tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-04 18:16:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

mtmd : add ultravox audio input (#13623)

Browse Source 
* convert ok, load ok

* warmup ok

* test

* still does not work?

* fix padding

* temporary give up

* fix merge conflict

* build_ultravox()

* rm test

* fix merge conflict

* add necessary mtmd APIs

* first working version (only 4s of audio)

* will this monster compile?

* fix compile

* please compile

* fPIC

* fix windows

* various fixes

* clean up audio_helpers

* fix conversion

* add some debug stuff

* long audio input ok

* adapt the api

* add --audio arg

* final touch UX

* add miniaudio to readme

* fix typo

* refactor kv metadata

* mtmd_default_marker()
This commit is contained in:
Xuan-Son Nguyen
2025-05-22 20:42:48 +02:00
committed by GitHub
parent ab86335760
commit 797990c4bc
21 changed files with 95401 additions and 259 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
gguf-py/gguf/gguf_writer.py
View File

@ -936,12 +936,18 @@ class GGUFWriter:
# for vision models
def add_clip_has_vision_encoder(self, value: bool) -> None:
self.add_bool(Keys.Clip.HAS_VISION_ENCODER, value)
def add_clip_has_audio_encoder(self, value: bool) -> None:
self.add_bool(Keys.Clip.HAS_AUDIO_ENCODER, value)
def add_clip_projector_type(self, value: str) -> None:
self.add_string(Keys.Clip.PROJECTOR_TYPE, value)
def add_vision_projection_dim(self, value: int) -> None:
self.add_uint32(Keys.ClipVision.PROJECTION_DIM, value)
def add_vision_has_vision_encoder(self, value: bool) -> None:
self.add_bool(Keys.ClipVision.HAS_VISION_ENCODER, value)
def add_vision_patch_size(self, value: int) -> None:
self.add_uint32(Keys.ClipVision.PATCH_SIZE, value)
@ -957,9 +963,6 @@ class GGUFWriter:
def add_vision_head_count(self, value: int) -> None:
self.add_uint32(Keys.ClipVision.Attention.HEAD_COUNT, value)
def add_vision_projector_type(self, value: str) -> None:
self.add_string(Keys.ClipVision.PROJECTOR_TYPE, value)
def add_vision_attention_layernorm_eps(self, value: float) -> None:
self.add_float32(Keys.ClipVision.Attention.LAYERNORM_EPS, value)
@ -987,6 +990,32 @@ class GGUFWriter:
def add_vision_n_wa_pattern(self, value: int) -> None:
self.add_uint32(Keys.ClipVision.N_WA_PATTERN, value)
# audio models
def add_audio_projection_dim(self, value: int) -> None:
self.add_uint32(Keys.ClipAudio.PROJECTION_DIM, value)
def add_audio_embedding_length(self, value: int) -> None:
self.add_uint32(Keys.ClipAudio.EMBEDDING_LENGTH, value)
def add_audio_feed_forward_length(self, value: int) -> None:
self.add_uint32(Keys.ClipAudio.FEED_FORWARD_LENGTH, value)
def add_audio_block_count(self, value: int) -> None:
self.add_uint32(Keys.ClipAudio.BLOCK_COUNT, value)
def add_audio_head_count(self, value: int) -> None:
self.add_uint32(Keys.ClipAudio.Attention.HEAD_COUNT, value)
def add_audio_attention_layernorm_eps(self, value: float) -> None:
self.add_float32(Keys.ClipAudio.Attention.LAYERNORM_EPS, value)
def add_audio_num_mel_bins(self, value: int) -> None:
self.add_uint32(Keys.ClipAudio.NUM_MEL_BINS, value)
def add_audio_stack_factor(self, value: int) -> None:
self.add_uint32(Keys.ClipAudio.Projector.STACK_FACTOR, value)
def _pack(self, fmt: str, value: Any, skip_pack_prefix: bool = False) -> bytes:
pack_prefix = ''
if not skip_pack_prefix: