988lovelyBugs2026-05-11 02:595090D 32G single card encounter error assert num_stages >= 1 on first request
- Issue comment#1988lhw3652026-05-09 20:325090D 32G single card encounter error assert num_stages >= 1 on first request
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 1,733 stars here means stars gained during the window, not the repo's star count.