]>
git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/log
Georgi Gerganov [Sun, 14 May 2023 16:06:45 +0000 (19:06 +0300)]
release : v1.4.2
Georgi Gerganov [Sun, 14 May 2023 15:56:46 +0000 (18:56 +0300)]
ggml : add AVX dot products
Georgi Gerganov [Sun, 14 May 2023 15:46:19 +0000 (18:46 +0300)]
talk-llama : fix build + sync latest llama.cpp
Jhen-Jie Hong [Sun, 14 May 2023 15:11:08 +0000 (23:11 +0800)]
readme : improve Core ML model conversion guidance (#915)
Georgi Gerganov [Sun, 14 May 2023 15:09:44 +0000 (18:09 +0300)]
coreml : support quantized model files
Georgi Gerganov [Sun, 14 May 2023 15:04:23 +0000 (18:04 +0300)]
ggml : sync latest ggml
- New Q4 and Q5 formats
- Various improvements
Rich Jones [Sun, 14 May 2023 14:54:57 +0000 (16:54 +0200)]
main : fix help for --no-timestamps arg (#908)
Georgi Gerganov [Sun, 14 May 2023 07:01:52 +0000 (10:01 +0300)]
extra : update ggml sync script
Jhen-Jie Hong [Sun, 14 May 2023 06:47:02 +0000 (14:47 +0800)]
whisper.objc : enable Core ML in example & fix segmentation fault (#910)
* coreml : update endcoder header import path
* coreml : force objc_arc in whisper-encoder.mm
* whisper.objc : create coreml/ group link
* whisper.objc : add coreml model link
* whisper.objc : update readme
* coreml : use -fobjc-arc for coreml/whisper-encoder.mm
* ci: create dummy .mlmodelc for pass ios build
* whisper.objc : update readme
---------
Co-authored-by: Georgi Gerganov <redacted>
Georgi Gerganov [Sun, 14 May 2023 06:42:19 +0000 (09:42 +0300)]
coreml : fix seg fault, double free (#919, #917, #899)
Georgi Gerganov [Tue, 9 May 2023 15:38:12 +0000 (18:38 +0300)]
coreml : fix memory leak (#899)
Jonathan Soo [Mon, 8 May 2023 18:08:09 +0000 (14:08 -0400)]
cmake : fix define used for COREML_ALLOW_FALLBACK (#893)
Luis Herrera [Mon, 8 May 2023 17:59:21 +0000 (12:59 -0500)]
talk-llama : only copy used KV cache in get / set state (#890)
---------
Co-authored-by: ejones <redacted>
Clifford Heath [Mon, 8 May 2023 17:58:36 +0000 (03:58 +1000)]
readme : add instructions on converting to GGML + "--no-config" to wget (#874)
ZaBlazzingZephyrus [Mon, 8 May 2023 17:45:53 +0000 (00:45 +0700)]
cmake : fix options disabling AVX and AVX2 flags (#885)
Georgi Gerganov [Thu, 4 May 2023 16:31:04 +0000 (19:31 +0300)]
cmake : add options to disable CPU flags (#860)
RelatedTitle [Wed, 3 May 2023 20:47:37 +0000 (14:47 -0600)]
ci : add cuBLAS build workflow and fix error causing lines in CMakeLists (#867)
* Add windows build with cuBLAS
* Remove error causing lines for cuBLAS on Windows
Vulcan [Wed, 3 May 2023 16:24:43 +0000 (21:54 +0530)]
readme : partial OpenCL GPU support via CLBlast (#863)
* ggml : CLBlast support as in llama.cpp
Building with CLBlast speeds up whisper.cpp ~2x on low end / older AMD APUs (CPU with integrated GPU) such as the A9.
Usage:
WHISPER_CLBLAST=1 make
* CMake/Makefile : CLBlast support as in llama.cpp
Building with CLBlast speeds up whisper.cpp ~2x on low end / older AMD APUs (CPU with integrated GPU) such as the A9.
Usage:
```
Makefile:
cd whisper.cpp
WHISPER_CLBLAST=1 make
CMake:
cd whisper.cpp ; mkdir build ; cd build
cmake -DWHISPER_CLBLAST=ON ..
make
```
* Update README.md
Added OpenCL Build Instructions
* Instruction: Partial OpenCL GPU support via CLBlast
Added build instructions and examples for Make and CMake to support OpenCL enabled GPUs.
Vulcan [Tue, 2 May 2023 19:50:32 +0000 (01:20 +0530)]
build : CLBlast support as in llama.cpp (#862)
* ggml : CLBlast support as in llama.cpp
Building with CLBlast speeds up whisper.cpp ~2x on low end / older AMD APUs (CPU with integrated GPU) such as the A9.
Usage:
WHISPER_CLBLAST=1 make
* CMake/Makefile : CLBlast support as in llama.cpp
Building with CLBlast speeds up whisper.cpp ~2x on low end / older AMD APUs (CPU with integrated GPU) such as the A9.
Usage:
```
Makefile:
cd whisper.cpp
WHISPER_CLBLAST=1 make
CMake:
cd whisper.cpp ; mkdir build ; cd build
cmake -DWHISPER_CLBLAST=ON ..
make
```
Georgi Gerganov [Tue, 2 May 2023 18:47:12 +0000 (21:47 +0300)]
ggml : fix 32-bit ARM build + quantization
Georgi Gerganov [Tue, 2 May 2023 18:23:54 +0000 (21:23 +0300)]
ggml : sync ggml (clBLAST + tensor names)
Luis Herrera [Tue, 2 May 2023 17:05:27 +0000 (12:05 -0500)]
talk-llama : fix session prompt load (#854)
CRD716 [Tue, 2 May 2023 16:51:52 +0000 (11:51 -0500)]
whisper : add detect-language mode (#853)
* add detectlanguage flag
* renaming and help
* no idea why that last one didn't commit
* run language detection if dl is set
* help message fix
* various fixes
* fix quitting
* fix language being english on print
Luis Herrera [Mon, 1 May 2023 17:18:10 +0000 (12:18 -0500)]
talk-llama : add --session support (#845)
* feat: adding session support
* readme: adding --session info in examples/talk-llama
* llama: adding session fixes
* readme: updating session doc
* talk-llama: update the value of need_to_save_session to true in order to save the session in the subsequent interaction
* talk-llama: adding missing function which updates session_tokens
Georgi Gerganov [Mon, 1 May 2023 11:44:39 +0000 (14:44 +0300)]
bench : improve benchmarks
Georgi Gerganov [Mon, 1 May 2023 07:03:56 +0000 (10:03 +0300)]
whisper : add memory sizes for Q8_0 (close #846)
Baffin Lee [Mon, 1 May 2023 06:28:05 +0000 (14:28 +0800)]
whisper.wasm : fix typo in readme (#832)
Georgi Gerganov [Sun, 30 Apr 2023 19:57:42 +0000 (22:57 +0300)]
release : v1.4.1
Georgi Gerganov [Sun, 30 Apr 2023 19:50:04 +0000 (22:50 +0300)]
whisper : fix quantize bug (#842)
* whisper : debug
* whisper : fix bug during quantization
Georgi Gerganov [Sun, 30 Apr 2023 19:27:30 +0000 (22:27 +0300)]
ggml : fix UB (int << 31)
Georgi Gerganov [Sun, 30 Apr 2023 16:23:37 +0000 (19:23 +0300)]
release : v1.4.0
Georgi Gerganov [Sun, 30 Apr 2023 16:12:49 +0000 (19:12 +0300)]
examples : fix + refactor Levenshtein distance
Georgi Gerganov [Sun, 30 Apr 2023 15:51:57 +0000 (18:51 +0300)]
whisper : add integer quantization support (#540)
* whisper : add integer quantization support
* examples : add common-ggml + prepare to add "quantize" tool
* whisper : quantization tool ready
* whisper : fix F32 support
* whisper : try to fix shared lib linkage
* wasm : update quantized models to Q5
* bench.wasm : remove "medium" button
* bench.wasm : fix custom model button
* ggml : add Q5_0 and Q5_1 WASM SIMD
* wasm : add quantized models to all WASM examples
* wasm : bump DB version number to 2
* talk-llama : update example to latest llama.cpp
* node : increase test timeout to 10s
* readme : add information for model quantization
* wasm : add links to other examples
Georgi Gerganov [Sun, 30 Apr 2023 09:14:33 +0000 (12:14 +0300)]
whisper : add GPU support via cuBLAS (#834)
* make : add WHISPER_CUBLAS
* make : fix CUBLAS build
* whisper : disable Flash Attention + adjust memory buffers
* whisper : remove old commented code
* readme : add cuBLAS instructions
* cmake : add WHISPER_CUBLAS option
* gitignore : ignore build-cublas
Georgi Gerganov [Sat, 29 Apr 2023 18:37:23 +0000 (21:37 +0300)]
ggml : fix WASM build
Georgi Gerganov [Sat, 29 Apr 2023 18:33:33 +0000 (21:33 +0300)]
ggml : fix 32-bit ARM NEON (#836)
* ggml : add support for 32-bit ARM
* ggml : fix
* ggml : fix
Georgi Gerganov [Sat, 29 Apr 2023 18:14:09 +0000 (21:14 +0300)]
ggml : use vzip instead of vuzp for consistency
Georgi Gerganov [Sat, 29 Apr 2023 17:21:25 +0000 (20:21 +0300)]
ggml : fix WASM build
Georgi Gerganov [Sat, 29 Apr 2023 16:30:22 +0000 (19:30 +0300)]
ggml : sync with ggml repo (warning fixes + asserts)
Thijs Raymakers [Sat, 29 Apr 2023 15:55:37 +0000 (17:55 +0200)]
whisper : use correct seek_end when offset is used (#833)
Whenever an `offset_ms` is provided, the value of `seek_end` is
calculated incorrectly. This causes Whisper to keep transcribing
after the end of the file.
The current behavior looks like
```
[00:34:40.000 --> 00:34:47.000] This is an example audio file.
[00:34:47.000 --> 00:34:49.000] The text has been redacted
[00:34:49.000 --> 00:34:51.000] This is the end of the audio.
[00:34:51.000 --> 00:34:52.000] ***
[00:34:52.000 --> 00:34:53.000] ***
[00:34:53.000 --> 00:34:54.000] ***
[00:34:55.000 --> 00:34:56.000] ***
...
```
The expected behavior should be
```
[00:34:40.000 --> 00:34:47.000] This is an example audio file.
[00:34:47.000 --> 00:34:49.000] The text has been redacted
[00:34:49.000 --> 00:34:51.000] This is the end of the audio.
- end of program -
```
This commit changes the calculation of the `seek_end` variable to
only add `seek_start` if a custom `duration_ms` is provided.
Otherwise, it defaults to the end of the file.
Signed-off-by: Thijs Raymakers <redacted>
Georgi Gerganov [Sat, 29 Apr 2023 09:32:18 +0000 (12:32 +0300)]
tests : add "threads" to run-tests.sh
Georgi Gerganov [Sat, 29 Apr 2023 09:32:05 +0000 (12:32 +0300)]
extra : add sync-ggml.sh script
Georgi Gerganov [Sat, 29 Apr 2023 09:31:52 +0000 (12:31 +0300)]
ggml : sync latest ggml + llama.cpp updates (quantization)
Zollner [Sat, 29 Apr 2023 08:00:20 +0000 (16:00 +0800)]
whisper.android : add some tips (#816)
Georgi Gerganov [Sat, 29 Apr 2023 07:55:24 +0000 (10:55 +0300)]
build : add WHISPER_COREML_ALLOW_FALLBACK to make / CMake (#812)
Canis Lupus [Sat, 29 Apr 2023 07:49:02 +0000 (08:49 +0100)]
whisper : allow non-CoreML fallback when Core ML cannot be loaded (#812)
if the Core ML model cannot be loaded, continue without Core ML instead of
returning. This allows a single build to transcribe using Core ML models
where available, and regular models when not.
Georgi Gerganov [Sat, 29 Apr 2023 07:42:14 +0000 (10:42 +0300)]
whisper : fix bug from previous commit
Georgi Gerganov [Sat, 29 Apr 2023 07:36:50 +0000 (10:36 +0300)]
whisper : avoid designated initializers
AsukaMinato [Sat, 29 Apr 2023 07:06:25 +0000 (16:06 +0900)]
minor : improve C++ and Python style (#768)
* use some STL functions
* use self.field than setattr, use pathlib.Path
* recover some format
* const some iter
* Keep the original
* 2 space
Georgi Gerganov [Fri, 28 Apr 2023 19:41:29 +0000 (22:41 +0300)]
readme : add logo
Laytan Laats [Sun, 23 Apr 2023 16:01:59 +0000 (18:01 +0200)]
main : escape quotes in csv output (#815)
Taras Glek [Sun, 23 Apr 2023 14:00:30 +0000 (17:00 +0300)]
stream : flush upon finishing inference (#811)
Philipp Zabel [Sun, 23 Apr 2023 13:52:52 +0000 (15:52 +0200)]
examples : add missing #include <cstdint> (#798)
common.cpp uses uint8_t and uint64_t, which are defined in <cstdint>.
Tauseef Mohiuddin [Sun, 23 Apr 2023 13:47:30 +0000 (08:47 -0500)]
main : update escape_double_quotes() function (#776)
Updated the escape_double_quotes() function such that the function now escapes both double quotes and backslashes in the input string.
Changes Made:
- Renamed the function to escape_quotes_and_backslashes
- Modified the condition in the first loop to increment the value of 'escaped_length' for both double quotes and backslashes.
- Modified the condition in second loop to add a backslash before the current character if it is a double quote or a backslash.
Resolves: #769
Georgi Gerganov [Sat, 15 Apr 2023 14:30:44 +0000 (17:30 +0300)]
release : v1.3.0
Georgi Gerganov [Sat, 15 Apr 2023 14:18:43 +0000 (17:18 +0300)]
whisper : pad audio instead of spectrogram (#579)
Also, fallback only if more temperatures are available and if we are
at least 3 seconds before the end of the audio
Georgi Gerganov [Sat, 15 Apr 2023 13:04:07 +0000 (16:04 +0300)]
whisper : restore decoder temperature fallbacks
I disabled this because there were many complaints about slow decoding.
The current implementation does not allow batching the decoders when
using the "best of" or "beam size" parameters, so the decoding time is
proportional to the number of decoders, which is obviously not great.
However, now there are even more complaints about wrong decodings and
repetition.
So, making a compromise by re-enabling the fallbacks, but defaulting to
just 2 "best of" / "beam size" decoders. Also, the temperature step is
increased from 0.2 to 0.4 - i.e. from maximum of 5 fallbacks to maximum
of 2.
Also, the stream example now has fallbacks enabled by default.
close #471 #477 #508 #612 #719 #731
Jhen-Jie Hong [Sat, 15 Apr 2023 11:21:58 +0000 (19:21 +0800)]
ggml, ci : fix build on whisper.android (ARM_NEON) + add CI (#764)
* ggml : fix undefined symbol by remove inline handle
* ggml : make own ggml_aligned_malloc function
* ci: add ios/android build
Georgi Gerganov [Sat, 15 Apr 2023 11:18:46 +0000 (14:18 +0300)]
whisper : slightly faster Log Mel computation + n-1 FFT threads (#568)
Georgi Gerganov [Sat, 15 Apr 2023 10:30:36 +0000 (13:30 +0300)]
readme : fix link
Georgi Gerganov [Sat, 15 Apr 2023 10:30:07 +0000 (13:30 +0300)]
readme : add usage instructions for Core ML
Georgi Gerganov [Sat, 15 Apr 2023 10:21:27 +0000 (13:21 +0300)]
whisper : add Core ML support (#566)
* coreml : use Core ML encoder inference
* coreml : simlpify whisper_encode + log messages
* whisper : resolve rebase conflicts
* coreml : add scripts for CoreML model generation
* bench-all : recognize COREML flag
Maximiliano Levi [Fri, 14 Apr 2023 19:35:34 +0000 (16:35 -0300)]
whisper : do not launch log_mel threads when n_thread is 1 (#763)
AfryMask [Fri, 14 Apr 2023 17:35:03 +0000 (01:35 +0800)]
whisper : fix the bug related to word splitting errors in the "tokenize" function. (#760)
Co-authored-by: AfryMask <redacted>
Aaron Taylor [Fri, 14 Apr 2023 17:24:00 +0000 (13:24 -0400)]
readme : add SwiftWhisper to listed bindings (#755)
Georgi Gerganov [Fri, 14 Apr 2023 17:13:47 +0000 (20:13 +0300)]
gitignore : add .test
Bader-eddine Ouaich [Fri, 14 Apr 2023 17:05:56 +0000 (17:05 +0000)]
whisper : fix potential memory leaks (#740)
* fix potential memory leak if whisper_init_state failed
* fix potential memory leak if gpt2_init failed
Anton Kostin [Fri, 14 Apr 2023 17:04:42 +0000 (00:04 +0700)]
license : update year (#739)
GitAritron [Fri, 14 Apr 2023 17:03:16 +0000 (20:03 +0300)]
whisper : fix typos in whisper.h (#737)
Fixed a couple of typos (in comments, so nothing major). Keep up the great work 😄
Ali Alameh [Fri, 14 Apr 2023 17:02:18 +0000 (20:02 +0300)]
stream : support language auto-detect (#501)
#445 fix Language auto-detect "auto" flag does not work using the stream tool
Alex Evgrashin [Fri, 14 Apr 2023 16:59:44 +0000 (19:59 +0300)]
readme : add unity bindings (#733)
DGdev91 [Fri, 14 Apr 2023 16:53:58 +0000 (18:53 +0200)]
talk, talk-llama : add basic example script for eleven-labs tts (#728)
Ivan Gorin [Fri, 14 Apr 2023 16:50:39 +0000 (19:50 +0300)]
models : change convert-pt-to-ggml to use .tiktoken tokenizer files (#725)
LittleLoli [Fri, 14 Apr 2023 16:36:38 +0000 (00:36 +0800)]
cmake : add msvc compiler args /utf-8 fix error C3688 (#721)
* force msvc compiler use utf-8 encode
* only enable on msvc
Maciek [Fri, 14 Apr 2023 16:36:09 +0000 (18:36 +0200)]
talk-llama : correct default speak.sh path (#720)
There is `speak.sh` file in `./examples/talk-llama` as described in README.
However `./examples/talk/speak.sh` is used in `talk-llama.cpp`, this commit corrects that.
LittleLoli [Fri, 14 Apr 2023 16:35:33 +0000 (00:35 +0800)]
main : add lrc output support (#718)
* add lrc output support.
* fix wrong comment
Sam [Fri, 14 Apr 2023 16:33:06 +0000 (16:33 +0000)]
readme : make the quick start instructions clearer. (#716)
Users wanting to make use of this implementation of the whisper model with no prior knowledge of C/C++ may download the Whisper model but fail to use of the "make" command as specified given that they forgot or didn't know they needed to clone the repository first. Hope this modification clears things up.
duthils [Fri, 14 Apr 2023 16:31:51 +0000 (12:31 -0400)]
make : disable avx in case f16c is not available (#706)
Why:
* ggml.c does not support AVX without F16C
bocytko [Fri, 14 Apr 2023 16:25:23 +0000 (18:25 +0200)]
readme : add shell command example for --print-colors (#710)
The section of the readme file explaining `--print-colors` includes only a screenshot with directories that are inconsistent with other examples. This commit adds an example shell command, consistent with the remaining examples.
Georgi Gerganov [Fri, 14 Apr 2023 16:20:39 +0000 (19:20 +0300)]
ggml : sync latest ggml
Georgi Gerganov [Fri, 14 Apr 2023 16:16:34 +0000 (19:16 +0300)]
whisper : fix bug in prompt processing (close #705)
Was dereferencing a dangling pointer
Brian Murray [Fri, 14 Apr 2023 15:52:10 +0000 (09:52 -0600)]
go : exposed various parts to the Go Interface (#697)
novag [Fri, 14 Apr 2023 10:34:20 +0000 (12:34 +0200)]
ggml : fix q4_1 dot product types (#759)
Co-authored-by: Georgi Gerganov <redacted>
Georgi Gerganov [Thu, 13 Apr 2023 15:53:44 +0000 (18:53 +0300)]
ggml : sync latest changes from ggml and llama.cpp
Georgi Gerganov [Mon, 10 Apr 2023 20:18:29 +0000 (23:18 +0300)]
ggml : fix WASM build
Georgi Gerganov [Mon, 10 Apr 2023 20:09:15 +0000 (23:09 +0300)]
talk-llama : increase context to 2048
Georgi Gerganov [Mon, 10 Apr 2023 19:59:13 +0000 (22:59 +0300)]
talk-llama : update to latest llama.cpp (improved performance)
Georgi Gerganov [Mon, 10 Apr 2023 19:28:54 +0000 (22:28 +0300)]
ggml : backport llama.cpp updates (close #709)
- About x2 overall performance improvement on Apple Silicon
- Results should now be the same for different number of threads (not
tested)
pajowu [Thu, 30 Mar 2023 17:29:29 +0000 (19:29 +0200)]
whisper : add progress callback (#600)
Zigfrid Zvezdin [Thu, 30 Mar 2023 04:51:33 +0000 (01:51 -0300)]
misc : typo (#688)
InconsolableCellist [Wed, 29 Mar 2023 21:10:20 +0000 (15:10 -0600)]
talk-llama : fixing usage message for talk-llama (#687)
"-ml" instead of "-mg" for specifying the llama file
Georgi Gerganov [Wed, 29 Mar 2023 20:59:45 +0000 (23:59 +0300)]
main : add <cstring> header
Lucas Zanek [Wed, 29 Mar 2023 20:59:17 +0000 (17:59 -0300)]
whisper.addon : fixed test to new async implementation (#686)
* fixed blocking code on node addon
* modify the example to run async
* format
* added logic to see the whisper output
* added logic to see the whisper output
* removed extra function for more clean example
* fixed whisper test to new async implementation
be-next [Wed, 29 Mar 2023 20:38:33 +0000 (22:38 +0200)]
models : handle spaces and special characters in shell script paths (#677)
This commit modifies the `get_script_path` function to correctly handle
spaces and special characters in directory paths. The fix involves adding
double quotes around variables and commands where needed to ensure proper
parsing of paths with spaces and special characters.
Egor Egorov [Wed, 29 Mar 2023 20:26:39 +0000 (23:26 +0300)]
main : fix typo in JSON output (#648)
* typo in JSON output
* fix double quotes in JSON output
Jhen-Jie Hong [Wed, 29 Mar 2023 20:23:23 +0000 (04:23 +0800)]
whisper : add initial_prompt param (#645)
clach04 [Wed, 29 Mar 2023 20:11:35 +0000 (13:11 -0700)]
make : 32-bit ARM flags (#486)
* issue #470 - working 32-bit ARM
* Update Makefile
* Update Makefile
---------
Co-authored-by: Georgi Gerganov <redacted>
Jonno [Wed, 29 Mar 2023 20:04:38 +0000 (06:04 +1000)]
whisper.swiftui : update README.md (#682)
- Slight tweaks to README for improved comprehension.
Evan Jones [Wed, 29 Mar 2023 20:01:14 +0000 (16:01 -0400)]
talk-llama : add alpaca support (#668)
Georgi Gerganov [Tue, 28 Mar 2023 07:50:49 +0000 (10:50 +0300)]
whisper : bump "large" scratch buffer even mode (close #671)