mtmd: add chunks and fix preproc for qwen3a (#23073)

* mtmd: add chunks and fix preproc for qwen3a

* add attn_mask

* limit mtmd_chunk size (avoid blow up memory)

* correct audio tokens

* re-order the set_input case

* remove attn_mask
This commit is contained in:
Xuan-Son Nguyen
2026-05-15 19:32:47 +02:00
committed by GitHub
parent 8be1786707
commit 72e60f500d
7 changed files with 200 additions and 71 deletions
+9
View File
@@ -96,6 +96,15 @@ struct mtmd_audio_preprocessor_gemma4a : mtmd_audio_preprocessor {
mtmd_audio_cache cache;
};
struct mtmd_audio_preprocessor_qwen3a : mtmd_audio_preprocessor {
mtmd_audio_preprocessor_qwen3a(const clip_ctx * ctx) : mtmd_audio_preprocessor(ctx) {}
void initialize() override;
bool preprocess(const float * samples, size_t n_samples, std::vector<mtmd_audio_mel> & output) override;
private:
mtmd_audio_cache cache;
};
//
// streaming ISTFT - converts spectrogram frames back to audio one frame at a time
//