ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

[pre-commit.ci] auto fixes from pre-commit.com hooks

Browse source 
for more information, see https://pre-commit.ci
This commit is contained in:
pre-commit-ci[bot] 2026-01-08 12:11:11 +00:00 • committed by Daniel Han
commit 13fced1b8a
2 changed files with 69 additions and 65 deletions
Download patch file Download diff file Expand all files Collapse all files

4
unsloth/models/loader.py
View file

@ -619,7 +619,9 @@ class FastLanguageModel(FastLlamaModel):
model_type == "qwen3"
or model_type == "qwen3_moe"
or model_type == "qwen3_omni_moe"
or (isinstance(model_type, (list, tuple)) and "qwen3_omni_moe" in model_type)
or (
isinstance(model_type, (list, tuple)) and "qwen3_omni_moe" in model_type
)
):
if not SUPPORTS_QWEN3 or not SUPPORTS_QWEN3_MOE:
raise ImportError(

130
unsloth/models/qwen3_moe.py
View file

@ -262,6 +262,7 @@ def _patch_qwen3_omni_model_class(model_class):
Patch Qwen3OmniMoeForConditionalGeneration with required methods for training.
The model delegates to its thinker submodule for actual forward pass.
"""
# Add embedding accessor methods that delegate to thinker
def get_input_embeddings(self):
return self.thinker.get_input_embeddings()
@ -278,48 +279,48 @@ def _patch_qwen3_omni_model_class(model_class):
# Add forward method that delegates to thinker for training
def forward(
self,
input_ids=None,
input_features=None,
pixel_values=None,
pixel_values_videos=None,
image_grid_thw=None,
video_grid_thw=None,
attention_mask=None,
feature_attention_mask=None,
audio_feature_lengths=None,
position_ids=None,
past_key_values=None,
inputs_embeds=None,
rope_deltas=None,
labels=None,
use_cache=None,
output_router_logits=None,
use_audio_in_video=None,
cache_position=None,
video_second_per_grid=None,
input_ids = None,
input_features = None,
pixel_values = None,
pixel_values_videos = None,
image_grid_thw = None,
video_grid_thw = None,
attention_mask = None,
feature_attention_mask = None,
audio_feature_lengths = None,
position_ids = None,
past_key_values = None,
inputs_embeds = None,
rope_deltas = None,
labels = None,
use_cache = None,
output_router_logits = None,
use_audio_in_video = None,
cache_position = None,
video_second_per_grid = None,
**kwargs,
):
"""Forward method that delegates to the thinker for training."""
return self.thinker.forward(
input_ids=input_ids,
input_features=input_features,
pixel_values=pixel_values,
pixel_values_videos=pixel_values_videos,
image_grid_thw=image_grid_thw,
video_grid_thw=video_grid_thw,
attention_mask=attention_mask,
feature_attention_mask=feature_attention_mask,
audio_feature_lengths=audio_feature_lengths,
position_ids=position_ids,
past_key_values=past_key_values,
inputs_embeds=inputs_embeds,
rope_deltas=rope_deltas,
labels=labels,
use_cache=use_cache,
output_router_logits=output_router_logits,
use_audio_in_video=use_audio_in_video,
cache_position=cache_position,
video_second_per_grid=video_second_per_grid,
input_ids = input_ids,
input_features = input_features,
pixel_values = pixel_values,
pixel_values_videos = pixel_values_videos,
image_grid_thw = image_grid_thw,
video_grid_thw = video_grid_thw,
attention_mask = attention_mask,
feature_attention_mask = feature_attention_mask,
audio_feature_lengths = audio_feature_lengths,
position_ids = position_ids,
past_key_values = past_key_values,
inputs_embeds = inputs_embeds,
rope_deltas = rope_deltas,
labels = labels,
use_cache = use_cache,
output_router_logits = output_router_logits,
use_audio_in_video = use_audio_in_video,
cache_position = cache_position,
video_second_per_grid = video_second_per_grid,
**kwargs,
)
@ -368,18 +369,18 @@ def _patch_qwen3_omni_attention_modules(model):
def _attention_wrapper(
self,
hidden_states,
position_embeddings=None,
attention_mask=None,
past_key_values=None,
position_embeddings = None,
attention_mask = None,
past_key_values = None,
**kwargs,
):
# FALLBACK: If caching is used (Inference), use original code
if past_key_values is not None:
return self._original_forward(
hidden_states,
position_embeddings=position_embeddings,
attention_mask=attention_mask,
past_key_values=past_key_values,
position_embeddings = position_embeddings,
attention_mask = attention_mask,
past_key_values = past_key_values,
**kwargs,
)
@ -387,9 +388,9 @@ def _patch_qwen3_omni_attention_modules(model):
return Qwen3Attention_fast_forward(
self,
hidden_states,
attention_mask=attention_mask,
position_embeddings=position_embeddings,
past_key_value=None,
attention_mask = attention_mask,
position_embeddings = position_embeddings,
past_key_value = None,
**kwargs,
)
@ -412,17 +413,17 @@ def _patch_qwen3_omni_attention_modules(model):
class FastQwen3OmniMoeModel(FastQwen3MoeModel):
@staticmethod
def from_pretrained(
model_name="Qwen/Qwen3-Omni-30B-A3B-Instruct",
max_seq_length=4096,
dtype=None,
load_in_4bit=True,
token=None,
device_map="sequential",
rope_scaling=None,
fix_tokenizer=True,
model_patcher=None,
tokenizer_name=None,
trust_remote_code=False,
model_name = "Qwen/Qwen3-Omni-30B-A3B-Instruct",
max_seq_length = 4096,
dtype = None,
load_in_4bit = True,
token = None,
device_map = "sequential",
rope_scaling = None,
fix_tokenizer = True,
model_patcher = None,
tokenizer_name = None,
trust_remote_code = False,
**kwargs,
):
# Import the model class
@ -445,11 +446,12 @@ class FastQwen3OmniMoeModel(FastQwen3MoeModel):
if load_in_4bit:
from transformers import BitsAndBytesConfig
model_kwargs["quantization_config"] = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_compute_dtype=dtype,
bnb_4bit_use_double_quant=True,
bnb_4bit_quant_type="nf4",
load_in_4bit = True,
bnb_4bit_compute_dtype = dtype,
bnb_4bit_use_double_quant = True,
bnb_4bit_quant_type = "nf4",
)
# Merge any additional kwargs
@ -464,8 +466,8 @@ class FastQwen3OmniMoeModel(FastQwen3MoeModel):
# Load processor/tokenizer
processor = AutoProcessor.from_pretrained(
model_name,
trust_remote_code=trust_remote_code,
token=token,
trust_remote_code = trust_remote_code,
token = token,
)
# Patch attention and MLP modules for faster training