IntenseWebs/openvino
3
0 Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
13,725 Commits 38 Branches 54 Tags
032ac898e21c4008fb8c8dc2da5778c1b3ab4afe
Commit Graph
100 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Vladimir Paramuzov 3a7acbb5cc [GPU] Allow precision conversion on get_state() call (#21658) 2023-12-15 09:57:49 +00:00
Vladimir Paramuzov 2bcc94064f [GPU] Fixed stateful KV cache issues (#21618) 2023-12-13 16:15:00 +04:00
Vladimir Paramuzov 7ab79be0f6 [GPU] Stateful LLMs initial support (#21235) 2023-11-28 14:59:54 +04:00
Vladimir Paramuzov eb94c80804 [GPU] Enable BroadcastTransition before CommonOptimizations (#21198) 2023-11-21 16:54:04 +01:00
Vladimir Paramuzov a6e4d8fc30 [GPU] Cleanup buffer optimization code (#21012) 2023-11-21 14:43:14 +04:00
Vladimir Paramuzov 8f406067d1 [GPU] Remove binary convolution primitive and all related code (#20889) 2023-11-09 09:54:46 +04:00
Vladimir Paramuzov 2afb64b9ea [GPU] Refactor KV cache test (#20802) 2023-11-02 09:54:19 +04:00
Vladimir Paramuzov cde757d66a [GPU] Adjust in/out layouts in codition prim to handle prealloc (#20740) 2023-10-27 16:54:57 +00:00
Vladimir Paramuzov 79f90839b9 [GPU] Int4 utils fix (#20726) 2023-10-27 20:14:08 +04:00
Vladimir Paramuzov cc10b14bab [GPU] Share compilation context with sub-networks (#20706) 2023-10-27 09:58:49 +04:00
Vladimir Paramuzov d9c4ca3021 [GPU] Allow setting remote output for dynamic model (#20608) 2023-10-26 09:09:54 +04:00
Vladimir Paramuzov 307176e5c6 [GPU] Fixed surfaces shape in create_tensor_nv12 helpers (#20539) 2023-10-25 12:46:47 +04:00
Vladimir Paramuzov 3313d1ce7a [GPU] FC with 4-bit weights compression support for iGPU (#20572) 2023-10-25 12:45:31 +04:00
Vladimir Paramuzov 5853509b3c Fixed ngraph reference impl for Range operation (#20631) 
* Fixed ngraph reference impl for Range operation

* Truncate range for integer output type

* explicit static cast
 2023-10-24 10:18:04 +04:00
Vladimir Paramuzov 90ad4c618d [GPU] Grouped decompression scale/zp support (#20491) 2023-10-18 15:56:35 +04:00
Vladimir Paramuzov a00d28aac0 [GPU] Type traits cleanup (#20455) 2023-10-16 14:54:32 +04:00
Vladimir Paramuzov 4da61fc7e7 [TESTS] Fix random generator for custom signed types (#20393) 
* [TESTS] Fix random generator for custom signed types

* Increase threashold for bf16 CPU tests
 2023-10-12 13:15:59 +04:00
Vladimir Paramuzov 0ee0b4d956 [GPU] Fix custom layer (#20220) 2023-10-04 10:18:34 +04:00
Vladimir Paramuzov a2c524019a [GPU] Use ov element type and float16 inside plugin (#20110) 2023-10-03 16:46:50 +04:00
Vladimir Paramuzov 5611f7d65c [GPU] Fix constants reuse (#20003) 2023-09-22 15:29:38 +04:00
Vladimir Paramuzov 541f2dc62f [GPU] Fixed static init order for serialization (#19768) 2023-09-13 00:31:05 +04:00
Vladimir Paramuzov 47fe50ca35 [GPU] 2.0 plugin api impl (#18920) 2023-09-12 11:13:59 +04:00
Vladimir Paramuzov 7e3e1e2480 [GPU] Support of int8 compressed weights for matmul (#19548) 2023-09-11 18:11:34 +04:00
Vladimir Paramuzov 77600c7701 [GPU] Add FullyConnected custom op (#19539) 2023-09-05 16:47:05 +04:00
Vladimir Paramuzov 2f782b2131 [GPU] Add permute primitive instead of manual copy for deconv weights (#19516) 2023-09-04 17:03:49 +04:00
Vladimir Paramuzov 38cad619af [GPU] Allow simple attached mem as input memory for network (#19419) 2023-09-01 09:54:30 +04:00
Vladimir Paramuzov a45e5e03c5 [GPU] Added some formats for pvc (#19388) 2023-08-25 15:09:42 +04:00
Vladimir Paramuzov 3b2e263879 [GPU] Fix reshape optimization (#19270) 2023-08-23 10:25:04 +00:00
Vladimir Paramuzov 526d76c81f [GPU] New headers and namespaces in some parts (#19229) 2023-08-18 15:57:15 +04:00
Vladimir Paramuzov cde0893271 [GPU] Use ov::Tensor intead of ngraph::HostTensor (#18925) 2023-08-10 09:43:45 +00:00
Vladimir Paramuzov 3e0b4b337b [GPU] LUID property added (#18815) 2023-07-28 19:16:25 +04:00
Vladimir Paramuzov 9177809bd7 [GPU] Fixed removal of reorder connected to non-default out port (#18826) 2023-07-28 14:55:46 +04:00
Vladimir Paramuzov 325d02b760 [GPU] Use stream executor and exceptions from new api (#18531) 
* [Common] Handle ov::Exception too in CompiledModel wrapper

* [GPU] Use new threading and exception interfaces where possible
 2023-07-14 20:51:40 +02:00
Vladimir ParamuzovandPavel Durandin 2939818278 [GPU] Enable shape agnostic quantize_scale_shift kernel (#18188) 
* [GPU] Enable shape agnostic quantize_scale_shift kernel

* Resolve conflicts in compile_graph.cpp

* Update compile_graph.cpp

---------

Co-authored-by: Pavel Durandin <pavel.durandin@intel.com>
 2023-07-07 10:38:41 +00:00
Vladimir Paramuzov 3d79bd1ac5 [GPU] Minor layout optimizer refactoring (#17553) 2023-06-16 10:33:53 +04:00
Vladimir Paramuzov 1ce447674e [GPU] Better device input memory reuse (#17853) 2023-06-05 09:30:22 +04:00
Vladimir Paramuzov ac26216869 [GPU] Functional fixes for nvidia (#17735) 2023-06-01 09:45:30 +04:00
Vladimir Paramuzov b95aa84b45 [GPU] Removed some redundant internal passes (#17552) 2023-05-19 13:34:42 +04:00
Vladimir Paramuzov 4cda67da15 [GPU] Fixed unit tests for dGPU (#17541) 2023-05-16 09:36:09 +04:00
Vladimir Paramuzov 8251399222 [GPU] convolution primitive updates (#17351) 2023-05-12 17:19:29 +04:00
Vladimir Paramuzov e282a25d94 [GPU] Minor binary size reduction (#17308) 2023-05-03 09:07:52 +04:00
Vladimir Paramuzov b6761beb55 [GPU] Don't throw exception if no devices are found (#17288) 2023-05-01 13:57:06 +00:00
Vladimir Paramuzov c6ec6d457f [GPU] Move functional tests to intel_gpu/tests (#17248) 2023-04-28 22:10:18 +04:00
Vladimir Paramuzov b019868653 [GPU] Added missing second input to splitted reshape in handle reshape pass (#17254) 2023-04-27 17:21:17 +01:00
Vladimir Paramuzov 6ed85178d5 [GPU] Fix layout propagation logic (#17199) 2023-04-26 14:20:48 +01:00
Vladimir Paramuzov ce5f65af14 [GPU] Use hash of test name for random generator initialization (#17213) 2023-04-26 12:52:38 +04:00
Vladimir Paramuzov f736c71feb [GPU] Fix reshape split for dynamic models + accuracy fix for SAM (#16911) 2023-04-25 09:21:31 +00:00
Vladimir Paramuzov ca1102b855 [GPU] Support MVN cases with axis=-1 w/o decomposition (#17020) 2023-04-25 12:59:03 +04:00
Vladimir Paramuzov d1f1fa2b39 [GPU] Enable broadcast transition pass (#17172) 2023-04-25 09:04:37 +04:00
Vladimir Paramuzov 3bb0fb61f6 [GPU] Support 8d tensors in activation and quantize primitives (#16947) 2023-04-25 09:02:54 +04:00
Vladimir Paramuzov faba5fb71e [Transformations] Add threshold for const comparison in Gelu fusion pass to fuse with fp16 precision (#17042) 2023-04-24 14:37:31 +04:00
Vladimir Paramuzov e8ae1e41ea [GPU] Skip FC fake alignment for some vector by matrix multiplications (#17051) 2023-04-24 14:34:50 +04:00
Vladimir Paramuzov 219a0eebdc [GPU] Fix 1d onednn convolutions (#17038) 2023-04-24 09:24:56 +04:00
Vladimir Paramuzov dbd20ec799 [GPU] Added try/catch for device detection loop to skip platforms which throw an exception (#17011) 2023-04-19 11:05:24 +04:00
Vladimir Paramuzov 231569db16 [GPU] Fix group axis value for blocking desc (#16936) 2023-04-14 14:42:21 +00:00
Vladimir ParamuzovandSergey Shlyapnikov 5299f26168 [GPU] Handle unsupported eltwise fusion for onednn gemm in dynamic cases (#16875) 
* [GPU] Handle unsupported eltwise fusion for onednn gemm in dynamic cases

* Update src/plugins/intel_gpu/tests/fusions/gemm_fusion_test.cpp

Co-authored-by: Sergey Shlyapnikov <Sergeishlyapnikov@gmail.com>

---------

Co-authored-by: Sergey Shlyapnikov <Sergeishlyapnikov@gmail.com>
 2023-04-13 15:55:44 +04:00
Vladimir Paramuzov 67c07ccebe [GPU] Support 7D and 8D tensors (#16810) 2023-04-13 09:04:14 +04:00
Vladimir Paramuzov 7d4496bb12 [GPU] Remove unused constants from the graph (#16873) 2023-04-12 16:52:26 +04:00
Vladimir Paramuzov 6d82f36050 Enable nop elimination for f16 type (#16749) 2023-04-07 09:18:27 +04:00
Vladimir Paramuzov b2a64e8c3a [GPU] Enable dynamic shapes support for gather elements (#16727) 2023-04-06 11:14:36 +04:00
Vladimir Paramuzov f5e199c494 [GPU] Don't reorder weights when can reinterpret (#16714) 
* [GPU] Don't reorder weights when can reinterpret

* [GPU] Test fixes
 2023-04-05 16:20:51 +04:00
Vladimir Paramuzov f7d15e12c8 [GPU] Refactor dimensions jitter and max rank related code (#16603) 
* [GPU] Refactor dimensions jitter and max rank related code
 2023-04-03 13:34:06 +02:00
Vladimir Paramuzov 906939a1f1 [GPU] Fixed invalid is_dynamic flag value for scalar inputs (#16565) 2023-03-28 10:03:51 +04:00
Vladimir Paramuzov 98237b06b5 [GPU] Update memory_statistics property impl (#16399) 2023-03-21 08:52:52 +04:00
Vladimir Paramuzov a46fc47e6a [GPU] Enable tile with dynamic input (#16364) 2023-03-20 12:49:35 +04:00
Vladimir Paramuzov bb59672639 [GPU] Fixed shape agnostic scatter nd update kernel (#16319) 2023-03-17 09:57:25 +04:00
Vladimir Paramuzov 28d3e1087e [GPU] Fix strided slice kernel with begin/end/stride as inputs (#16302) 2023-03-15 16:25:45 +04:00
Vladimir Paramuzov e44fd03d2a [GPU] Shape agnostic concat kernel + refactoring (#16170) 2023-03-15 09:47:31 +04:00
Vladimir Paramuzov 75b48f2153 [GPU] Changed impls cache key type to avoid hash collisions (#16130) 2023-03-08 16:09:55 +04:00
Vladimir Paramuzov a1eb76ad06 [GPU] Move is_local_block_io_supported WA to kernel selector (#15235) 2023-03-07 15:12:08 +04:00
Vladimir ParamuzovandEddy Kim eff0bce7e3 [GPU] Move some op parameters from node to primitive class (#16070) 
* [GPU] Move parameters of conv and quantize primitive from node to primitive

---------

Co-authored-by: Eddy Kim <eddy.kim@intel.com>
 2023-03-07 08:56:00 +04:00
Vladimir Paramuzov 27ac7d9092 [GPU] backend independent code for fuse params in program_node (#16028) 2023-03-02 09:18:29 +04:00
Vladimir Paramuzov c5c7e4ff65 [GPU] Cleanup tuning cache methods (#16000) 2023-03-01 16:30:47 +04:00
Vladimir Paramuzov 3de00347f3 [GPU] Code cleanup (#16014) 
* [GPU] Improve exception message for program build

* [GPU] Code cleanup
 2023-03-01 14:05:59 +04:00
Vladimir Paramuzov 9b840a789c [GPU] Minor fixes for dynamic models (#15543) 2023-02-13 10:34:41 +04:00
Vladimir Paramuzov 34d16b8777 [GPU] Move tuning cache loading to kernel selector (#15112) 
* [GPU] Move tuning cache loading to kernel selector. Remove tuning modes

* [GPU] Removed kernel runner
 2023-01-20 15:17:12 +04:00
Vladimir Paramuzov 5b389860a1 [GPU] Remove legacy activation fused ops from program (#15075) 
* [GPU] Fix winograd kernel

* [GPU] Remove fusion via legacy mechanism

* [GPU] Conversion to legacy activations in kernel selector helper
 2023-01-20 09:55:53 +04:00
Vladimir Paramuzov 2201a5f83e Added common ov::execution_mode hint (#15048) 
* [GPU] Added common exec mode hint and gpu support

* Add ov::inference precision and update usages. Deprecate ov::hint::inference_precision property

* [GPU] Execution mode tests and fixes

* Fixed code style

* Moved execution_mode test to common. Fixes for python API

* Remove deprecations for hint::inference_precision and just keep both

* Fix test
 2023-01-18 20:13:00 +04:00
Vladimir ParamuzovandMingyu Kim 0c79caa45c [GPU] Use in-order queue for devices with dpas support (#15061) 
* [GPU] Use in-order queue for devices with dpas support

* [GPU] Don't query properties in inference time

Co-authored-by: Mingyu Kim <mingyu.kim@intel.com>
 2023-01-16 15:19:56 +04:00
Vladimir Paramuzov 9c7beee5e5 [GPU] Added missing supports_image initialization in EngineInfo. Fix some kernels with image support (#14885) 2023-01-12 12:06:01 +03:00
Vladimir Paramuzov 4feaeaad68 [GPU] Remote context reuse and internal config update (#14635) 2023-01-11 11:14:03 +00:00
Vladimir Paramuzov 028cf7a34d [GPU] Code cleanup (#14974) 2023-01-09 15:43:26 +04:00
Vladimir Paramuzov e23b8492c5 [GPU] Change default infer precision to fp16 (#14752) 2023-01-04 15:55:44 +04:00
Vladimir Paramuzov 5a4dbb6760 [GPU] Enable oneDNN primitives cache (#14835) 2023-01-02 13:44:43 +04:00
Vladimir Paramuzov 13c8b4fdc7 [GPU] Better extension requirements checks in kernels. Subgroups basic emulation (#13926) 2022-12-29 09:08:05 +04:00
Vladimir Paramuzov d2d952fb50 [GPU] Rename GPU unit and func tests (#14719) 2022-12-28 11:13:28 +01:00
Vladimir Paramuzov 31d91033d9 [GPU] new shape infer for several trivial primitives (#14803) 
* [GPU] new shape infer for several primitives

* [GPU] Make forwatd_input0_shape method tempalte
 2022-12-28 09:12:46 +04:00
Vladimir Paramuzov 06c2d3da52 [GPU] Fixed debug graph dump (#14806) 2022-12-26 18:47:35 +04:00
Vladimir Paramuzov 3414fc402a [GPU] Rename exec mode options (#14798) 2022-12-23 17:02:31 +04:00
Vladimir Paramuzov 04c0dbbf60 [GPU] Fix shape infer for 0d broadcast (#14744) 2022-12-21 09:39:20 +04:00
Vladimir Paramuzov 6bc4f00695 [GPU] Refactored GPU verbose utils (#14549) 
* [GPU] Refactored GPU verbose utils

* [GPU] More refacoring

* Fix build
 2022-12-20 11:06:01 +04:00
Vladimir Paramuzov ee256e801c [GPU] Removed legacy public gpu config and related processing (#14707) 2022-12-19 11:19:07 +04:00
Vladimir Paramuzov fa1cddf278 [GPU] Use ASSERT_* instead of EXPECT_* in most of the cases (#14535) 2022-12-16 09:09:57 +04:00
Vladimir Paramuzov 78bd280a30 [GPU] Fix invalid memory access in lrn fsv16 kernel (#14536) 2022-12-13 17:16:13 +04:00
Vladimir Paramuzov c581bbd6dd [GPU] Fixed invalid offset for tensor with batch padding (#14460) 2022-12-13 09:14:59 +04:00
Vladimir Paramuzov 6df011c9f6 [GPU] Use transformed gemm shapes for layout optimizer checks (#14407) 2022-12-08 09:12:02 +04:00
Vladimir Paramuzov 7b7e1d19b0 [GPU] allow softmax_bf for axis=Y && X==1 case (#14443) 2022-12-07 13:04:47 +04:00
Vladimir Paramuzov d0a6122cb2 [GPU] Fixed incorrect explicit transpose op inserted for MatMul (#14400) 2022-12-06 09:50:14 +04:00
Vladimir Paramuzov 48c0897b0f [GPU] Adjust default num streams (#14308) 2022-12-05 09:59:03 +04:00
Vladimir Paramuzov b59aea873c [GPU] Fix exception for dyn reshape followed by static node (#14305) 
* [GPU] Fix exception for dyn reshape followed by static node

* [GPU] Weights reorder fix
 2022-12-01 00:52:03 +03:00