vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2025-09-06 04:30:03 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 2
698 commits 41 branches 19 tags 46 MiB
Commit graph

61 commits

Author SHA1 Message Date
djw
3f9bbf1181 support qwen3, dont speak human language 2025-04-28 08:44:47 +00:00
Atream
46493789eb
fix chat template encoding 2025-04-24 12:44:16 +08:00
qiyuxinlin
4f9950e30c kill serve lead to kill sched and engine 2025-04-22 09:25:44 +00:00
qiyuxinlin
03a65d6bea roll back ktransformers backend, add max_tokens, max_completion_tokens param 2025-04-21 12:55:37 +00:00
qiyuxinlin
38e841900d Move KV cache creation to balance_serve 2025-04-18 10:10:07 +00:00
sean.su
8699109129 Refactor the chat interface to support tool calling and parameter processing 
Defined new data structures in chat.py to replace OpenAI's original implementation, adding support for tool calling.

Implemented logic for extracting and processing tool calls, enabling dynamic function invocation during conversations.

Added methods in balance_serve.py to retrieve sampling parameters, handling default values and edge cases.

Updated ktransformers.py and transformers.py to support the passing of tool parameters.

Modified the default value of top_p in config.py to 1.0 to increase generation diversity.

Extended the message model in chat.py to support the transmission of tool call information.

These changes enhance the system's flexibility and functionality, enabling more complex interaction patterns.
 2025-04-14 15:23:37 +08:00
wangkuigang-yewu-cmss
4538bdae97 prevent rpc process from crashing on long prompt 
当prompt超过cache_len的时候，rpc进程会crash掉，导致整体不可用。
这里增加一个检查，让过长的prompt在请求早期就被提前过滤掉
 2025-04-13 16:13:16 +08:00
dongjw
ec03bcbd7f fix temperature=0, flashinfer sample error 2025-04-07 12:30:47 +08:00
dongjw
5c7ed7b579 fix top_p = 0 bug 2025-04-01 20:38:33 +08:00
Azure-Tang
31677181c3 Fix ktransformers-server flashinfer wrapper position arg issue; 
Fix db position issue
 2025-04-01 07:30:23 +00:00
Atream
25cee5810e add balance-serve, support concurrence 2025-03-31 22:55:32 +08:00
Atream
d453c320f1 fix flashinfer precision 2025-03-07 14:07:00 +00:00
BITcyman
299c4dca64 [update] support openai chat completion api 2025-03-07 08:51:09 +00:00
Azure
662c1e4c14 small fix about max new token 2025-03-05 09:25:41 +00:00
wang jiahao
48b9800790
Merge pull request #759 from 3wweiweiwu/fix_top_p_typo 
fix typo for top_p
 2025-03-02 13:58:11 +08:00
1668068727@qq.com
7cdf8139f0 fix ollama api temperature bug 2025-03-02 13:55:26 +08:00
Wix Woo
3aa0cfc29d fix typo for top_p 2025-03-01 20:15:36 +00:00
Atream
ca1dc1e7d1
Merge branch 'main' into main 2025-03-01 23:24:10 +08:00
Atream
fa03ea48dd Merge branch 'main' into feat-chunk-prefill-flashinfer 2025-03-01 11:35:09 +00:00
Atream
f35e8d41d8 support chunk prefill, support 139K context for 24G VRAM 2025-03-01 11:28:25 +00:00
liam
80e0536fb0 Merge branch 'main' of https://github.com/KMSorSMS/ktransformers into main 2025-03-01 00:12:21 +08:00
liam
8ddc990668 ⚡ fix server cache lens 2025-03-01 00:09:57 +08:00
qiyuxinlin
22df52e94e fix temperature 2025-02-27 21:00:44 +08:00
lazymio
b121ca4df8
Fix according to upstream changes 2025-02-27 18:11:35 +08:00
wang jiahao
26f7b4af11
Merge branch 'main' into temperature_top_p_from_request 2025-02-27 18:08:55 +08:00
Atream
b443c7dfa2
Merge pull request #657 from kvcache-ai/feat-absorb-for-long-prefill 
Feat absorb for long prefill
 2025-02-25 16:53:21 +08:00
Atream
f4c198bd42 support absorb for prefill long context 2025-02-25 08:52:02 +00:00
Azure
36fbeee341 Update doc 2025-02-25 08:21:18 +00:00
Azure
4dc5518e4d update fp8 kernel tutorial 2025-02-24 15:37:01 +00:00
lazymio
07eb712a73
Left out 2025-02-24 21:51:14 +08:00
lazymio
76487c4dcb
Revert repetition_penalty as it is not in API spec 2025-02-24 21:30:03 +08:00
lazymio
bf36547f98
Also allow repetition_penalty 2025-02-24 21:07:35 +08:00
lazymio
8704c09192
Allow temperature and top_p from requests 2025-02-24 21:01:33 +08:00
Atream
024009675e Merge branch 'main' into feat-more-context 2025-02-22 06:17:39 +00:00
Atream
7e1fe256c8 optimize GPU 2025-02-21 05:06:57 +00:00
Atream
a529518346 clean PR code and disable flashinfer 2025-02-19 04:42:47 +00:00
ceerrep
73d072f609 Merge branch 'fix_precision_MLA' of https://github.com/kvcache-ai/ktransformers into server-prefix-cache 2025-02-18 11:44:28 +08:00
Xie Weiyu
f029588b61 fix server warmup 2025-02-18 11:39:45 +08:00
ceerrep
c70b6f4d5b fix: use 'cuda:0' by default if torch_device is 'cuda' 2025-02-18 11:15:17 +08:00
Xie Weiyu
c176e516b5 server mix mla 2025-02-17 20:40:28 +08:00
ceerrep
ee24eb8dc3 fix: fix server for triton kernel 2025-02-17 18:08:45 +08:00
ceerrep
cd9f7f8f34 fix: server: drop <think> tag in chat template 2025-02-17 14:25:27 +08:00
ceerrep
bb0ccc7b1a feat: add prefix cache for server 2025-02-17 00:10:55 +08:00
MuWinds
f74c2d1d17
Solve torch.backends.cuda.sdp_kernel() is deprecated. 2025-02-15 12:41:51 +08:00
hrz6976
2c3dcd9774 Add a lock to server inference() 2025-02-13 10:05:22 +00:00
liam
4385e85096 ⚡ support force thinking 2025-02-12 12:43:53 +08:00
liam
6f3a39be08 ⚡ update force_think config 2025-02-12 12:10:16 +08:00
liam
e536e1420d ⚡ update force_think 2025-02-12 11:42:55 +08:00
liam
bf1d413be0 Merge branch 'feat-DeepSeekV3' of github.com:kvcache-ai/ktransformers into feat-DeepSeekV3 2025-02-08 13:17:10 +08:00
liam
c18ecd7b7f ⚡ add flush print in local_chat output and change default optimize yaml of deepseekv3 to single gpu 2025-02-08 13:15:52 +08:00