Files
calculet-npu-research-archive/history/topics/07-llm-prefill-decode.tsv
T

21 lines
3.1 KiB
Plaintext

904993432e59dbe1ba6e2060cac875117e9e18b9 4e049fb678264751ec60712f9ef98cca347526b8 2026-05-24T15:54:52+08:00 patrickyang update llm hparam
b12ca8edd20180a3b88acbab21a897a37bcd54d6 a6cf81326bb629c4171d4c418e923b57c9a7ebdb 2026-05-12T20:41:09+08:00 wantaobo upt python cnn and llm model runner
a6cf81326bb629c4171d4c418e923b57c9a7ebdb 6e5ff4dd84adecc61c8c8dc735b1a67c41f5f025 2026-05-12T19:32:51+08:00 wantaobo upt llm hparam.py
b305139fdbb35aef4808522799edb69f8251e0f2 f6d8084b98f5a1566ea62b7745ec844c0fa8abea 2026-05-12T11:49:53+08:00 wantaobo upt llm hparam
59c305237c0acf014343759c3718442e48a138e0 6cef0a060d9e606e8659a2a78087dc5e0bf68efd 2026-05-12T09:46:43+08:00 wantaobo upt llm_hparam
871d477d202a6d1038848649de134dd20b69ca2e 0b549165d875ef3274e07b8a83ac87d1cd1add5e 2026-04-14T21:41:57+08:00 wantaobo upt model runner for llm_dynamic
9afe49bb208a5b0d155a5b9a5308abeb52f82c8b 30748706555cf7d9450554ae8df1c04590399fb2 2026-04-14T20:30:30+08:00 wantaobo upt llm_dynamic
59e91eb3907a13e4d7b59722dd3c2eb2f9f54fdb 3f6eec29def96796f3ecf3e4106481d438a70afb 2026-04-07T17:12:32+08:00 wantaobo upt model runner for llm_dynamic
862e41d1e589b079afcfcd355dcfe754c0935af5 f61c26c3511600eaa0ff8f766e72c374162a6752 2026-03-04T17:55:30+08:00 wantaobo upt model runner fix bug: llm hparam, input data from start not from posid
ef98c26aecda65a8958c5e23167526483bfe75a9 e9f95f4f06bf0742eda6649a322f327c2dfa5b38 2026-02-06T17:28:53+08:00 wantaobo upt llmtest
6a0c84728f933ca0b2323e4baa8d7c957716a911 5df50e9aa1549960b4d53d71042e4d3e2066176f 2026-02-05T11:05:14+08:00 ylsun add N_MAX_BATCH macro
d73d0ac515f2ab3bb3cf7020d4de8185b46f73be ab558e259c68679451635899066d05b7ef8310b9 2026-01-05T11:49:08+08:00 patrickyang update llmtest model runner
ab558e259c68679451635899066d05b7ef8310b9 5a4dc80636bb3846533e7d3e063f73e0eeb5d330 2026-01-05T11:35:38+08:00 patrickyang rm duplicated namespace. add exception handling for llmtest
d975419f1d9ef1b99d42879074fb5e1580a937de 78934176cb86aaa93fe876b6ef7c5cb5c679074f 2025-12-12T15:06:41+08:00 wantaobo [model runner] upt decode name search
3fe116ddce59f35ee6d0dbb04324fe54d38eb6d9 0af5d1a87a41b7c1a7faeba56710581fdf6614eb 2025-12-12T15:03:31+08:00 wantaobo [model runner] upt prefill and decode hparam feature
962167ff709da8f2fd8308d54681d8a940d17966 34b73c812458a9e379daeb9acb4d049ca1dc918e 2025-12-11T09:42:08+08:00 wantaobo upt llmtest model runner
b117576b511b674a5fa0edbf9ce78a23cb586efd 4d44acc3df6ae93e3daaef968bfd8e3c7428553c 2025-12-04T15:04:40+08:00 wantaobo [model runner] 1. udpate llm test; 2. udpate calxtensor
23752d5d997d2a78c9f43cda35d13aebceee1b70 c885a0e98c6cc19aa5adefd772b638d0030f4e63 2025-11-27T12:06:24+08:00 wantaobo update decode model runner, support hparam txt
90747c80fe69a6cffe5ddeca4a70b6538f3e1594 0dc2fe92d7ce08325a2f83a9889d54919bd14472 2025-11-12T18:02:56+08:00 wantaobo update decode
11ee8ad7fa9869d12a3e7c2b29f0c07dc3bed4cd 95de7ba869bc455002b14a227cd25c279dd49acc 2025-10-30T17:14:38+08:00 wantaobo upt prefill and decode model runner
62478f7825eafe925baa1b7e4b1d854151399f70 2d2380dd8f878444d01e735f5f2df4f4054c1be8 2025-08-04T17:49:59+08:00 patrickyang update calbin Class. Now it is able to provide LLM seperately