Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

module update #6

Closed
wants to merge 260 commits into from
Closed
Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
260 commits
Select commit Hold shift + click to select a range
8ee06dc
Energy op (#2121)
turbotoribio Jul 13, 2023
e687c60
Initialize test output variables (#2124)
rascani Jul 13, 2023
c3fddae
Move kTfLiteAbort to micro_context.h (#2072)
tanmaydas82 Jul 14, 2023
6da07de
Generic benchmarking. (#2125)
tanmaydas82 Jul 14, 2023
2548022
Filter_bank ops (#2123)
turbotoribio Jul 14, 2023
699c517
Explicit load for py_library. (#2127)
advaitjain Jul 14, 2023
15154aa
Add RISC-V as a requirement for the issue on error. (#2128)
advaitjain Jul 14, 2023
52007f6
Fix benchmarking utils on RISC-V (#2130)
rascani Jul 15, 2023
70aed11
Adds FFT Auto Scale Op (#2134)
suleshahid Jul 19, 2023
61bc56d
Fix numpy array compare for Numpy 1.25 (#2136)
suleshahid Jul 19, 2023
ed11500
Delay OP python extension (#2138)
suleshahid Jul 19, 2023
55037d2
Adds IRFFT Op to Signal Library (#2137)
suleshahid Jul 20, 2023
40a6936
Increase tolerance for fft_test (#2139)
suleshahid Jul 20, 2023
bb3fda3
Python extensions for energy, framer, overlap_add Signal OPs (#2142)
suleshahid Jul 24, 2023
e2cdefd
Python extensions for filter bank and stacker Signal Ops. (#2143)
suleshahid Jul 25, 2023
c20bab1
No longer sync TfLite OpResolver from upstream. (#2145)
advaitjain Jul 25, 2023
aa945a0
Add the AddEmbeddingLookup method to micro_mutable_op_resolver.h (#2148)
ddavis-2015 Jul 26, 2023
ca74563
feat: build and test a Python distribution package `tflite_micro` (#2…
rkuester Aug 1, 2023
a7846a1
Add Q16x8 Depthwise Conv Support (#2140)
mbrooksx Aug 1, 2023
f19f21d
Disable whl_test since we are not building python package internally.…
advaitjain Aug 4, 2023
ed7830f
build(bazel): upgrade @rules_python to 0.24.0 (#2156)
rkuester Aug 7, 2023
aa61a91
Automated sync from github.com/tensorflow/tensorflow (#2158)
TFLM-bot Aug 9, 2023
d5f819d
Fix armclang build issue (#2155)
mansnils Aug 9, 2023
80cb11b
`load()`s are being added in preparation for changes being made to Bl…
turbotoribio Aug 10, 2023
3c1c19b
Automated sync from github.com/tensorflow/tensorflow (#2161)
TFLM-bot Aug 10, 2023
901d830
Add code generator experiment (#2162)
rascani Aug 10, 2023
feb2669
Add hello world codegen example (#2163)
rascani Aug 11, 2023
e3d1a5a
Add py load statement for codegen (#2166)
rascani Aug 14, 2023
46ccaf8
Remove unused and combine similar loads (#2165)
rascani Aug 14, 2023
3c291cf
Update int16 LSTM test case with zero_point=0 (#2171)
advaitjain Aug 16, 2023
f4c2d08
feat(py): give package a developmental version label (#2168)
rkuester Aug 16, 2023
a7b6b85
Remove LICENSE export for non-existent file (#2173)
rascani Aug 16, 2023
704d307
Remove unnecessary visibility. (#2174)
advaitjain Aug 16, 2023
3bd11ea
build(py): add target that publishes to PyPI (#2172)
rkuester Aug 17, 2023
389d8f5
Automated sync from github.com/tensorflow/tensorflow (#2177)
TFLM-bot Aug 18, 2023
a0f8970
Generate op table and subgraph invoke functions (#2176)
rascani Aug 18, 2023
0228e3e
Add expand_stamp_vars docstring and fix import (#2178)
rascani Aug 21, 2023
b4060ef
Automated sync from github.com/tensorflow/tensorflow (#2182)
TFLM-bot Aug 23, 2023
e01fdf0
`PCAN C++` (#2180)
turbotoribio Aug 23, 2023
34691a6
codegen: Add TFLiteContext and TfLiteNodes (#2179)
rascani Aug 23, 2023
17987e1
Optimized kernel reference fallbacks: CONV (#1911)
ddavis-2015 Aug 24, 2023
f8ab03d
Generate BuiltinData for FullyConnected (#2184)
rascani Aug 24, 2023
a73eb10
Automated sync from github.com/tensorflow/tensorflow (#2185)
TFLM-bot Aug 24, 2023
c795f31
Automated sync from github.com/tensorflow/tensorflow (#2187)
TFLM-bot Aug 25, 2023
563f79f
Increase arena size to accomodate larger models. (#2188)
advaitjain Aug 25, 2023
1e7c710
remove unused symbol from BUILD file. (#2190)
advaitjain Aug 28, 2023
ad83c8c
`Implementation preserve_all_tensors and GetTensor features in TFLM i…
turbotoribio Aug 28, 2023
4d07bd0
Generate static TfLiteEvalTensors (#2193)
rascani Aug 29, 2023
b45012b
Fix RISC-V codegen example (#2199)
rascani Aug 29, 2023
8f72639
Move benchmarking model to static memory (#2200)
rascani Aug 29, 2023
fd4b198
Increase Corstone-300 SRAM size (#2201)
rascani Aug 30, 2023
1dfe2bc
Deduplicate StridedSlice Init and Prepare (#2203)
rascani Aug 31, 2023
a72c055
Automated sync from github.com/tensorflow/tensorflow (#2198)
TFLM-bot Sep 1, 2023
4944999
`Python PCAN` (#2204)
turbotoribio Sep 1, 2023
87b99b2
Automated sync from github.com/tensorflow/tensorflow (#2205)
TFLM-bot Sep 5, 2023
0cc7719
Add python load rules to person detection utils (#2208)
rascani Sep 5, 2023
5d542d7
Port BATCH_MATMUL (#2147)
ddavis-2015 Sep 6, 2023
4813acf
Define preprocessor schema (#2211)
rascani Sep 7, 2023
5ec9d47
Update readmes (#2216)
mansnils Sep 11, 2023
c3cfd34
Make MicroContext and MicroGraph abstract (#2202)
rascani Sep 11, 2023
28b143f
Adding micro_log.h header to micro_interpreter_context.cc and batch_m…
turbotoribio Sep 12, 2023
f800eb8
Automated sync from github.com/tensorflow/tensorflow (#2214)
TFLM-bot Sep 12, 2023
3323a41
Implement MicroContext/Graph for codegen (#2218)
rascani Sep 12, 2023
77e2cdb
Create codegen_preprocessor (#2219)
rascani Sep 14, 2023
ad8d238
Automated sync from github.com/tensorflow/tensorflow (#2221)
TFLM-bot Sep 14, 2023
d1bc163
Add Xtensa dockerfiles (#2223)
rascani Sep 15, 2023
bc75bf6
Update Xtensa actions to use new images (#2224)
rascani Sep 15, 2023
7d1dee8
Fix Hifi5 docker image name (#2227)
rascani Sep 15, 2023
2f2c744
Automated sync from github.com/tensorflow/tensorflow (#2226)
TFLM-bot Sep 15, 2023
127f88f
remove unused visibility (#2230)
suleshahid Sep 18, 2023
3022d09
CMSIS-NN: Move kernel sums to prepare for FC and SVDF INT8 (#2233)
mansnils Sep 19, 2023
9145957
Automated sync from github.com/tensorflow/tensorflow (#2234)
TFLM-bot Sep 19, 2023
e0e052d
feat(py): add scripts for building and uploading to PyPI (#2236)
rkuester Sep 20, 2023
4cdcb6e
build: Add -Wno-unused-parameter to bazel build (#2237)
afoxley Sep 21, 2023
d027f2a
Update python version in Xtensa images (#2228)
rascani Sep 21, 2023
1e9b4c5
Create Make helpers for running codegen (#2222)
rascani Sep 21, 2023
5beca9a
Remove unwanted ;(semicolon) in micro_profiler.h (#2235)
tilakrayal Sep 21, 2023
ce593cf
xtensa: Add more flexible way of selecting optimized kernels (#2240)
afoxley Sep 22, 2023
4bcc144
xtensa: Add BUILD file and workspace for nnlib_hifi4 (#2239)
afoxley Sep 22, 2023
e4fa17c
Disable Corstone 300 FVP codegen targets (#2243)
rascani Sep 25, 2023
6b43adc
Automated sync from github.com/tensorflow/tensorflow (#2242)
TFLM-bot Sep 25, 2023
0fb5805
Restore cell_state scale changed with #2171 (#2245)
advaitjain Sep 26, 2023
266d49d
Automated sync from github.com/tensorflow/tensorflow (#2250)
TFLM-bot Sep 27, 2023
4bbdef5
Automated sync from github.com/tensorflow/tensorflow (#2251)
TFLM-bot Sep 29, 2023
cdf028c
fix(pypi): pass some env v'bles through during containerized whl buil…
rkuester Oct 2, 2023
13564c7
Add UINT32 support to CAST (#2255)
ddavis-2015 Oct 2, 2023
6da4e75
Add INT32 support to DIV (#2256)
ddavis-2015 Oct 2, 2023
a438ba8
TFLM vs TfLite debugger script python (#2259)
turbotoribio Oct 2, 2023
f391e68
Download Cadence Nature DSP from GitHub (#2241)
shlmregev Oct 5, 2023
7b5bd33
build(bazel): fix hashes for Py package wrapt for Python 3.11 (#2265)
rkuester Oct 9, 2023
a6e6ea1
Use fallback for 8-bit FC kernel with non-symmetric weights in CMSIS-…
AdrianLundell Oct 9, 2023
568d181
Update op resolver generate script (#2266)
mansnils Oct 10, 2023
778d6f8
Remove extra Xtensa tools install (#2271)
rascani Oct 12, 2023
d5a83db
Change micro_speech example (#2249)
ddavis-2015 Oct 12, 2023
f2f53c1
Automated sync from github.com/tensorflow/tensorflow (#2269)
TFLM-bot Oct 13, 2023
ba8364a
ci: remove obsolete ci/flatbuffers_for_tf_sync (#2273)
rkuester Oct 13, 2023
ceb8cc8
Add missing include path for hifimini (#2275)
rascani Oct 17, 2023
6ec8c54
micro_speech: Add py_rules loads and import paths (#2276)
rascani Oct 17, 2023
ea4c7a6
Remove audio_preprocessor golden tests (#2277)
rascani Oct 17, 2023
49ab008
Automated sync from github.com/tensorflow/tensorflow (#2280)
TFLM-bot Oct 18, 2023
6bb1e3a
build: update flatbuffers dependency to v23.5.26 (#2274)
rkuester Oct 18, 2023
4e903b1
`Adds a test to MicroInterpreter Test for GetTensor Failure Case` (#…
turbotoribio Oct 19, 2023
fa16acd
Add #include of mul.h (#2282)
shlmregev Oct 19, 2023
005e2fe
Automated sync from github.com/tensorflow/tensorflow (#2283)
TFLM-bot Oct 20, 2023
96bfbdb
`Layer by Layer TFLM vs TfLite python debug tool` (#2284)
turbotoribio Oct 20, 2023
7ac0b66
Automated sync from github.com/tensorflow/tensorflow (#2285)
TFLM-bot Oct 21, 2023
89475eb
Adds optimized kernels for Signal ops (#2287)
suleshahid Oct 25, 2023
28389e0
Automated sync from github.com/tensorflow/tensorflow (#2292)
TFLM-bot Oct 30, 2023
e38f488
renaming Flatbuffers BUILD (#2294)
turbotoribio Oct 30, 2023
6d337dc
Microfrontend warnings (#2296)
ericheim Oct 31, 2023
00067fb
add layer_by_layer debugging tool to makefile (#2299)
turbotoribio Nov 6, 2023
58d8f84
Makefile changes for optimized signal directories (#2303)
suleshahid Nov 6, 2023
c845961
CMSIS-NN: Add transpose conv operator (#2304)
mansnils Nov 7, 2023
38c657a
Automated sync from github.com/tensorflow/tensorflow (#2297)
TFLM-bot Nov 7, 2023
a4d6c5c
Automated sync from github.com/tensorflow/tensorflow (#2307)
TFLM-bot Nov 8, 2023
904edd2
Automated sync from github.com/tensorflow/tensorflow (#2308)
TFLM-bot Nov 10, 2023
ecb3b32
Automated sync from github.com/tensorflow/tensorflow (#2311)
TFLM-bot Nov 14, 2023
5beca7c
CMSIS-NN: Add int4 kernel support to conv2d, depthwise and fully conn…
mansnils Nov 14, 2023
0996efe
Add old benchmarks back to make build (#2315)
rascani Nov 15, 2023
7811156
PyPI whl build and upload to PyPI test server nightly job (#2317)
jwithers Nov 15, 2023
0264234
adding twine to pypi upload script (#2318)
jwithers Nov 15, 2023
34635a6
Add support for building Position Independent Code (PIC) for Hexagon …
shlmregev Nov 21, 2023
a8421cf
Enable Corstone 300 to run on CI backend (#2321)
ddavis-2015 Nov 21, 2023
743e85c
Move branch_protection to queue_rules (#2322)
rascani Nov 22, 2023
ecf56f4
Fix zero-point calculation rounding error (#2326)
ddavis-2015 Nov 22, 2023
944b23e
Use branch_protection_injection_mode for mergify (#2328)
rascani Nov 25, 2023
5043319
SPACE_TO_BATCH_ND: update output tensor shape (#2335)
ddavis-2015 Nov 28, 2023
c1a5f16
Adding ReadMe for Layer by Layer Debug Targets (#2305)
turbotoribio Nov 28, 2023
2f97d82
Initialize ArithmeticParams & OpDataReduce (#2337)
rascani Nov 30, 2023
83bbff9
Revert "SPACE_TO_BATCH_ND: update output tensor shape" (#2339)
rascani Nov 30, 2023
a7ff71a
Automated sync from github.com/tensorflow/tensorflow (#2346)
TFLM-bot Dec 7, 2023
5f5fcc8
Unique Init, Prepare, Eval functions Kernels A-M (#2344)
turbotoribio Dec 8, 2023
5261ba0
CMSIS-NN: Update cortex generic target with M85 for GCC (#2341)
ArmRyan Dec 8, 2023
1a9fd2e
Automated sync from github.com/tensorflow/tensorflow (#2349)
TFLM-bot Dec 8, 2023
1c13347
Add MicroSnPrintf (#2343)
ddavis-2015 Dec 9, 2023
c64b691
Fix cortex_m_generic C++ linkage errors (#2350)
AdrianLundell Dec 11, 2023
c0043da
Fix bazel 7.0 bzlmod update (#2352)
suleshahid Dec 12, 2023
b654b1b
Fix Bazel dependencies for targets under signal/ (#2353)
shlmregev Dec 12, 2023
ec57f3b
Match MicroProfiler docs to behavior (#2340)
rascani Dec 12, 2023
93d4b1c
Unique Init, Prepare, Eval functions Kernels N-Z (#2351)
turbotoribio Dec 12, 2023
47f77ab
Benchmark (#2348)
ddavis-2015 Dec 14, 2023
b24a9ed
Automated sync from github.com/tensorflow/tensorflow (#2361)
TFLM-bot Dec 18, 2023
b814f38
Fix metrics deps (#2364)
rascani Dec 19, 2023
bfd5866
Enable subpackage visibility for signal (#2363)
rascani Dec 19, 2023
c5c3b3e
codegen: Remove preprocessor (#2355)
rascani Dec 19, 2023
436b6d2
mutable_op_resolver script: remove dependency to tensorflow-cpu (#2365)
mansnils Dec 20, 2023
b766947
codegen: Add Bazel macro for inference library (#2366)
rascani Dec 21, 2023
6576ef7
Automated sync from github.com/tensorflow/tensorflow (#2370)
TFLM-bot Jan 2, 2024
17d0e7f
Xtensa LSTM: (#2150)
cad-audio Jan 2, 2024
bb54330
Compute output shapes for some kernels (#2356)
ddavis-2015 Jan 2, 2024
77f655c
build(bazel): add missing dep on python package `pillow` (#2371)
rkuester Jan 3, 2024
0c03d3d
build(bazel): pin bazel version to 7.0.0 (#2360)
rkuester Jan 3, 2024
276deb7
Update Ethos-U download links (#2376)
AdrianLundell Jan 5, 2024
95467d2
removing usage of `tflm_runtime` and replacing usage with new runtime…
turbotoribio Jan 5, 2024
515cfd2
Adjust checks in reshape (#2383)
mansnils Jan 8, 2024
833f7bb
Add dep for codegen build (#2385)
suleshahid Jan 9, 2024
a9405e3
Exclude xa_nn_activations_asym8_asym8.c from NNLib build for Hifi4 (#…
shlmregev Jan 9, 2024
06499e8
codegen: Remove make helpers for codegen (#2378)
rascani Jan 9, 2024
85e19dd
Add missing ops to OpResolvers (#2386)
rascani Jan 10, 2024
ca5358f
DTLN example: (#2012)
cad-audio Jan 10, 2024
7a63bce
Revert "Compute output shapes for some kernels (#2356)" (#2390)
rascani Jan 11, 2024
a3296d6
Disable DTLN test on hifimini & VP6 (#2391)
rascani Jan 11, 2024
5399638
Remove dynamic allocation from VP6 Reduce kernel (#2394)
rascani Jan 12, 2024
fdf7876
Remove unused BUILD load (#2397)
suleshahid Jan 17, 2024
2b5f0fd
Update TF Sync PR reviewer (#2398)
rascani Jan 17, 2024
21d7203
Automated sync from github.com/tensorflow/tensorflow (#2395)
TFLM-bot Jan 19, 2024
2d80ee4
Enable tflm_benchmark on HiFi5 (#2400)
rascani Jan 19, 2024
73e419e
Enable optimized ADD/SUB for HiFi5 (#2399)
rascani Jan 19, 2024
4b2bdd1
Remove unused visibility spec (#2403)
rascani Jan 19, 2024
324ae1e
Enable bool input for CAST (#2401)
rascani Jan 22, 2024
0c4738a
cmsis-nn: refactor conv to get rid of duplicate code (#2402)
mansnils Jan 29, 2024
c93be16
Add a makefile for generic Cortex-A targets. (#2405)
shlmregev Jan 30, 2024
fd3ca73
adds dep needed for xtensa fft (#2409)
suleshahid Jan 31, 2024
f2b8cfc
cortex_m_corstone_300_makefile: add build fix workaround for armclang…
mansnils Jan 31, 2024
13de409
Automated sync from github.com/tensorflow/tensorflow (#2414)
TFLM-bot Jan 31, 2024
9d1dd7b
revert xtensa include vector (#2441)
suleshahid Jan 31, 2024
42f4bb8
Automated sync from github.com/tensorflow/tensorflow (#2445)
TFLM-bot Feb 6, 2024
a549448
RISC-V target makefile: make the arch, abi and mcmodel configurable. …
shlmregev Feb 7, 2024
0ec823e
Update CMSIS-NN CONV and LSTM implementations (#2446)
AdrianLundell Feb 9, 2024
66034be
Generic Benchmark Tool changes (#2451)
ddavis-2015 Feb 14, 2024
a00b9a0
Remove initialized but unused variables. (#2453)
shlmregev Feb 14, 2024
749892a
Remove unused variables in Xtensa implmenetation of LSTM. (#2456)
shlmregev Feb 15, 2024
699564d
Fix broken hifi3 NDSP lib patch file (#2457)
shlmregev Feb 16, 2024
ce72f7b
Fix a linker error: undefined kMaxBufferCount (#2458)
shlmregev Feb 16, 2024
09c796e
Fix Corstone Ethos-U CI failure (#2460)
ddavis-2015 Feb 20, 2024
ab76710
Add missing micro_context dependency (#2462)
rascani Feb 22, 2024
3119632
Add workflow_call to cortex_m.yml (#2466)
rascani Feb 22, 2024
a7238a1
Disable per channel quantization for FC layers (#2465)
rascani Feb 22, 2024
29526e4
Replace tensorflow-cpu usage with tensorflow (#2396)
rascani Feb 23, 2024
f5e498b
cmsis-nn: update fully connected int8 (#2469)
mansnils Feb 23, 2024
8085cec
Add kernel build flag for prioritizing speed or size (#2408)
mansnils Feb 23, 2024
cfa4c91
Fix array out-of-bounds access in WideDynamicFuncLut (#2468)
kismeter Feb 24, 2024
0445acc
Use absl::Status instead of tsl::Status (#2477)
rascani Feb 27, 2024
b991e72
Fix a linker error: undefined kMaxSmallSize (#2476)
shlmregev Feb 27, 2024
20fd5b2
Prevent runtime_shape.cc removal during sync (#2484)
rascani Feb 28, 2024
9adee08
Remove experimental flag for per-channel FC quantization (#2482)
rascani Feb 28, 2024
1f7f54b
Fix GitHub workflow conditionals (#2489)
rascani Mar 4, 2024
ec086d2
Fix tests_entry & tests_post workflows (#2493)
rascani Mar 7, 2024
74c4632
Unique Init, Prepare, Eval functions of signal kernels. (#2497)
shlmregev Mar 11, 2024
8f143b5
Add workaround for newer GitHub runners (#2505)
rascani Mar 15, 2024
32876bc
Automated sync from github.com/tensorflow/tensorflow (#2502)
TFLM-bot Mar 15, 2024
f9044b0
Remove Wave as dependency (#2508)
rascani Mar 15, 2024
d308847
Update examples and documentation for Arm(R) Corstone(TM)-300 FVP (#2…
mansnils Mar 15, 2024
98de885
Automated sync from github.com/tensorflow/tensorflow (#2510)
TFLM-bot Mar 18, 2024
178cfeb
CMSIS-NN LSTM issue fixes (#2504)
AdrianLundell Mar 18, 2024
5e17f07
Add option to rename symbols in Nature DSP library (#2509)
shlmregev Mar 19, 2024
79e3f35
Remove the `output_to_genfiles = True` parameter from Starlark rules …
turbotoribio Mar 19, 2024
15092df
IF kernel nulltpr fix (#2515)
suleshahid Mar 21, 2024
cf0a560
Add GNU Make rule to print GENDIR (#2514)
shlmregev Mar 22, 2024
d20aa31
Automated sync from github.com/tensorflow/tensorflow (#2516)
TFLM-bot Mar 25, 2024
0ef3452
Add support for architecture armv8.2-a in Cortex-A target (#2519)
shlmregev Mar 26, 2024
7005d60
Compile with -ffp-mode=full for armclang (#2517)
AdrianLundell Mar 26, 2024
1b2245f
Download CMSIS for ref kernels as well for cortex_m_generic (#2525)
mansnils Mar 28, 2024
bb4fc83
Pass cc_op_defs as deps, not srcs, to underlying cc_library rule (#2522)
rascani Mar 28, 2024
7aa0d93
Basic makefile for MIPS architecture (#2536)
shlmregev Apr 3, 2024
f99fcf8
Use scratch memory instead of perm buffer for the RFFT kernel's work…
shlmregev Apr 12, 2024
deec0d8
gen_micro_mutable script: update import path (#2545)
mansnils Apr 17, 2024
26fbc4b
Replace test with invalid-enum-load (#2546)
rascani Apr 17, 2024
a17682d
Quantize, dequantize optimizations for HiFi targets (#2544)
cad-audio Apr 18, 2024
9741578
Enable int16_t support for CMSIS-NN LSTM kernel (#2551)
AdrianLundell Apr 23, 2024
fc498a8
Automated sync from github.com/tensorflow/tensorflow (#2553)
TFLM-bot Apr 24, 2024
a3be93b
Enable hifi5 optimized kernels (#2555)
rascani Apr 26, 2024
0fc313b
Revert "gen_micro_mutable script: update import path" (#2556)
rascani Apr 26, 2024
fe4229c
Automated sync from github.com/tensorflow/tensorflow (#2558)
TFLM-bot Apr 29, 2024
e5e2305
CMSIS-NN: Add int32 bias support int16xint8 conv (#2557)
mansnils Apr 29, 2024
a32a33b
Fix misspelling of integer (#2554)
heavyimage Apr 30, 2024
ed36c46
Fix patch file for Nature NDSP library for hifi3 (#2541)
shlmregev May 2, 2024
cf10f94
Enable optimized dequantize on HIFI3 (#2560)
rascani May 3, 2024
6b766e3
Update CMSIS-NN download SHA-1 (#2563)
mansnils May 7, 2024
8a1fe96
Update documentation (#2565)
ddavis-2015 May 8, 2024
5e4e51b
Check whether the bias tensor is `nullptr` before accessing the type.…
dansuh17 May 9, 2024
c377b74
Automated sync from github.com/tensorflow/tensorflow (#2567)
TFLM-bot May 14, 2024
7dbd9d2
Add meta-data to generic benchmark (#2496)
ddavis-2015 May 14, 2024
b281a0d
Automated sync from github.com/tensorflow/tensorflow (#2571)
TFLM-bot May 15, 2024
1973d40
Add cortex300 download for aarch64 linux host (#2573)
ArmRyan May 16, 2024
b2cf5b3
Automated sync from github.com/tensorflow/tensorflow (#2572)
TFLM-bot May 17, 2024
ae690a1
Remove #definitions of MAX_RFFT_PWR, MIN_RFFT_PWR from Xtensa flags (…
shlmregev May 17, 2024
94e7cba
Fix corstone-300 simulator path (#2579)
ddavis-2015 May 17, 2024
6028493
Fixes for generic benchmark wrt corstone-300 (#2581)
ddavis-2015 May 18, 2024
04d071e
Fix test_cortex_m_corstone_300 ci job (#2582)
ArmRyan May 20, 2024
9045e1b
Patch NDP Hifi3 to allow override of MAX_RFFT_LEN (#2588)
shlmregev May 20, 2024
8e22946
Automated sync from github.com/tensorflow/tensorflow (#2585)
TFLM-bot May 20, 2024
dfdd666
more fixes for generic benchmark wrt Corstone-300 (#2592)
ddavis-2015 May 21, 2024
d069ba8
Remove unused visibility spec (#2594)
rascani May 22, 2024
efa694b
Import visualize from tensorflow, not tflite-micro (#2595)
rascani May 23, 2024
be11bd7
Add support for 16 bit activations in ExpandDims operation. (#2589)
robertkamara May 24, 2024
71fba59
Fail per-channel quantized FullyConnected layers (#2602)
rascani May 29, 2024
8be43d1
zephyr: module.yml: added
XenuIsWatching Apr 17, 2023
73258a3
zephyr_static_export.sh: added
JordanYates Jun 13, 2024
48613f7
third_party_static: exported
JordanYates Jun 27, 2024
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Update int16 LSTM test case with zero_point=0 (tensorflow#2171)
* tensorflow@a76d154 has the changes needed to create test data for the current PR. I reverted the changes back to have the script be mostly unchanged (except for some typos and trailing whitespaces).
 * To get the new quantization parameters:
```bash
bazel run tensorflow/lite/micro/kernels/testdata:lstm_test_data_generator
```
And then manually copy over the parameters into `lstm_test_data.cc`

BUG=http://b/296130372
  • Loading branch information
advaitjain authored Aug 16, 2023
commit 3c291cf7d495a83df1d515d8791f5febdcedb052
14 changes: 8 additions & 6 deletions tensorflow/lite/micro/kernels/testdata/lstm_test_data.cc
Original file line number Diff line number Diff line change
Expand Up @@ -251,12 +251,14 @@ NodeQuantizationParameters Get2X2Int16LstmQuantizationSettings() {

// state quantization parameters
quantization_settings.input = {/*scale=*/3.0518044e-5, /*zp=*/0,
/*symmetry=*/false};
quantization_settings.output = {/*scale=*/1.8310826e-5, /*zp=*/-5461,
/*symmetry=*/false};
quantization_settings.hidden_state = {/*scale=*/1.8310826e-5, /*zp=*/-5461,
/*symmetry=*/false};
quantization_settings.cell_state = {/*scale=*/0.00024414062, /*zp=*/0,
/*symmetry=*/true};
quantization_settings.output = {/*scale=*/2.1362956633198035e-05, /*zp=*/0,
/*symmetry=*/true};
quantization_settings.hidden_state = {/*scale=*/2.1362956633198035e-05,
/*zp=*/0,
/*symmetry=*/true};
quantization_settings.cell_state = {/*scale=*/0.00024414807580797754,
/*zp=*/0,
/*symmetry=*/true};

// gate quantization parameters
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -17,15 +17,15 @@
2. Print the intermediate step outputs inside the LSTM for a single step LSTM invocation (Get2X2GateOutputCheckData in .cc)
3. Print the outputs for multi-step LSTM invocation (Get2X2LstmEvalCheckData in .cc)

Every invocation gives three types information:
1. Quantized output: kernel output in integer
Every invocation gives three types information:
1. Quantized output: kernel output in integer
2. Dequantized output: Quantized output in floating point representation
3. Float output: output from the floating point computation (i.e., float kernel)

Note:
Note:
1. Change quantization settings in _KERNEL_CONFIG to see the outcomes from various quantization schema (e.g., 8x8 Vs. 16x8)
2. Only single batch inference is supporte here. Change _GATE_TEST_DATA or _MULTISTEP_TEST_DATA to see kernel outputs on different input data
3. The quantization computation here is not the exact as the c++ implementation. The integer calculation is mimiced here using floating point.
3. The quantization computation here is not the exact as the c++ implementation. The integer calculation is emulated here using floating point.
No fixed point math is implemented here. The purpose is to illustrate the computation procedure and possible quantization error accumulation, not for bit exactness.
"""
from absl import app
Expand Down Expand Up @@ -88,7 +88,7 @@
_MULTISTEP_TEST_DATA = {
'init_hidden_state_vals': [0, 0],
'init_cell_state_vals': [0, 0],
'input_data': [0.2, 0.3, 0.2, 0.3, 0.2, 0.3], # three time steps
'input_data': [0.2, 0.3, 0.2, 0.3, 0.2, 0.3], # three time steps
'hidden_state_range': (-0.5, 0.7),
'cell_state_range': [-8, 8],
'input_data_range': [-1, 1]
Expand Down