jeongin601
1bf905ddaa
[Bugfix][SpecDecode] apply sampling parameters to target probabilities for consistency in rejection sampling. ( #10198 )
...
Signed-off-by: jeongin601 <0200angela@gmail.com>
Signed-off-by: jeong_in.bae <jeong_in.bae@navercorp.com>
2024-11-27 05:07:30 +00:00
Chendi.Xue
0a71900bc9
Remove hard-dependencies of Speculative decode to CUDA workers ( #10587 )
...
Signed-off-by: Chendi Xue <chendi.xue@intel.com>
2024-11-26 17:57:11 -08:00
Murali Andoorveedu
db66e018ea
[Bugfix] Fix for Spec model TP + Chunked Prefill ( #10232 )
...
Signed-off-by: andoorve <37849411+andoorve@users.noreply.github.com>
Signed-off-by: Sourashis Roy <sroy@roblox.com>
Co-authored-by: Sourashis Roy <sroy@roblox.com>
2024-11-26 09:11:16 -08:00
sroy745
f6778620a9
Disable spec-decode + chunked-prefill for draft models with tensor parallelism > 1 ( #10136 )
...
Signed-off-by: Sourashis Roy <sroy@roblox.com>
2024-11-08 15:56:18 +00:00
Nicolò Lucchesi
9d43afcc53
[Feature] [Spec decode]: Combine chunked prefill with speculative decoding ( #9291 )
...
Signed-off-by: NickLucche <nlucches@redhat.com>
2024-11-07 08:15:14 -08:00
Sungjae Lee
0c63c34f72
[Bugfix][SpecDecode] kv corruption with bonus tokens in spec decode ( #9730 )
...
Co-authored-by: LiuXiaoxuanPKU <lilyliupku@gmail.com>
2024-11-06 01:45:45 +00:00
youkaichao
e893795443
[2/N] executor pass the complete config to worker/modelrunner ( #9938 )
...
Signed-off-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
2024-11-02 07:35:05 -07:00
wangshuai09
4e2d95e372
[Hardware][ROCM] using current_platform.is_rocm ( #9642 )
...
Signed-off-by: wangshuai09 <391746016@qq.com>
2024-10-28 04:07:00 +00:00
Cody Yu
d11bf435a0
[MISC] Consolidate cleanup() and refactor offline_inference_with_prefix.py ( #9510 )
2024-10-18 14:30:55 -07:00
Kuntai Du
81ede99ca4
[Core] Deprecating block manager v1 and make block manager v2 default ( #8704 )
...
Removing the block manager v1. This is the initial piece of prefix-caching-centric design. In order to achieve prefix-caching-centric design, we need to simplify the code path so that we only use v2 block manager (which has much higher performance on prefix caching).
2024-10-17 11:38:15 -05:00
Lily Liu
89feb4c84d
[SpecDec] Remove Batch Expansion (2/3) ( #9298 )
2024-10-12 05:13:37 +00:00
sroy745
f3a507f1d3
[Core] Add an environment variable which needs to be set explicitly to allow BlockSpaceManagerV1 ( #9149 )
2024-10-10 14:17:17 +08:00
sroy745
91add85ec4
Fix failing spec decode test ( #9054 )
2024-10-03 23:07:29 +00:00
Lily Liu
1570203864
[Spec Decode] (1/2) Remove batch expansion ( #8839 )
2024-10-01 16:04:42 -07:00
Lily Liu
bce324487a
[CI][SpecDecode] Fix spec decode tests, use flash attention backend for spec decode CI tests. ( #8975 )
2024-10-01 00:51:40 +00:00
Travis Johnson
01b6f9e1f0
[Core][Bugfix] Support prompt_logprobs returned with speculative decoding ( #8047 )
...
Signed-off-by: Travis Johnson <tsjohnso@us.ibm.com>
2024-09-24 17:29:56 -07:00
Lily Liu
c6bd70d772
[SpecDec][Misc] Cleanup, remove bonus token logic. ( #8701 )
2024-09-22 12:34:14 -07:00
Cyrus Leung
0455c46ed4
[Core] Factor out common code in SequenceData and Sequence ( #8675 )
2024-09-21 02:30:39 +00:00
Lily Liu
775f00f81e
[Speculative Decoding] Test refactor ( #8317 )
...
Co-authored-by: youkaichao <youkaichao@126.com>
2024-09-11 14:07:34 -07:00
Lily Liu
e6a26ed037
[SpecDecode][Kernel] Flashinfer Rejection Sampling ( #7244 )
2024-09-01 21:23:29 -07:00
afeldman-nm
428dd1445e
[Core] Logprobs support in Multi-step ( #7652 )
2024-08-29 19:19:08 -07:00
Jonas M. Kübler
f205c09854
[Bugfix] Unify rank computation across regular decoding and speculative decoding ( #7899 )
2024-08-28 22:18:13 -07:00
Nick Hill
1856aff4d6
[Spec Decoding] Streamline batch expansion tensor manipulation ( #7851 )
2024-08-25 15:45:14 -07:00
Travis Johnson
cc0eaf12b1
[Bugfix] spec decode handle None entries in topk args in create_sequence_group_output ( #7232 )
...
Signed-off-by: Travis Johnson <tsjohnso@us.ibm.com>
2024-08-22 09:33:48 -04:00
Abhinav Goyal
a3fce56b88
[Speculative Decoding] EAGLE Implementation with Top-1 proposer ( #6830 )
2024-08-22 02:42:24 -07:00
Abhinav Goyal
312f761232
[Speculative Decoding] Fixing hidden states handling in batch expansion ( #7508 )
2024-08-19 17:58:14 -07:00
SangBin Cho
ff7ec82c4d
[Core] Optimize SPMD architecture with delta + serialization optimization ( #7109 )
2024-08-18 17:57:20 -07:00
jon-chuang
50b8d08dbd
[Misc/Testing] Use torch.testing.assert_close ( #7324 )
2024-08-16 04:24:04 +00:00
shangmingc
b67ae00cdb
[Misc] Add quantization config support for speculative model. ( #7343 )
2024-08-15 19:34:28 -07:00
Wallas Henrique
70b746efcf
[Misc] Deprecation Warning when setting --engine-use-ray ( #7424 )
...
Signed-off-by: Wallas Santos <wallashss@ibm.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: Nick Hill <nickhill@us.ibm.com>
Co-authored-by: youkaichao <youkaichao@126.com>
2024-08-14 09:44:27 -07:00
Travis Johnson
99b4cf5f23
[Bugfix] Fix speculative decoding with MLPSpeculator with padded vocabulary ( #7218 )
...
Signed-off-by: Travis Johnson <tsjohnso@us.ibm.com>
2024-08-08 22:08:46 -07:00
Cade Daniel
82a1b1a82b
[Speculative decoding] Add periodic log with time spent in proposal/scoring/verification ( #6963 )
2024-08-05 08:46:44 +00:00
Nick Hill
5cf9254a9c
[BugFix] Fix use of per-request seed with pipeline parallel ( #6698 )
2024-07-30 10:40:08 -07:00
Nick Hill
2cf0df3381
[Bugfix] Fix speculative decode seeded test ( #6743 )
2024-07-24 08:58:31 -07:00
Nick Hill
c882a7f5b3
[SpecDecoding] Update MLPSpeculator CI tests to use smaller model ( #6714 )
2024-07-24 07:34:22 +00:00
sroy745
14f91fe67c
[Spec Decode] Disable Log Prob serialization to CPU for spec decoding for both draft and target models. ( #6485 )
2024-07-20 23:58:58 -07:00
Thomas Parnell
f0bbfaf917
[Bugfix] [SpecDecode] AsyncMetricsCollector: update time since last collection ( #6578 )
2024-07-19 14:01:03 -07:00
Thomas Parnell
a5314e8698
[Model] RowParallelLinear: pass bias to quant_method.apply ( #6327 )
...
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
2024-07-19 07:15:22 -06:00
Woo-Yeon Lee
a921e86392
[BUGFIX] Raise an error for no draft token case when draft_tp>1 ( #6369 )
2024-07-19 06:01:09 -07:00
Thomas Parnell
d4201e06d5
[Bugfix] Make spec. decode respect per-request seed. ( #6034 )
...
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
Co-authored-by: Nick Hill <nickhill@us.ibm.com>
2024-07-18 19:22:08 -07:00
Alexander Matveev
e76466dde2
[Core] draft_model_runner: Implement prepare_inputs on GPU for advance_step ( #6338 )
2024-07-17 14:30:28 -07:00
Cody Yu
160e1d8c99
[Misc] Log spec decode metrics ( #6454 )
2024-07-16 20:37:10 +00:00
sroy745
ae151d73be
[Speculative Decoding] Enabling bonus token in speculative decoding for KV cache based models ( #5765 )
2024-07-10 16:02:47 -07:00
Abhinav Goyal
2416b26e11
[Speculative Decoding] Medusa Implementation with Top-1 proposer ( #4978 )
2024-07-09 18:34:02 -07:00
Swapnil Parekh
4d6ada947c
[CORE] Adding support for insertion of soft-tuned prompts ( #4645 )
...
Co-authored-by: Swapnil Parekh <swapnilp@ibm.com>
Co-authored-by: Joe G <joseph.granados@h2o.ai>
Co-authored-by: Antoni Baum <antoni.baum@protonmail.com>
2024-07-09 13:26:36 -07:00
Qubitium-ModelCloud
ee93f4f92a
[CORE] Quantized lm-head Framework ( #4442 )
...
Co-authored-by: Robert Shaw <rshaw@neuralmagic.com>
Co-authored-by: ZX <zx@lbx.dev>
2024-07-02 22:25:17 +00:00
Murali Andoorveedu
c5832d2ae9
[Core] Pipeline Parallel Support ( #4412 )
...
Signed-off-by: Muralidhar Andoorveedu <muralidhar.andoorveedu@centml.ai>
2024-07-02 10:58:08 -07:00
Sirej Dua
15aba081f3
[Speculative Decoding] MLPSpeculator Tensor Parallel support (1/2) ( #6050 )
...
Co-authored-by: Sirej Dua <sirej.dua@databricks.com>
Co-authored-by: Sirej Dua <Sirej Dua>
2024-07-02 07:20:29 -07:00
xwjiang2010
98d6682cd1
[VLM] Remove image_input_type from VLM config ( #5852 )
...
Signed-off-by: Xiaowei Jiang <xwjiang2010@gmail.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Co-authored-by: Roger Wang <ywang@roblox.com>
2024-07-02 07:57:09 +00:00
Alexander Matveev
3476ed0809
[Core] Optimize block_manager_v2 vs block_manager_v1 (to make V2 default) ( #5602 )
2024-07-01 20:10:37 -07:00
