sleepy/llama.cpp
1
0
Fork 0
Code Issues 11 Pull Requests Actions 107 Packages Projects Releases Wiki Activity

mtmd : Adding support for Nvidia Music Flamingo Model (#18470)

Browse Source 
* Inital commit, debugging q5_k_s quant

* Made hf_to_gguf extend whisper to reduce code duplication

* addressed convert_hf_to_gguf pull request issue

---------

Co-authored-by: Henry D <henrydorsey147@gmail.com>
This commit is contained in:
Henry147147
2025-12-31 06:13:23 -05:00
committed by GitHub
parent 9a6369bb60
commit 9b8329de7a
7 changed files with 49 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/mtmd/clip.cpp
+18 -1
View File
@@ -818,6 +818,7 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
case PROJECTOR_TYPE_VOXTRAL:
case PROJECTOR_TYPE_QWEN2A:
case PROJECTOR_TYPE_GLMA:
case PROJECTOR_TYPE_MUSIC_FLAMINGO:
{
builder = std::make_unique<clip_graph_whisper_enc>(ctx, img);
} break;
@@ -1176,6 +1177,7 @@ struct clip_model_loader {
case PROJECTOR_TYPE_QWEN2A:
case PROJECTOR_TYPE_GLMA:
case PROJECTOR_TYPE_VOXTRAL:
case PROJECTOR_TYPE_MUSIC_FLAMINGO:
{
bool require_stack = model.proj_type == PROJECTOR_TYPE_ULTRAVOX ||
model.proj_type == PROJECTOR_TYPE_VOXTRAL ||
@@ -1576,6 +1578,17 @@ struct clip_model_loader {
model.mm_1_w = get_tensor(string_format(TN_MM_AUDIO_MLP, 1, "weight"));
model.mm_2_w = get_tensor(string_format(TN_MM_AUDIO_MLP, 2, "weight"));
} break;
case PROJECTOR_TYPE_MUSIC_FLAMINGO:
{
model.conv1d_1_w = get_tensor(string_format(TN_CONV1D, 1, "weight"));
model.conv1d_1_b = get_tensor(string_format(TN_CONV1D, 1, "bias"));
model.conv1d_2_w = get_tensor(string_format(TN_CONV1D, 2, "weight"));
model.conv1d_2_b = get_tensor(string_format(TN_CONV1D, 2, "bias"));
model.mm_1_w = get_tensor(string_format(TN_MM_AUDIO_MLP, 1, "weight"));
model.mm_1_b = get_tensor(string_format(TN_MM_AUDIO_MLP, 1, "bias"));
model.mm_2_w = get_tensor(string_format(TN_MM_AUDIO_MLP, 2, "weight"));
model.mm_2_b = get_tensor(string_format(TN_MM_AUDIO_MLP, 2, "bias"));
} break;
case PROJECTOR_TYPE_INTERNVL:
{
model.mm_0_w = get_tensor(string_format(TN_MVLM_PROJ_MLP, 0, "weight"));
@@ -3031,6 +3044,7 @@ int clip_n_output_tokens(const struct clip_ctx * ctx, struct clip_image_f32 * im
case PROJECTOR_TYPE_VOXTRAL:
case PROJECTOR_TYPE_ULTRAVOX:
case PROJECTOR_TYPE_QWEN2A:
case PROJECTOR_TYPE_MUSIC_FLAMINGO:
{
n_patches = img->nx;
@@ -3403,6 +3417,7 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
case PROJECTOR_TYPE_ULTRAVOX:
case PROJECTOR_TYPE_LFM2:
case PROJECTOR_TYPE_VOXTRAL:
case PROJECTOR_TYPE_MUSIC_FLAMINGO:
case PROJECTOR_TYPE_JANUS_PRO:
case PROJECTOR_TYPE_COGVLM:
{
@@ -3526,6 +3541,7 @@ int clip_n_mmproj_embd(const struct clip_ctx * ctx) {
return ctx->model.projection->ne[1];
case PROJECTOR_TYPE_ULTRAVOX:
case PROJECTOR_TYPE_VOXTRAL:
case PROJECTOR_TYPE_MUSIC_FLAMINGO:
return ctx->model.mm_2_w->ne[1];
case PROJECTOR_TYPE_INTERNVL:
return ctx->model.mm_3_w->ne[1];
@@ -3587,7 +3603,8 @@ bool clip_has_whisper_encoder(const struct clip_ctx * ctx) {
return ctx->proj_type() == PROJECTOR_TYPE_ULTRAVOX
|| ctx->proj_type() == PROJECTOR_TYPE_QWEN2A
|| ctx->proj_type() == PROJECTOR_TYPE_GLMA
|| ctx->proj_type() == PROJECTOR_TYPE_VOXTRAL;
|| ctx->proj_type() == PROJECTOR_TYPE_VOXTRAL
|| ctx->proj_type() == PROJECTOR_TYPE_MUSIC_FLAMINGO;
}
bool clip_encode_float_image (struct clip_ctx * ctx, int n_threads, float * img, int h, int w, float * vec) {