Skip to content

[BugFix] Fix load_weights error when loading HunyuanImage3.0#1598

Merged
hsliuustc0106 merged 1 commit intovllm-project:mainfrom
Semmer2:load_weight_error
Mar 2, 2026
Merged

[BugFix] Fix load_weights error when loading HunyuanImage3.0#1598
hsliuustc0106 merged 1 commit intovllm-project:mainfrom
Semmer2:load_weight_error

Conversation

@Semmer2
Copy link
Copy Markdown
Contributor

@Semmer2 Semmer2 commented Mar 2, 2026 •
edited
Loading

Move some submodule load weights code of HunyuanImage3Pipeline to AutoWeightsLoader:load_weights, fix weights not initialized error.

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

DiffusersPipelineLoader:load_weights added strictly weights load gap verification. Which reports bug when loading HunyuanImage3.0 mode. Move some submodule loading code to AutoWeightsLoader:load_weights to fix this bug.

Test Plan

python examples/offline_inference/text_to_image/text_to_image.py --mode /data/HunyuanImage-3.0/ --prompt "A brown and white dog is running on the grass" --output output_image.png --num-inference-steps 50 --guidance-scale 5.0 --tensor-parallel-size 8 --seed 1234

Test Result

INFO 03-01 22:53:10 [omni.py:181] Initializing stages for model: /data/HunyuanImage-3.0/
INFO 03-01 22:53:10 [omni.py:313] No omni_master_address provided, defaulting to localhost (127.0.0.1)
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
INFO 03-01 22:53:10 [config.py:379] Replacing legacy 'type' key with 'rope_type'
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
INFO 03-01 22:53:10 [config.py:379] Replacing legacy 'type' key with 'rope_type'
INFO 03-01 22:53:10 [initialization.py:35] No OmniTransferConfig provided
INFO 03-01 22:53:10 [omni.py:347] [Orchestrator] Loaded 1 stages
INFO 03-01 22:53:10 [omni.py:458] [Orchestrator] Waiting for 1 stages to initialize (timeout: 300s)
[Stage-0] INFO 03-01 22:53:32 [omni_stage.py:679] Starting stage worker with model: /data/HunyuanImage-3.0/
[Stage-0] INFO 03-01 22:53:32 [omni_stage.py:694] [Stage] Set VLLM_WORKER_MULTIPROC_METHOD=spawn
[Stage-0] INFO 03-01 22:53:32 [omni_stage.py:725] [Stage-0] ZMQ transport detected; disabling SHM IPC (shm_threshold_bytes set to maxsize)
[Stage-0] INFO 03-01 22:53:32 [omni_stage.py:85] Using sequential init locks (nvml_available=True, pid_host=False)
[Stage-0] INFO 03-01 22:53:34 [multiproc_executor.py:74] Starting server...
[Stage-0] INFO 03-01 22:53:57 [diffusion_worker.py:349] Worker 0 created result MessageQueue
[Stage-0] INFO 03-01 22:53:59 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:53:59 [vllm.py:689] Asynchronous scheduling is enabled.
[Stage-0] INFO 03-01 22:53:59 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:53:59 [vllm.py:689] Asynchronous scheduling is enabled.
[Stage-0] INFO 03-01 22:53:59 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:53:59 [vllm.py:689] Asynchronous scheduling is enabled.
[Stage-0] INFO 03-01 22:54:00 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:54:00 [vllm.py:689] Asynchronous scheduling is enabled.
[Stage-0] INFO 03-01 22:54:03 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:54:03 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:54:03 [vllm.py:689] Asynchronous scheduling is enabled.
[Stage-0] INFO 03-01 22:54:03 [vllm.py:689] Asynchronous scheduling is enabled.
[Stage-0] INFO 03-01 22:54:03 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:54:03 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=2048.
[Stage-0] INFO 03-01 22:54:03 [vllm.py:689] Asynchronous scheduling is enabled.
[Stage-0] INFO 03-01 22:54:03 [vllm.py:689] Asynchronous scheduling is enabled.
[Gloo] Rank 7[Gloo] Rank  is connected to 67 is connected to  peer ranks. 7Expected number of connected peer ranks is :  peer ranks. 7Expected number of connected peer ranks is : 
7
[Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank [Gloo] Rank 45 is connected to  is connected to 77 peer ranks.  peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 77

[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 7: Initialized device and distributed environment.
[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 6: Initialized device and distributed environment.
[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 0: Initialized device and distributed environment.
[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 1: Initialized device and distributed environment.
[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 2: Initialized device and distributed environment.
[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 4: Initialized device and distributed environment.
[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 5: Initialized device and distributed environment.
[Stage-0] INFO 03-01 22:54:03 [diffusion_worker.py:114] Worker 3: Initialized device and distributed environment.
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0[Gloo] Rank  peer ranks. [Gloo] Rank 0Expected number of connected peer ranks is : 0 is connected to 0 is connected to 0
0 peer ranks.  peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 00

[Gloo] Rank 0 is connected to [Gloo] Rank 00 peer ranks.  is connected to Expected number of connected peer ranks is : 00 peer ranks. 
Expected number of connected peer ranks is : 0[Gloo] Rank 
0[Gloo] Rank  is connected to 00 is connected to  peer ranks. 0Expected number of connected peer ranks is :  peer ranks. 0Expected number of connected peer ranks is : 
0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0[Gloo] Rank  is connected to 00 peer ranks.  is connected to Expected number of connected peer ranks is : 00[Gloo] Rank  peer ranks. 
Expected number of connected peer ranks is : 00 is connected to 
0[Gloo] Rank  peer ranks. 0[Gloo] Rank Expected number of connected peer ranks is :  is connected to 000 is connected to 
 peer ranks. 0Expected number of connected peer ranks is :  peer ranks. 0Expected number of connected peer ranks is : 
0
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 0: sp_group=[0], ulysses_group=[0], ring_group=[0]
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 7: sp_group=[7], ulysses_group=[7], ring_group=[7]
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 5: sp_group=[5], ulysses_group=[5], ring_group=[5]
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 6: sp_group=[6], ulysses_group=[6], ring_group=[6]
[Gloo] Rank 0 is connected to 0[Gloo] Rank  peer ranks. 0[Gloo] Rank Expected number of connected peer ranks is :  is connected to 000
 is connected to  peer ranks. 0Expected number of connected peer ranks is :  peer ranks. 0Expected number of connected peer ranks is : 
0
[Gloo] Rank 0[Gloo] Rank  is connected to 00 peer ranks.  is connected to Expected number of connected peer ranks is : 00 peer ranks. 
Expected number of connected peer ranks is : 0
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 4: sp_group=[4], ulysses_group=[4], ring_group=[4]
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 3: sp_group=[3], ulysses_group=[3], ring_group=[3]
[Gloo] Rank 0 is connected to 0 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 00 is connected to 
0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to [Gloo] Rank 0 peer ranks. 0Expected number of connected peer ranks is :  is connected to 00
 peer ranks. Expected number of connected peer ranks is : 0
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:575] Building SP subgroups from explicit sp_group_ranks (sp_size=1, ulysses=1, ring=1, use_ulysses_low=True).
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 1: sp_group=[1], ulysses_group=[1], ring_group=[1]
[Stage-0] INFO 03-01 22:54:03 [parallel_state.py:617] SP group details for rank 2: sp_group=[2], ulysses_group=[2], ring_group=[2]
[Gloo] Rank 0[Gloo] Rank  is connected to 00 peer ranks.  is connected to Expected number of connected peer ranks is : 00 peer ranks. 
Expected number of connected peer ranks is : 0
[Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 1 is connected to 7 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 70
 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank [Gloo] Rank 76 is connected to  is connected to 77 peer ranks.  peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 77

[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0[Gloo] Rank  is connected to 00[Gloo] Rank  peer ranks.  is connected to 0Expected number of connected peer ranks is : 0 is connected to 0 peer ranks. 0
Expected number of connected peer ranks is :  peer ranks. 0Expected number of connected peer ranks is : 
[Gloo] Rank 00
 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
The module name  (originally ) is not a valid Python identifier. Please rename the original module to avoid import issues.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
You are using a model of type hunyuan_image_3_moe to instantiate a model of type Hunyuan. This is not supported for all configurations of models and can yield errors.
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [config.py:379] Replacing legacy 'type' key with 'rope_type'
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:04 [platform.py:77] Defaulting to diffusion attention backend FLASH_ATTN
[Stage-0] INFO 03-01 22:54:05 [unquantized.py:131] Using TRITON backend for Unquantized MoE
Loading safetensors checkpoint shards:   0% Completed | 0/32 [00:00<?, ?it/s]
Loading safetensors checkpoint shards:   3% Completed | 1/32 [00:02<01:13,  2.36s/it]
Loading safetensors checkpoint shards:   6% Completed | 2/32 [00:05<01:31,  3.05s/it]
Loading safetensors checkpoint shards:   9% Completed | 3/32 [00:09<01:37,  3.35s/it]
Loading safetensors checkpoint shards:  12% Completed | 4/32 [00:13<01:41,  3.64s/it]
Loading safetensors checkpoint shards:  16% Completed | 5/32 [00:17<01:38,  3.65s/it]
Loading safetensors checkpoint shards:  19% Completed | 6/32 [00:21<01:38,  3.80s/it]
Loading safetensors checkpoint shards:  22% Completed | 7/32 [00:24<01:32,  3.69s/it]
Loading safetensors checkpoint shards:  25% Completed | 8/32 [00:28<01:28,  3.67s/it]
Loading safetensors checkpoint shards:  28% Completed | 9/32 [00:32<01:23,  3.62s/it]
Loading safetensors checkpoint shards:  31% Completed | 10/32 [00:35<01:18,  3.58s/it]
Loading safetensors checkpoint shards:  34% Completed | 11/32 [00:39<01:14,  3.57s/it]
Loading safetensors checkpoint shards:  38% Completed | 12/32 [00:42<01:11,  3.58s/it]
Loading safetensors checkpoint shards:  41% Completed | 13/32 [00:46<01:07,  3.54s/it]
Loading safetensors checkpoint shards:  44% Completed | 14/32 [00:49<01:04,  3.56s/it]
Loading safetensors checkpoint shards:  47% Completed | 15/32 [00:53<01:00,  3.55s/it]
Loading safetensors checkpoint shards:  50% Completed | 16/32 [00:56<00:56,  3.54s/it]
Loading safetensors checkpoint shards:  53% Completed | 17/32 [01:00<00:52,  3.52s/it]
Loading safetensors checkpoint shards:  56% Completed | 18/32 [01:03<00:49,  3.54s/it]
Loading safetensors checkpoint shards:  59% Completed | 19/32 [01:08<00:49,  3.77s/it]
Loading safetensors checkpoint shards:  62% Completed | 20/32 [01:11<00:44,  3.73s/it]
Loading safetensors checkpoint shards:  66% Completed | 21/32 [01:15<00:40,  3.66s/it]
Loading safetensors checkpoint shards:  69% Completed | 22/32 [01:19<00:36,  3.68s/it]
Loading safetensors checkpoint shards:  72% Completed | 23/32 [01:22<00:33,  3.68s/it]
Loading safetensors checkpoint shards:  75% Completed | 24/32 [01:26<00:29,  3.67s/it]
Loading safetensors checkpoint shards:  78% Completed | 25/32 [01:29<00:25,  3.60s/it]
Loading safetensors checkpoint shards:  81% Completed | 26/32 [01:33<00:21,  3.59s/it]
Loading safetensors checkpoint shards:  84% Completed | 27/32 [01:36<00:17,  3.55s/it]
[Stage-0] INFO 03-01 22:55:45 [diffusers_loader.py:301] Loading weights took 97.34 seconds
[Stage-0] INFO 03-01 22:55:46 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 102.764971 seconds
[Stage-0] INFO 03-01 22:55:46 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:46 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:46 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:47 [diffusion_worker.py:142] Worker 5: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:55:47 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:5, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:55:47 [diffusion_worker.py:84] Worker 5: Initialization complete.
[Stage-0] INFO 03-01 22:55:47 [diffusion_worker.py:483] Worker 5: Scheduler loop started.
[Stage-0] INFO 03-01 22:55:47 [diffusion_worker.py:406] Worker 5 ready to receive requests via shared memory
Loading safetensors checkpoint shards:  88% Completed | 28/32 [01:39<00:13,  3.41s/it]
Loading safetensors checkpoint shards:  91% Completed | 29/32 [01:42<00:09,  3.30s/it]
Loading safetensors checkpoint shards:  94% Completed | 30/32 [01:45<00:06,  3.22s/it]
Loading safetensors checkpoint shards:  97% Completed | 31/32 [01:48<00:02,  2.93s/it]
Loading safetensors checkpoint shards: 100% Completed | 32/32 [01:49<00:00,  2.57s/it]
Loading safetensors checkpoint shards: 100% Completed | 32/32 [01:49<00:00,  3.44s/it]

[Stage-0] INFO 03-01 22:55:57 [diffusers_loader.py:301] Loading weights took 110.06 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusers_loader.py:301] Loading weights took 110.57 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusers_loader.py:301] Loading weights took 110.77 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusers_loader.py:301] Loading weights took 110.89 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusers_loader.py:301] Loading weights took 110.90 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusers_loader.py:301] Loading weights took 110.92 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusers_loader.py:301] Loading weights took 110.96 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 115.113843 seconds
[Stage-0] INFO 03-01 22:55:58 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:58 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:58 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 115.452985 seconds
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:59 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 115.529497 seconds
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:59 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 115.653362 seconds
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:59 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 115.673151 seconds
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:59 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 115.692698 seconds
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:59 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:142] Worker 0: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:142] Worker 1: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:142] Worker 7: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:118] Model loading took 23.6234 GiB and 115.866355 seconds
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:123] Model runner: Model loaded successfully.
[Stage-0] WARNING 03-01 22:55:59 [diffusion_model_runner.py:141] Model runner: torch.compile failed with error: 'HunyuanImage3Pipeline' object has no attribute 'transformer'. Using eager mode.
[Stage-0] INFO 03-01 22:55:59 [diffusion_model_runner.py:163] Model runner: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:0, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:84] Worker 0: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:483] Worker 0: Scheduler loop started.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:406] Worker 0 ready to receive requests via shared memory
[Stage-0] INFO 03-01 22:55:59 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:7, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:84] Worker 7: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:483] Worker 7: Scheduler loop started.
[Stage-0] INFO 03-01 22:55:59 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:1, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:406] Worker 7 ready to receive requests via shared memory
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:84] Worker 1: Initialization complete.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:483] Worker 1: Scheduler loop started.
[Stage-0] INFO 03-01 22:55:59 [diffusion_worker.py:406] Worker 1 ready to receive requests via shared memory
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:142] Worker 6: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:142] Worker 2: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:142] Worker 4: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:56:00 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:6, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:84] Worker 6: Initialization complete.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:483] Worker 6: Scheduler loop started.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:406] Worker 6 ready to receive requests via shared memory
[Stage-0] INFO 03-01 22:56:00 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:2, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:84] Worker 2: Initialization complete.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:483] Worker 2: Scheduler loop started.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:406] Worker 2 ready to receive requests via shared memory
[Stage-0] INFO 03-01 22:56:00 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:4, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:84] Worker 4: Initialization complete.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:483] Worker 4: Scheduler loop started.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:406] Worker 4 ready to receive requests via shared memory
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:142] Worker 3: Process-scoped GPU memory after model loading: 0.00 GiB.
[Stage-0] INFO 03-01 22:56:00 [manager.py:91] Initializing DiffusionLoRAManager: device=cuda:3, dtype=torch.bfloat16, max_cached_adapters=1, static_lora_path=None
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:84] Worker 3: Initialization complete.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:483] Worker 3: Scheduler loop started.
[Stage-0] INFO 03-01 22:56:00 [diffusion_worker.py:406] Worker 3 ready to receive requests via shared memory
[Stage-0] INFO 03-01 22:56:00 [scheduler.py:41] SyncScheduler initialized result MessageQueue
[Stage-0] INFO 03-01 22:56:00 [diffusion_engine.py:341] dummy run to warm up the model
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
[Stage-0] INFO 03-01 22:56:00 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:00 [kv_transfer_manager.py:517] Request has no ID, cannot receive KV cache
  0%|                                                                                                                                                                                                             | 0/1 [00:00<?, ?it/s][Stage-0] WARNING 03-01 22:56:12 [fused_moe.py:1087] Using default MoE config. Performance might be sub-optimal! Config file not found at /usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/fused_moe/configs/E=64,N=384,device_name=NVIDIA_A100-SXM4-80GB.json
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.45s/it]

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.42s/it]
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.68s/it]
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.67s/it]
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.67s/it]
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.67s/it]
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.68s/it]
[Stage-0] INFO 03-01 22:56:23 [omni_stage.py:794] Max batch size: 1
INFO 03-01 22:56:23 [omni.py:448] [Orchestrator] Stage-0 reported ready
INFO 03-01 22:56:23 [omni.py:477] [Orchestrator] All stages initialized successfully

============================================================
Generation Configuration:
  Model: /data/HunyuanImage-3.0/
  Inference steps: 50
  Cache backend: None (no acceleration)
  Quantization: None (BF16)
  Parallel configuration: tensor_parallel_size=8, ulysses_degree=1, ring_degree=1, cfg_parallel_size=1, vae_patch_parallel_size=1
  CPU offload: False
  Image size: 1024x1024
============================================================

Adding requests:   0%|                                                                                                                                                                                            | 0/1 [00:00<?, ?it/s[Stage-0] INFO 03-01 22:56:23 [manager.py:566] Deactivating all adapters: 0 layers                                                                             | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 unit/s, output: 0.00 unit/s]
[Stage-0] WARNING 03-01 22:56:23 [kv_transfer_manager.py:421] No connector available for receiving KV cache
[Stage-0] INFO 03-01 22:56:23 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:23 [kv_transfer_manager.py:421] No connector available for receiving KV cache
  0%|                                                                                                                                                                                                            | 0/50 [00:00<?, ?it/s][Stage-0] INFO 03-01 22:56:23 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:23 [kv_transfer_manager.py:421] No connector available for receiving KV cache
[Stage-0] INFO 03-01 22:56:23 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:23 [kv_transfer_manager.py:421] No connector available for receiving KV cache
  0%|                                                                                                                                                                                                            | 0/50 [00:00<?, ?it/s][Stage-0] INFO 03-01 22:56:24 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:24 [kv_transfer_manager.py:421] No connector available for receiving KV cache
[Stage-0] INFO 03-01 22:56:24 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:24 [kv_transfer_manager.py:421] No connector available for receiving KV cache
  0%|                                                                                                                                                                                                            | 0/50 [00:00<?, ?it/s][Stage-0] INFO 03-01 22:56:24 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:24 [kv_transfer_manager.py:421] No connector available for receiving KV cache
  0%|                                                                                                                                                                                                            | 0/50 [00:00<?, ?it/s][Stage-0] INFO 03-01 22:56:24 [manager.py:566] Deactivating all adapters: 0 layers
[Stage-0] WARNING 03-01 22:56:24 [kv_transfer_manager.py:421] No connector available for receiving KV cache
 72%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▍                                                      | 36/50 [00:57<00:22,  1.59s/it][Stage-0] INFO 03-01 22:57:23 [shm_broadcast.py:542] No available shared memory broadcast block found in 60 seconds. This typically happens when some processes are hanging or doing some time-consuming work (e.g. compilation, weight/kv cache quantization).
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 50/50 [01:19<00:00,  1.60s/it]
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 50/50 [01:20<00:00,  1.61s/it]
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 50/50 [01:20<00:00,  1.60s/it]


100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 50/50 [01:19<00:00,  1.60s/it]


[Stage-0] INFO 03-01 22:57:49 [diffusion_engine.py:80] Generation completed successfully.
[Stage-0] INFO 03-01 22:57:49 [diffusion_engine.py:98] Post-processing completed in 0.0000 seconds
Processed prompts: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [01:26<00:00, 86.04s/img, est. speed stage-0 img/s: 0.00, avg e2e_lat: 0.0ms]
Total generation time: 86.0467 seconds (86046.67 ms)█████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [01:26<00:00, 86.04s/img, est. speed stage-0 img/s: 0.00, avg e2e_lat: 0.0ms]
INFO 03-01 22:57:49 [text_to_image.py:401] Outputs: [OmniRequestOutput(request_id='', finished=True, stage_id=0, final_output_type='image', request_output=[OmniRequestOutput(request_id='0_ae4da804-a11b-4c26-826a-28bd46516550', finished=True, stage_id=None, final_output_type='image', request_output=None, images=[1 PIL Images], prompt={'prompt': 'A brown and white dog is running on the grass', 'negative_prompt': None, 'additional_information': {'global_request_id': ['0_ae4da804-a11b-4c26-826a-28bd46516550']}}, latents=None, metrics={'image_num': 1, 'resolution': 640, 'postprocess_time_ms': 0.023603439331054688}, multimodal_output={})], images=[], prompt=None, latents=None, metrics={}, multimodal_output={})]
Adding requests:   0%|                                                                                                                                                                                            | 0/1 [01:26<?, ?it/s]
Saved generated image to output_image.png
WARNING 03-01 22:57:50 [omni_stage.py:542] Failed to send shutdown to in_q: Socket operation on non-socket
[Stage-0] INFO 03-01 22:57:50 [omni_stage.py:842] Received shutdown signal
/usr/lib/python3.12/multiprocessing/resource_tracker.py:279: UserWarning: resource_tracker: There appear to be 8 leaked semaphore objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '
/usr/lib/python3.12/multiprocessing/resource_tracker.py:279: UserWarning: resource_tracker: There appear to be 1 leaked shared_memory objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '
076EF8B6-120F-4274-A4EA-EBA265BA76A4
Essential Elements of an Effective PR Description Checklist
  • The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
  • The test plan. Please provide the test scripts & test commands. Please state the reasons if your codes don't require additional test scripts. For test file guidelines, please check the test style doc
  • The test results. Please paste the results comparison before and after, or the e2e results.
  • (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model. Please run mkdocs serve to sync the documentation editions to ./docs.
  • (Optional) Release notes update. If your change is user-facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

@Semmer2 Semmer2 requested a review from hsliuustc0106 as a code owner March 2, 2026 07:18
chatgpt-codex-connector[bot]
chatgpt-codex-connector Bot reviewed Mar 2, 2026
View reviewed changes
Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 01276c1bea

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

Comment thread vllm_omni/diffusion/models/hunyuan_image_3/pipeline_hunyuan_image_3.py
custom_pos_emb = self.get_pos_emb(custom_pos_emb, position_ids)

inputs_embeds = self.model.wte(input_ids)
inputs_embeds = self.model.embed_tokens(input_ids)
Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector Bot Mar 2, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Avoid calling embed_tokens on non-first PP ranks

forward_call now unconditionally does self.model.embed_tokens(input_ids), but HunyuanImage3Model.__init__ only creates embed_tokens on the first PP rank (or last when tied embeddings); other pipeline-parallel ranks get PPMissingLayer. With pipeline_parallel_size > 1 and default tie_word_embeddings=False, this change makes non-first ranks invoke a missing layer and fail during inference, whereas the previous self.model.wte path existed on every rank.

Useful? React with 👍 / 👎.

Copy link
Copy Markdown
Contributor Author

@Semmer2 Semmer2 Mar 2, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Avoid calling embed_tokens on non-first PP ranks

forward_call now unconditionally does self.model.embed_tokens(input_ids), but HunyuanImage3Model.__init__ only creates embed_tokens on the first PP rank (or last when tied embeddings); other pipeline-parallel ranks get PPMissingLayer. With pipeline_parallel_size > 1 and default tie_word_embeddings=False, this change makes non-first ranks invoke a missing layer and fail during inference, whereas the previous self.model.wte path existed on every rank.

Useful? React with 👍 / 👎.

Current model do not support PP, so PP.is_first_rank is always true. No need to check for now.

@Semmer2 Semmer2 force-pushed the load_weight_error branch from 01276c1 to 883a04c Compare March 2, 2026 07:26
@Semmer2
[BugFix] Fix load_weights error when loading HunyuanImage3.0
86bbf58 
Move some submodule load weights code of HunyuanImage3Pipeline to
AutoWeightsLoader:load_weights, fix weights not initialized error.

Signed-off-by: Semmer2 <semmer@live.cn>
@Semmer2 Semmer2 force-pushed the load_weight_error branch from 883a04c to 86bbf58 Compare March 2, 2026 07:32
@princepride princepride added the ready label to trigger buildkite CI label Mar 2, 2026
@hsliuustc0106 hsliuustc0106 enabled auto-merge (squash) March 2, 2026 08:30
@hsliuustc0106 hsliuustc0106 merged commit 1ca198e into vllm-project:main Mar 2, 2026
6 of 7 checks passed
yJader pushed a commit to omni-nicelab/vllm-omni-batching that referenced this pull request Mar 3, 2026
@Semmer2 @yJader
[BugFix] Fix load_weights error when loading HunyuanImage3.0 (vllm-pr…
066057e 
…oject#1598)

Signed-off-by: Semmer2 <semmer@live.cn>

Signed-off-by: jader <yjader@foxmail.com>
yJader pushed a commit to omni-nicelab/vllm-omni-batching that referenced this pull request Mar 3, 2026
@Semmer2
[BugFix] Fix load_weights error when loading HunyuanImage3.0 (vllm-pr…
50218a0 
…oject#1598)

Signed-off-by: Semmer2 <semmer@live.cn>

Signed-off-by: jader <yjader@foxmail.com>
yJader pushed a commit to omni-nicelab/vllm-omni-batching that referenced this pull request Mar 3, 2026
@Semmer2
[BugFix] Fix load_weights error when loading HunyuanImage3.0 (vllm-pr…
f94bd67 
…oject#1598)

Signed-off-by: Semmer2 <semmer@live.cn>
hsliuustc0106 added a commit to hsliuustc0106/vllm-omni-skills that referenced this pull request Mar 4, 2026
@hsliuustc0106
chore: auto-update skills based on vllm-omni changes
49c85a9 
### vllm-omni-perf
- Source: [PR #1619](vllm-project/vllm-omni#1619) - [Bugfix] Fix Qwen3-TTS code predictor crash due to missing vLLM config context
- Changes:
  - Bug fix: [Bugfix] Fix Qwen3-TTS code predictor crash due to missing vLLM config context

### vllm-omni-contrib
- Source: [PR #1615](vllm-project/vllm-omni#1615) - [Doc] Fix links in the configuration doc
- Changes:
  - Bug fix: [Doc] Fix links in the configuration doc

### vllm-omni-perf
- Source: [PR #1609](vllm-project/vllm-omni#1609) - [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Changes:
  - Bug fix: [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation

### vllm-omni-image-gen
- Source: [PR #1609](vllm-project/vllm-omni#1609) - [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Changes:
  - Bug fix: [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Additions:
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image

### vllm-omni-api
- Source: [PR #1609](vllm-project/vllm-omni#1609) - [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Changes:
  - Bug fix: [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation

### vllm-omni-serving
- Source: [PR #1602](vllm-project/vllm-omni#1602) - [Bugfix] fix kernel error for qwen3-omni
- Changes:
  - Bug fix: [Bugfix] fix kernel error for qwen3-omni

### vllm-omni-perf
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0

### vllm-omni-image-gen
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Additions:
  - HunyuanImage3
  - HunyuanImage3Pipeline
  - HunyuanImage3
  - HunyuanImage-3
  - HunyuanImage-3
  - HunyuanImage-3
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage-3

### vllm-omni-quantization
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0

### vllm-omni-distributed
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0

### vllm-omni-contrib
- Source: [PR #1576](vllm-project/vllm-omni#1576) - 0.16.0 release

### vllm-omni-audio-tts
- Source: [PR #1570](vllm-project/vllm-omni#1570) - [bugfix] Fix unexpected argument 'is_finished' in function llm2code2wav_async_chunk of mimo-audio
- Changes:
  - Bug fix: [bugfix] Fix unexpected argument 'is_finished' in function llm2code2wav_async_chunk of mimo-audio

### vllm-omni-api
- Source: [PR #1566](vllm-project/vllm-omni#1566) - [Bugfix] Import InputPreprocessor into Renderer
- Changes:
  - Bug fix: [Bugfix] Import InputPreprocessor into Renderer

### vllm-omni-perf
- Source: [PR #1565](vllm-project/vllm-omni#1565) - [BugFix]: fix a lot of bug
- Changes:
  - Bug fix: [BugFix]: fix a lot of bug

### vllm-omni-contrib
- Source: [PR #1564](vllm-project/vllm-omni#1564) - [NPU][Bugfix] Align GPU side and recover qwen3-tts
- Changes:
  - Bug fix: [NPU][Bugfix] Align GPU side and recover qwen3-tts

### vllm-omni-audio-tts
- Source: [PR #1564](vllm-project/vllm-omni#1564) - [NPU][Bugfix] Align GPU side and recover qwen3-tts
- Changes:
  - Bug fix: [NPU][Bugfix] Align GPU side and recover qwen3-tts

### vllm-omni-perf
- Source: [PR #1562](vllm-project/vllm-omni#1562) - [BugFix] Fix unexpected crash when init OmniDiffusion
- Changes:
  - Bug fix: [BugFix] Fix unexpected crash when init OmniDiffusion

### vllm-omni-api
- Source: [PR #1562](vllm-project/vllm-omni#1562) - [BugFix] Fix unexpected crash when init OmniDiffusion
- Changes:
  - Bug fix: [BugFix] Fix unexpected crash when init OmniDiffusion

### vllm-omni-quantization
- Source: [PR #1562](vllm-project/vllm-omni#1562) - [BugFix] Fix unexpected crash when init OmniDiffusion
- Changes:
  - Bug fix: [BugFix] Fix unexpected crash when init OmniDiffusion

### vllm-omni-distributed
- Source: [PR #1562](vllm-project/vllm-omni#1562) - [BugFix] Fix unexpected crash when init OmniDiffusion
- Changes:
  - Bug fix: [BugFix] Fix unexpected crash when init OmniDiffusion

### vllm-omni-api
- Source: [PR #1554](vllm-project/vllm-omni#1554) - fix(qwen3-tts): fix Base ICL voice clone producing corrupted audio
- Changes:
  - Bug fix: fix(qwen3-tts): fix Base ICL voice clone producing corrupted audio

### vllm-omni-cicd
- Source: [PR #1543](vllm-project/vllm-omni#1543) - [CI] Modify some CI test cases to run on L4 environment to reduce H100 resource usage.

### vllm-omni-perf
- Source: [PR #1540](vllm-project/vllm-omni#1540) - Fix no embed text spk tokens
- Changes:
  - Bug fix: Fix no embed text spk tokens

### vllm-omni-distributed
- Source: [PR #1540](vllm-project/vllm-omni#1540) - Fix no embed text spk tokens
- Changes:
  - Bug fix: Fix no embed text spk tokens

### vllm-omni-perf
- Source: [PR #1539](vllm-project/vllm-omni#1539) - [Debug] Enable curl retry aligned with openai

### vllm-omni-quantization
- Source: [PR #1539](vllm-project/vllm-omni#1539) - [Debug] Enable curl retry aligned with openai

### vllm-omni-distributed
- Source: [PR #1539](vllm-project/vllm-omni#1539) - [Debug] Enable curl retry aligned with openai

### vllm-omni-image-gen
- Source: [PR #1538](vllm-project/vllm-omni#1538) - [CI][skip ci]Update H100 image link based on #1518

### vllm-omni-perf
- Source: [PR #1536](vllm-project/vllm-omni#1536) - [Bugfix] Fix transformers 5.x compat issues in online TTS serving
- Changes:
  - Bug fix: [Bugfix] Fix transformers 5.x compat issues in online TTS serving

### vllm-omni-serving
- Source: [PR #1536](vllm-project/vllm-omni#1536) - [Bugfix] Fix transformers 5.x compat issues in online TTS serving
- Changes:
  - Bug fix: [Bugfix] Fix transformers 5.x compat issues in online TTS serving

### vllm-omni-cicd
- Source: [PR #1534](vllm-project/vllm-omni#1534) - [Debug] Merge vllm pull 35368

### vllm-omni-contrib
- Source: [PR #1530](vllm-project/vllm-omni#1530) - [Docs] update async chunk docs diagram [skip ci]

### vllm-omni-distributed
- Source: [PR #1524](vllm-project/vllm-omni#1524) - [BugFix] Restore talker's config
- Changes:
  - Bug fix: [BugFix] Restore talker's config

### vllm-omni-api
- Source: [PR #1522](vllm-project/vllm-omni#1522) - [Bugfix] Use uds for zmq address if not set --stage-id
- Changes:
  - New feature: [Bugfix] Use uds for zmq address if not set --stage-id

### vllm-omni-perf
- Source: [PR #1521](vllm-project/vllm-omni#1521) - Revert gpu_1 job to use regular image

### vllm-omni-perf
- Source: [PR #1518](vllm-project/vllm-omni#1518) - Use pull through cache image for H100 pool

### vllm-omni-perf
- Source: [PR #1515](vllm-project/vllm-omni#1515) - [Bugfix] fix offline text_to_image error from #1009
- Changes:
  - Bug fix: [Bugfix] fix offline text_to_image error from #1009

### vllm-omni-image-gen
- Source: [PR #1515](vllm-project/vllm-omni#1515) - [Bugfix] fix offline text_to_image error from #1009
- Changes:
  - Bug fix: [Bugfix] fix offline text_to_image error from #1009
- Additions:
  - num-images-per-prompt

### vllm-omni-quantization
- Source: [PR #1515](vllm-project/vllm-omni#1515) - [Bugfix] fix offline text_to_image error from #1009
- Changes:
  - Bug fix: [Bugfix] fix offline text_to_image error from #1009

### vllm-omni-distributed
- Source: [PR #1515](vllm-project/vllm-omni#1515) - [Bugfix] fix offline text_to_image error from #1009
- Changes:
  - Bug fix: [Bugfix] fix offline text_to_image error from #1009

### vllm-omni-api
- Source: [PR #1509](vllm-project/vllm-omni#1509) - [Chore] remove unused logger in omni_diffusion (#531)

### vllm-omni-perf
- Source: [PR #1505](vllm-project/vllm-omni#1505) - [Doc] Update installation instructions for vllm 0.16.0

### vllm-omni-quantization
- Source: [PR #1505](vllm-project/vllm-omni#1505) - [Doc] Update installation instructions for vllm 0.16.0

### vllm-omni-distributed
- Source: [PR #1505](vllm-project/vllm-omni#1505) - [Doc] Update installation instructions for vllm 0.16.0

### vllm-omni-contrib
- Source: [PR #1505](vllm-project/vllm-omni#1505) - [Doc] Update installation instructions for vllm 0.16.0

### vllm-omni-video-gen
- Source: [PR #1504](vllm-project/vllm-omni#1504) - [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading
- Changes:
  - New feature: [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading

### vllm-omni-perf
- Source: [PR #1504](vllm-project/vllm-omni#1504) - [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading
- Changes:
  - New feature: [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading

### vllm-omni-api
- Source: [PR #1504](vllm-project/vllm-omni#1504) - [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading
- Changes:
  - New feature: [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading

### vllm-omni-cicd
- Source: [PR #1504](vllm-project/vllm-omni#1504) - [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading
- Changes:
  - New feature: [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading

### vllm-omni-contrib
- Source: [PR #1500](vllm-project/vllm-omni#1500) - [ROCm] [CI] [Docker] Point to use the latest vLLM v0.16.0 stable version

### vllm-omni-cicd
- Source: [PR #1492](vllm-project/vllm-omni#1492) - [Platform] Enable layerwise offload on all hardware

### vllm-omni-image-gen
- Source: [PR #1491](vllm-project/vllm-omni#1491) - [CI] Update Dockerfile for vllm-omni CI image and remove obsolete dep…

### vllm-omni-cicd
- Source: [PR #1488](vllm-project/vllm-omni#1488) - [XPU][NPU][ROCM] enable cpu_offloading flag for non_cuda

### vllm-omni-audio-tts
- Source: [PR #1482](vllm-project/vllm-omni#1482) - [Fix][Chore] Qwen3-TTS Modeling Minor Code Sanity Improvements
- Changes:
  - Bug fix: [Fix][Chore] Qwen3-TTS Modeling Minor Code Sanity Improvements

### vllm-omni-perf
- Source: [PR #1468](vllm-project/vllm-omni#1468) - [BugFix] process request.num_cached_tokens if it equals to the initial value
- Changes:
  - Bug fix: [BugFix] process request.num_cached_tokens if it equals to the initial value

### vllm-omni-audio-tts
- Source: [PR #1455](vllm-project/vllm-omni#1455) - [Bugfix] Fix case-sensitive task_type matching in Qwen3TTSModelForGeneration
- Changes:
  - Bug fix: [Bugfix] Fix case-sensitive task_type matching in Qwen3TTSModelForGeneration

### vllm-omni-cicd
- Source: [PR #1449](vllm-project/vllm-omni#1449) - [Test] Reduce Perf test case and fix modify stage config
- Changes:
  - Bug fix: [Test] Reduce Perf test case and fix modify stage config

### vllm-omni-cicd
- Source: [PR #1448](vllm-project/vllm-omni#1448) - [Bugfix] Race condition in MultiprocExecutor when concurent access to Scheduler
- Changes:
  - Bug fix: [Bugfix] Race condition in MultiprocExecutor when concurent access to Scheduler

### vllm-omni-cicd
- Source: [PR #1438](vllm-project/vllm-omni#1438) - [Qwen3TTS][Feat] Streaming output
- Changes:
  - New feature: [Qwen3TTS][Feat] Streaming output

### vllm-omni-api
- Source: [PR #1438](vllm-project/vllm-omni#1438) - [Qwen3TTS][Feat] Streaming output
- Changes:
  - New feature: [Qwen3TTS][Feat] Streaming output

### vllm-omni-contrib
- Source: [PR #1438](vllm-project/vllm-omni#1438) - [Qwen3TTS][Feat] Streaming output
- Changes:
  - New feature: [Qwen3TTS][Feat] Streaming output

### vllm-omni-audio-tts
- Source: [PR #1438](vllm-project/vllm-omni#1438) - [Qwen3TTS][Feat] Streaming output
- Changes:
  - New feature: [Qwen3TTS][Feat] Streaming output

### vllm-omni-cicd
- Source: [PR #1435](vllm-project/vllm-omni#1435) - [Doc][Test][Misc] ComfyUI test, more screenshot, and code cleaning

### vllm-omni-video-gen
- Source: [PR #1433](vllm-project/vllm-omni#1433) - [Debug] Multi-Request for Qwen 3 Omni use_audio_in_video

### vllm-omni-audio-tts
- Source: [PR #1433](vllm-project/vllm-omni#1433) - [Debug] Multi-Request for Qwen 3 Omni use_audio_in_video
hsliuustc0106 added a commit to hsliuustc0106/vllm-omni-skills that referenced this pull request Mar 7, 2026
@hsliuustc0106
chore: auto-update skills based on vllm-omni changes
214888e 
### vllm-omni-api
- Source: [PR #1724](vllm-project/vllm-omni#1724) - Revert "[Profile] Adding metrics for Diffusion/DiT Single diffusion Pipeline (#668)"
- Changes:
  - New feature: Revert "[Profile] Adding metrics for Diffusion/DiT Single diffusion Pipeline (#668)"

### vllm-omni-contrib
- Source: [PR #1724](vllm-project/vllm-omni#1724) - Revert "[Profile] Adding metrics for Diffusion/DiT Single diffusion Pipeline (#668)"
- Changes:
  - New feature: Revert "[Profile] Adding metrics for Diffusion/DiT Single diffusion Pipeline (#668)"

### vllm-omni-api
- Source: [PR #1716](vllm-project/vllm-omni#1716) - [Feature]:  Add vae-patch-parallel CLI argument in online serving
- Changes:
  - New feature: [Feature]:  Add vae-patch-parallel CLI argument in online serving

### vllm-omni-contrib
- Source: [PR #1716](vllm-project/vllm-omni#1716) - [Feature]:  Add vae-patch-parallel CLI argument in online serving
- Changes:
  - New feature: [Feature]:  Add vae-patch-parallel CLI argument in online serving

### vllm-omni-contrib
- Source: [PR #1693](vllm-project/vllm-omni#1693) - [skip CI][Docs] Add TTS model developer guide
- Changes:
  - New feature: [skip CI][Docs] Add TTS model developer guide

### vllm-omni-audio-tts
- Source: [PR #1688](vllm-project/vllm-omni#1688) - [MiMo-Audio] Bugfix tp lg than 1
- Changes:
  - Bug fix: [MiMo-Audio] Bugfix tp lg than 1

### vllm-omni-distributed
- Source: [PR #1688](vllm-project/vllm-omni#1688) - [MiMo-Audio] Bugfix tp lg than 1
- Changes:
  - Bug fix: [MiMo-Audio] Bugfix tp lg than 1

### vllm-omni-perf
- Source: [PR #1688](vllm-project/vllm-omni#1688) - [MiMo-Audio] Bugfix tp lg than 1
- Changes:
  - Bug fix: [MiMo-Audio] Bugfix tp lg than 1

### vllm-omni-perf
- Source: [PR #1687](vllm-project/vllm-omni#1687) - [BugFix] Return proper HTTP status for ErrorResponse in create_speech
- Changes:
  - Bug fix: [BugFix] Return proper HTTP status for ErrorResponse in create_speech

### vllm-omni-distributed
- Source: [PR #1687](vllm-project/vllm-omni#1687) - [BugFix] Return proper HTTP status for ErrorResponse in create_speech
- Changes:
  - Bug fix: [BugFix] Return proper HTTP status for ErrorResponse in create_speech

### vllm-omni-api
- Source: [PR #1687](vllm-project/vllm-omni#1687) - [BugFix] Return proper HTTP status for ErrorResponse in create_speech
- Changes:
  - Bug fix: [BugFix] Return proper HTTP status for ErrorResponse in create_speech
- Additions:
  - `/v1/audio/speech`

### vllm-omni-quantization
- Source: [PR #1687](vllm-project/vllm-omni#1687) - [BugFix] Return proper HTTP status for ErrorResponse in create_speech
- Changes:
  - Bug fix: [BugFix] Return proper HTTP status for ErrorResponse in create_speech

### vllm-omni-cicd
- Source: [PR #1683](vllm-project/vllm-omni#1683) - [CI] Remove high concurrency tests before issue #1374 fixed.
- Changes:
  - Bug fix: [CI] Remove high concurrency tests before issue #1374 fixed.

### vllm-omni-audio-tts
- Source: [PR #1678](vllm-project/vllm-omni#1678) - Add non-async chunk support for Qwen3-TTS
- Changes:
  - New feature: Add non-async chunk support for Qwen3-TTS

### vllm-omni-cicd
- Source: [PR #1678](vllm-project/vllm-omni#1678) - Add non-async chunk support for Qwen3-TTS
- Changes:
  - New feature: Add non-async chunk support for Qwen3-TTS

### vllm-omni-cicd
- Source: [PR #1677](vllm-project/vllm-omni#1677) - Replace hard-coded cuda generator with current_omni_platform.device_type

### vllm-omni-perf
- Source: [PR #1677](vllm-project/vllm-omni#1677) - Replace hard-coded cuda generator with current_omni_platform.device_type

### vllm-omni-serving
- Source: [PR #1675](vllm-project/vllm-omni#1675) - [Misc] remove logits_processor_pattern this field, because vllm have …

### vllm-omni-cicd
- Source: [PR #1666](vllm-project/vllm-omni#1666) - [Cleanup] Move cosyvoice3 tests to model subdirectory

### vllm-omni-audio-tts
- Source: [PR #1664](vllm-project/vllm-omni#1664) - [Bugfix] Fix all-silence TTS output: use float32 for speech tokenizer decoder
- Changes:
  - Bug fix: [Bugfix] Fix all-silence TTS output: use float32 for speech tokenizer decoder

### vllm-omni-cicd
- Source: [PR #1664](vllm-project/vllm-omni#1664) - [Bugfix] Fix all-silence TTS output: use float32 for speech tokenizer decoder
- Changes:
  - Bug fix: [Bugfix] Fix all-silence TTS output: use float32 for speech tokenizer decoder

### vllm-omni-distributed
- Source: [PR #1656](vllm-project/vllm-omni#1656) - [Optimize][Qwen3-Omni] Reduce inter-packet latency in async chunk

### vllm-omni-contrib
- Source: [PR #1656](vllm-project/vllm-omni#1656) - [Optimize][Qwen3-Omni] Reduce inter-packet latency in async chunk

### vllm-omni-quantization
- Source: [PR #1652](vllm-project/vllm-omni#1652) - [UX] Add progress bar for diffusion models
- Changes:
  - New feature: [UX] Add progress bar for diffusion models

### vllm-omni-perf
- Source: [PR #1652](vllm-project/vllm-omni#1652) - [UX] Add progress bar for diffusion models
- Changes:
  - New feature: [UX] Add progress bar for diffusion models

### vllm-omni-distributed
- Source: [PR #1651](vllm-project/vllm-omni#1651) - docs: Announce vllm-omni-skills community project

### vllm-omni-quantization
- Source: [PR #1651](vllm-project/vllm-omni#1651) - docs: Announce vllm-omni-skills community project

### vllm-omni-perf
- Source: [PR #1651](vllm-project/vllm-omni#1651) - docs: Announce vllm-omni-skills community project

### vllm-omni-contrib
- Source: [PR #1649](vllm-project/vllm-omni#1649) - [Misc] update wechat

### vllm-omni-perf
- Source: [PR #1642](vllm-project/vllm-omni#1642) - [chore] add _repeated_blocks for regional compilation support
- Changes:
  - New feature: [chore] add _repeated_blocks for regional compilation support

### vllm-omni-api
- Source: [PR #1641](vllm-project/vllm-omni#1641) - [Bugfix] Add TTS request validation to prevent engine crashes
- Changes:
  - New feature: [Bugfix] Add TTS request validation to prevent engine crashes

### vllm-omni-cicd
- Source: [PR #1641](vllm-project/vllm-omni#1641) - [Bugfix] Add TTS request validation to prevent engine crashes
- Changes:
  - New feature: [Bugfix] Add TTS request validation to prevent engine crashes

### vllm-omni-image-gen
- Source: [PR #1640](vllm-project/vllm-omni#1640) - [FP8 Quantization] Add FP8 quantization support for Flux transformer
- Changes:
  - New feature: [FP8 Quantization] Add FP8 quantization support for Flux transformer
- Additions:
  - text-to-image
  - Text-to-Image
  - Flux

### vllm-omni-quantization
- Source: [PR #1640](vllm-project/vllm-omni#1640) - [FP8 Quantization] Add FP8 quantization support for Flux transformer
- Changes:
  - New feature: [FP8 Quantization] Add FP8 quantization support for Flux transformer
- Additions:
  - FP8 support or improvements

### vllm-omni-contrib
- Source: [PR #1640](vllm-project/vllm-omni#1640) - [FP8 Quantization] Add FP8 quantization support for Flux transformer
- Changes:
  - New feature: [FP8 Quantization] Add FP8 quantization support for Flux transformer

### vllm-omni-perf
- Source: [PR #1640](vllm-project/vllm-omni#1640) - [FP8 Quantization] Add FP8 quantization support for Flux transformer
- Changes:
  - New feature: [FP8 Quantization] Add FP8 quantization support for Flux transformer

### vllm-omni-contrib
- Source: [PR #1631](vllm-project/vllm-omni#1631) - [BugFix] Fix LongCat Sequence Parallelism / Small Cleanup
- Changes:
  - Bug fix: [BugFix] Fix LongCat Sequence Parallelism / Small Cleanup

### vllm-omni-cicd
- Source: [PR #1628](vllm-project/vllm-omni#1628) - [Test][Qwen3-Omni]Modify Qwen3-Omni benchmark test cases

### vllm-omni-perf
- Source: [PR #1628](vllm-project/vllm-omni#1628) - [Test][Qwen3-Omni]Modify Qwen3-Omni benchmark test cases

### vllm-omni-perf
- Source: [PR #1619](vllm-project/vllm-omni#1619) - [Bugfix] Fix Qwen3-TTS code predictor crash due to missing vLLM config context
- Changes:
  - Bug fix: [Bugfix] Fix Qwen3-TTS code predictor crash due to missing vLLM config context

### vllm-omni-perf
- Source: [PR #1617](vllm-project/vllm-omni#1617) - [Refactor][Perf] Qwen3-TTS: re-prefill Code Predictor with torch.compile + enable Code2Wav decoder CUDA Graph
- Changes:
  - Performance improvement: [Refactor][Perf] Qwen3-TTS: re-prefill Code Predictor with torch.compile + enable Code2Wav decoder CUDA Graph

### vllm-omni-contrib
- Source: [PR #1615](vllm-project/vllm-omni#1615) - [Doc] Fix links in the configuration doc
- Changes:
  - Bug fix: [Doc] Fix links in the configuration doc

### vllm-omni-audio-tts
- Source: [PR #1614](vllm-project/vllm-omni#1614) - perf: replace per-element .item() GPU syncs with batch .tolist() in TTS code predictor
- Changes:
  - Performance improvement: perf: replace per-element .item() GPU syncs with batch .tolist() in TTS code predictor

### vllm-omni-perf
- Source: [PR #1614](vllm-project/vllm-omni#1614) - perf: replace per-element .item() GPU syncs with batch .tolist() in TTS code predictor
- Changes:
  - Performance improvement: perf: replace per-element .item() GPU syncs with batch .tolist() in TTS code predictor

### vllm-omni-image-gen
- Source: [PR #1609](vllm-project/vllm-omni#1609) - [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Changes:
  - Bug fix: [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Additions:
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image
  - GLM-Image

### vllm-omni-api
- Source: [PR #1609](vllm-project/vllm-omni#1609) - [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Changes:
  - Bug fix: [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation

### vllm-omni-perf
- Source: [PR #1609](vllm-project/vllm-omni#1609) - [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation
- Changes:
  - Bug fix: [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation

### vllm-omni-contrib
- Source: [PR #1604](vllm-project/vllm-omni#1604) - [Model]: support Helios  from ByteDance

### vllm-omni-perf
- Source: [PR #1604](vllm-project/vllm-omni#1604) - [Model]: support Helios  from ByteDance

### vllm-omni-serving
- Source: [PR #1602](vllm-project/vllm-omni#1602) - [Bugfix] fix kernel error for qwen3-omni
- Changes:
  - Bug fix: [Bugfix] fix kernel error for qwen3-omni

### vllm-omni-distributed
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0

### vllm-omni-image-gen
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Additions:
  - HunyuanImage3
  - HunyuanImage3Pipeline
  - HunyuanImage3
  - HunyuanImage-3
  - HunyuanImage-3
  - HunyuanImage-3
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage3Pipeline
  - HunyuanImage-3

### vllm-omni-quantization
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0

### vllm-omni-perf
- Source: [PR #1598](vllm-project/vllm-omni#1598) - [BugFix] Fix load_weights error when loading HunyuanImage3.0
- Changes:
  - Bug fix: [BugFix] Fix load_weights error when loading HunyuanImage3.0

### vllm-omni-audio-tts
- Source: [PR #1583](vllm-project/vllm-omni#1583) - [Feat][Qwen3TTS] reduce TTFA with flexible initial phase
- Changes:
  - New feature: [Feat][Qwen3TTS] reduce TTFA with flexible initial phase

### vllm-omni-api
- Source: [PR #1583](vllm-project/vllm-omni#1583) - [Feat][Qwen3TTS] reduce TTFA with flexible initial phase
- Changes:
  - New feature: [Feat][Qwen3TTS] reduce TTFA with flexible initial phase

### vllm-omni-cicd
- Source: [PR #1583](vllm-project/vllm-omni#1583) - [Feat][Qwen3TTS] reduce TTFA with flexible initial phase
- Changes:
  - New feature: [Feat][Qwen3TTS] reduce TTFA with flexible initial phase

### vllm-omni-contrib
- Source: [PR #1583](vllm-project/vllm-omni#1583) - [Feat][Qwen3TTS] reduce TTFA with flexible initial phase
- Changes:
  - New feature: [Feat][Qwen3TTS] reduce TTFA with flexible initial phase

### vllm-omni-api
- Source: [PR #1579](vllm-project/vllm-omni#1579) - [1/N][Refactor] Clean up dead code in output processor

### vllm-omni-serving
- Source: [PR #1579](vllm-project/vllm-omni#1579) - [1/N][Refactor] Clean up dead code in output processor

### vllm-omni-distributed
- Source: [PR #1578](vllm-project/vllm-omni#1578) - [Feature][Bagel] Add CFG parallel mode
- Changes:
  - New feature: [Feature][Bagel] Add CFG parallel mode

### vllm-omni-cicd
- Source: [PR #1578](vllm-project/vllm-omni#1578) - [Feature][Bagel] Add CFG parallel mode
- Changes:
  - New feature: [Feature][Bagel] Add CFG parallel mode

### vllm-omni-perf
- Source: [PR #1578](vllm-project/vllm-omni#1578) - [Feature][Bagel] Add CFG parallel mode
- Changes:
  - New feature: [Feature][Bagel] Add CFG parallel mode

### vllm-omni-contrib
- Source: [PR #1576](vllm-project/vllm-omni#1576) - 0.16.0 release

### vllm-omni-audio-tts
- Source: [PR #1570](vllm-project/vllm-omni#1570) - [bugfix] Fix unexpected argument 'is_finished' in function llm2code2wav_async_chunk of mimo-audio
- Changes:
  - Bug fix: [bugfix] Fix unexpected argument 'is_finished' in function llm2code2wav_async_chunk of mimo-audio

### vllm-omni-api
- Source: [PR #1566](vllm-project/vllm-omni#1566) - [Bugfix] Import InputPreprocessor into Renderer
- Changes:
  - Bug fix: [Bugfix] Import InputPreprocessor into Renderer

### vllm-omni-distributed
- Source: [PR #1539](vllm-project/vllm-omni#1539) - [Debug] Enable curl retry aligned with openai

### vllm-omni-quantization
- Source: [PR #1539](vllm-project/vllm-omni#1539) - [Debug] Enable curl retry aligned with openai

### vllm-omni-perf
- Source: [PR #1539](vllm-project/vllm-omni#1539) - [Debug] Enable curl retry aligned with openai

### vllm-omni-image-gen
- Source: [PR #1537](vllm-project/vllm-omni#1537) - [NPU] [Features] [Bugfix] Support mindiesd adaln
- Changes:
  - New feature: [NPU] [Features] [Bugfix] Support mindiesd adaln
- Additions:
  - mindiesd
  - mindiesd
  - Qwen-Image-Edit-2509
  - mindiesd
  - mindiesd
  - mindiesd
  - mindiesd

### vllm-omni-perf
- Source: [PR #1537](vllm-project/vllm-omni#1537) - [NPU] [Features] [Bugfix] Support mindiesd adaln
- Changes:
  - New feature: [NPU] [Features] [Bugfix] Support mindiesd adaln

### vllm-omni-serving
- Source: [PR #1536](vllm-project/vllm-omni#1536) - [Bugfix] Fix transformers 5.x compat issues in online TTS serving
- Changes:
  - Bug fix: [Bugfix] Fix transformers 5.x compat issues in online TTS serving

### vllm-omni-perf
- Source: [PR #1536](vllm-project/vllm-omni#1536) - [Bugfix] Fix transformers 5.x compat issues in online TTS serving
- Changes:
  - Bug fix: [Bugfix] Fix transformers 5.x compat issues in online TTS serving
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@chatgpt-codex-connector chatgpt-codex-connector[bot] chatgpt-codex-connector[bot] left review comments

@princepride princepride princepride approved these changes

@hsliuustc0106 hsliuustc0106 Awaiting requested review from hsliuustc0106 hsliuustc0106 is a code owner

Assignees

No one assigned

Labels

ready label to trigger buildkite CI

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@Semmer2 @princepride @hsliuustc0106