Add dtw preset for large-v3-turbo#2481
Merged
ggerganov merged 1 commit intoggml-org:masterfrom Oct 15, 2024
rotemdan:add-large-v3-turbo-dtw-preset
Merged
Add dtw preset for large-v3-turbo#2481ggerganov merged 1 commit intoggml-org:masterfrom rotemdan:add-large-v3-turbo-dtw-preset
ggerganov merged 1 commit intoggml-org:masterfrom
rotemdan:add-large-v3-turbo-dtw-preset
Conversation
ggerganov
approved these changes
Oct 15, 2024
bygreencn
added a commit
to bygreencn/whisper.cpp
that referenced
this pull request
Oct 21, 2024
* ggerganov/master: readme : update links and make commands (ggml-org#2489) ruby : fix bindings (ggml-org#2484) readme : add Vulkan notice (ggml-org#2488) make : fix GGML_VULKAN=1 build (ggml-org#2485) whisper : add dtw preset for large-v3-turbo (ggml-org#2481) convert : handle max_target_positions (ggml-org#2477) readme : update the Quick Start section (ggml-org#2475) whisper : add OpenVINO init with state (ggml-org#2464)
adutilleul
pushed a commit
to adutilleul/whisper.cpp
that referenced
this pull request
Nov 19, 2024
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See issue #2480 for more information.
The reference alignment heads for
large-v3-turboare:(extracted from reference Python implementation, see #2480 for more details)
Which, for
n_text_head = 20should imply:(Note: not 100% sure - it's an extrapolation based on how other
whisper.cppalignment head indice lists relate to the reference)I added a new preset:
{ WHISPER_AHEADS_LARGE_V3_TURBO, { 6, g_aheads_large_v3_turbo } },I did a few tests (in Echogarden) and it did seem to work correctly. The timing accuracy seemed consistent with other models.
Note: I named the preset
large.v3.turboto try to be consistent with how periods are used in all other presets. If you prefer something else likelarge.v3-turbothen change it.