Skip to content

MTMD: Fix Qwen3-tts-0.6b #28231

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ngxson merged 2 commits into from
Sep 2, 2026
Merged

MTMD: Fix Qwen3-tts-0.6b #28231

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
17bcb2d
mtmd: load the qwen3-tts code predictor proj_in as optional
ServeurpersoCom Sep 2, 2026
e5cf5d9
mtmd: keep the qwen3-tts code predictor ffn_down in F32
ServeurpersoCom Sep 2, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions conversion/qwen3tts.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -276,6 +276,10 @@ def tensor_force_quant(self, name, new_name, bid, n_dims):
# ConvTranspose1d kernels: only F16/F32 are implemented, no BF16
if new_name.endswith(".conv.weight") and (".up.blk." in new_name or ".dac.blk." in new_name):
return gguf.GGMLQuantizationType.F32
# the code predictor FFN intermediate peaks around 1.5e5, above the F16 range, and mul_mat
# casts its input to the weight type
if new_name.startswith("a.gen.code.blk.") and new_name.endswith(".ffn_down.weight"):
return gguf.GGMLQuantizationType.F32
return super().tensor_force_quant(name, new_name, bid, n_dims)

@classmethod
Expand Down
6 changes: 3 additions & 3 deletions tools/mtmd/clip.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2988,9 +2988,9 @@ struct clip_model_loader {
} break;
case PROJECTOR_TYPE_QWEN3TTS_GEN:
{
// code_predictor
model.gen_code_proj_in_w = get_tensor(string_format(TN_A_GEN_CODE_PROJ_IN, "weight"));
model.gen_code_proj_in_b = get_tensor(string_format(TN_A_GEN_CODE_PROJ_IN, "bias"));
// code_predictor, proj_in is absent when the talker and the predictor share the hidden size
model.gen_code_proj_in_w = get_tensor(string_format(TN_A_GEN_CODE_PROJ_IN, "weight"), false);
model.gen_code_proj_in_b = get_tensor(string_format(TN_A_GEN_CODE_PROJ_IN, "bias"), false);
model.gen_code_embd_w = get_tensor(string_format(TN_A_GEN_CODE_EMBD, "weight"));
model.gen_code_head_w = get_tensor(string_format(TN_A_GEN_CODE_HEAD, "weight"));
model.gen_code_out_embd_w = get_tensor(string_format(TN_A_GEN_CODE_OUT_EMBD, "weight"));
Expand Down
Loading