huggingface / huggingface/diffusers

MiniMax-H3 VAE decoder loads in fp32 but CUDA decode runs it in fp16

Aperta
#14,746 1 commento 0 reazioni 0 assegnatari Vedi su GitHub
bug modular-pipelines needs-env-info
Lingua principale
Python
Stelle
34.5k
Fork
7.3k
Merge medio
3g 3h
PR unite (30g)
91

Descrizione

### Describe the bug

The VAE keeps everything in fp32 no matter what dtype you pass, but the decode block wraps vae.decode in a fp16 autocast, so the decoder gets downcast to fp16 anyway on every call. On CUDA this just costs VRAM, RAM and disk space.

Decoding 192 frames at 1344x768:

| |weights|peak|
|---|---|---|
|main|9.70 GiB|16.69 GiB|
|decode in fp16|5.19 GiB|7.67 GiB|

As seen on the table, this prevents to use this model on 16GB GPUs for that resolution and duration.

### Reproduction

Just normal inference on a cuda GPU will reproduce it

```py
import torch

from diffusers import ComponentsManager, ModularPipeline

manager = ComponentsManager()
manager.enable_auto_cpu_offload(device="cuda")

pipe = ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3", components_manager=manager)
pipe.load_components(workflow="t2va", dtype=torch.bfloat16)

results = pipe(
prompt="An astronaut hiking through the mountains, humming a tune",
num_frames=192,
output=["videos", "audio", "sampling_rate"],
)
```

code to test a PoC fix:

```py
import torch

from diffusers import AutoencoderKLMiniMaxH3

class Fp16DecoderVae(AutoencoderKLMiniMaxH3):
_keep_in_fp32_modules = [
"encoder",
"quant_conv",
"post_quant_conv",
"norm1",
"norm2",
"norm_out",
"scale1",
"scale2",
]

@classmethod
def from_pretrained(cls, *args, **kwargs):
kwargs.pop("torch_dtype", None)
kwargs["dtype"] = torch.float16
return super().from_pretrained(*args, **kwargs)

vae = Fp16DecoderVae.from_pretrained("MiniMaxAI/MiniMax-H3", subfolder="vae", dtype=torch.bfloat16)
```

### main
https://github.com/user-attachments/assets/5bfbfedb-cfed-4645-87b7-58d9c8e8fa9f

### PoC fix
https://github.com/user-attachments/assets/72429905-03b6-4590-bee0-f4fd2e4a39d4

### System Info

diffusers from main
any cuda platform

### Who can help?

Guida per i contributori

Apri la guida per i contributori

Direzione di ricerca

Start at the AutoencoderKLMiniMaxH3 implementation and trace the decode block's autocast behavior during the provided CUDA inference reproduction. Compare it with the Fp16DecoderVae PoC, then verify that the decoder uses the intended dtype and that the 192-frame, 1344x768 case no longer incurs the reported fp32 memory usage.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Valutazione

Stack tecnologico
python, pytorch
Ambito
machine-learning, performance
Tipo di issue
Bug
Difficoltà
4/5
Tempo stimato
3-5 giorni
Stato di attività
Attiva
Chiarezza
Abbastanza chiara
Idoneità per principianti
55/100

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.