Skip to content

Fix Qwen3 MoE GGUF architecture mismatch #39976

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
ArthurZucker merged 5 commits into from
Aug 12, 2025
+9 −3
Merged

Fix Qwen3 MoE GGUF architecture mismatch #39976

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
14af3be
fix qwen3moe gguf architecture
ctcanbol Aug 7, 2025
b891987
Merge branch 'main' into main
MekkCyber Aug 7, 2025
344eaa3
Merge branch 'main' into main
MekkCyber Aug 7, 2025
f43d297
Fix Qwen3Moe GGUF loading
jusjinuk Aug 10, 2025
fec7841
Merge pull request #1 from jusjinuk/main
ctcanbol Aug 10, 2025
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
3 changes: 2 additions & 1 deletion src/transformers/integrations/ggml.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,13 +102,14 @@
"attention.layer_norm_rms_epsilon": "rms_norm_eps",
"vocab_size": "vocab_size",
},
"qwen3moe": {
"qwen3_moe": {
"context_length": "max_position_embeddings",
"block_count": "num_hidden_layers",
"feed_forward_length": "intermediate_size",
"embedding_length": "hidden_size",
"rope.dimension_count": None,
"rope.freq_base": "rope_theta",
"attention.key_length": "head_dim",
"attention.head_count": "num_attention_heads",
"attention.head_count_kv": "num_key_value_heads",
"attention.layer_norm_rms_epsilon": "rms_norm_eps",
Expand Down
9 changes: 7 additions & 2 deletions src/transformers/modeling_gguf_pytorch_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -246,6 +246,7 @@ def process(self, weights, name, **kwargs):
TENSOR_PROCESSORS = {
"llama": LlamaTensorProcessor,
"qwen2moe": Qwen2MoeTensorProcessor,
"qwen3moe": Qwen2MoeTensorProcessor,
"bloom": BloomTensorProcessor,
"t5": T5TensorProcessor,
"t5encoder": T5TensorProcessor,
Expand Down Expand Up @@ -295,6 +296,8 @@ def get_gguf_hf_weights_map(
model_type = "command-r"
elif model_type == "qwen2_moe":
model_type = "qwen2moe"
elif model_type == "qwen3_moe":
model_type = "qwen3moe"
elif model_type == "gemma3_text":
model_type = "gemma3"
arch = None
Expand All @@ -316,8 +319,8 @@ def get_gguf_hf_weights_map(
gguf_to_hf_name_map = {}
state_dict = hf_model.state_dict()
for hf_name in state_dict:
# An exception for qwen2moe model, where the expert layers are packed
if model_type == "qwen2moe" and "mlp.experts." in hf_name:
# An exception for qwen2moe/qwen3moe model, where the expert layers are packed
if model_type in ("qwen2moe", "qwen3moe") and "mlp.experts." in hf_name:
hf_name = re.sub(r"mlp.experts.\d+.", "mlp.experts.", hf_name)

name, suffix = hf_name, ""
Expand Down Expand Up @@ -391,6 +394,8 @@ def load_gguf_checkpoint(gguf_checkpoint_path, return_tensors=False, model_to_lo

if "qwen2moe" in architecture:
updated_architecture = "qwen2_moe"
elif "qwen3moe" in architecture:
updated_architecture = "qwen3_moe"

# For stablelm architecture, we need to set qkv_bias and use_parallel_residual from tensors
# If `qkv_bias=True`, qkv_proj with bias will be present in the tensors
Expand Down