Cosmos 3 FP8: wrapping pipe.transformer (e.g. torch.compile) silently disables the checkpoint's W8A16 step policy
I maintainer di solito rispondono entro 1 giorno
Nessuno ha ancora preso questa issue.
Valutazione
- Difficoltà
- 4/5
- Tempo stimato
- 3-5 giorni
- Idoneità per principianti
- 42/100
- Tipo di issue
- Bug
- Chiarezza
- Abbastanza chiara
- Stato di attività
- Attiva
- Ambito
- ai-infra-agents
Direzione di ricerca
Start in src/diffusers/pipelines/cosmos/mixed_precision.py: the loader lookup getattr(type(module), "load_config", None) in _on_disk_quantization_config() is the cause, and resolve() falls back to format="none" when it fails. Try the suggested fix of looking up load_config on the instance, or unwrapping with unwrap_module in src/diffusers/utils/torch_utils.py. Done when Cosmos3MixedPrecisionConfig.resolve(torch.compile(model)) equals resolve(model) on a small model whose config declares the ModelOpt FP8 policy, with a regression test added.
Scritto dal modello di indicizzazione a partire dal testo della issue.
Descrizione
Describe the bug
For the ModelOpt FP8 Cosmos 3 checkpoints, the per-step W8A16/W8A8 policy (quantization_config.runtime.diffusion_step_policy in transformer/config.json, added in #14664) is resolved from self.transformer on every pipeline call:
Cosmos3MixedPrecisionConfig.resolve() → quantization_config_from_module() → _on_disk_quantization_config(), which looks up the config loader on the module's class:
loader = getattr(type(module), "load_config", None)
The in-memory config does not carry quantization_config.runtime (see the docstring of quantization_config_from_module, L409-L426), so this on-disk read is the only source of the policy. When pipe.transformer is a wrapper, type(module) is the wrapper class, which has no load_config. The on-disk config is never read, _parsed_checkpoint_policy() returns None, and resolve() returns format="none" (L157-L158). Every step then runs W8A8 instead of the checkpoint's W8A16 first/last steps, with no warning. The Cosmos 3 docs note that all-W8A8 multi-step video can flicker (cosmos3.md L50-L54).
pipe.transformer = torch.compile(pipe.transformer), the pattern shown in several docs pages (e.g. optimization/cache_dit.md L260, quantization/autoround.md L102), is such a wrapper (OptimizedModule). In-place pipe.transformer.compile() and compile_repeated_blocks() (used on the Cosmos 3 page) keep the class and are not affected. The line is unchanged on main: mixed_precision.py#L450.
Reproduction
The policy is printed with the same resolve() call the pipeline makes, before any generation:
import modelopt.torch.quantization.backends.fp8_per_tensor_gemm # noqa: F401 - registers the FP8 GEMM
import torch
from diffusers import Cosmos3OmniPipeline
from diffusers.pipelines.cosmos.mixed_precision import Cosmos3MixedPrecisionConfig
from modelopt.torch.opt import enable_huggingface_checkpointing
enable_huggingface_checkpointing()
pipe = Cosmos3OmniPipeline.from_pretrained("nvidia/Cosmos3-Nano", revision="fp8", dtype=torch.bfloat16)
pipe.vae.to("cuda")
pipe.transformer.enable_group_offload(
onload_device=torch.device("cuda"), offload_device=torch.device("cpu"),
offload_type="block_level", num_blocks_per_group=1, use_stream=True,
)
pipe.transformer = torch.compile(pipe.transformer, backend="eager") # the only difference between the two runs
print(f"type(pipe.transformer) = {type(pipe.transformer).__name__}")
print(f"resolved policy = {Cosmos3MixedPrecisionConfig.resolve(pipe.transformer)}")
Logs
Without the torch.compile line:
type(pipe.transformer) = Cosmos3OmniTransformer
resolved policy = Cosmos3MixedPrecisionConfig(format='fp8', first_steps=3, last_steps=3, reasoner_policy='high_precision', overlap='a16')
With it:
type(pipe.transformer) = OptimizedModule
resolved policy = Cosmos3MixedPrecisionConfig(format='none', first_steps=3, last_steps=3, reasoner_policy='high_precision', overlap='a16')
Effect on generation (T2V 848×480, 49 frames, 8 steps, seed 123, transformer with enable_group_offload(offload_type="block_level", num_blocks_per_group=1, use_stream=True) to fit a 16 GB GPU):
pipe.transformer |
call | resolved policy | mixed-precision runtime installed on the transformer |
|---|---|---|---|
Cosmos3OmniTransformer |
default | fp8, first/last 3 steps W8A16 |
yes (252 reasoner + 252 generation linears) |
Cosmos3OmniTransformer |
mixed_precision_format="none" |
none |
no; different video from the default run |
OptimizedModule (torch.compile(pipe.transformer, backend="eager")) |
default | none |
— (see below) |
With the compiled ModelOpt FP8 transformer, generation currently stops earlier inside Dynamo, so the end-to-end output of the compiled run cannot be shown; the resolved policy above is what the pipeline applies to it. Any wrapper that forwards attribute access and does run reaches the none schedule.
Dynamo traceback of the compiled run (separate issue, context only)
Traceback (most recent call last):
File "repro.py", line 73, in <module>
video = pipe(
^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 120, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/diffusers/pipelines/cosmos/pipeline_cosmos3_omni.py", line 1756, in __call__
preds_vision, preds_sound, preds_action = self.transformer(
^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/_dynamo/eval_frame.py", line 414, in __call__
return super().__call__(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/_dynamo/eval_frame.py", line 832, in compile_wrapper
return fn(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/diffusers/hooks/hooks.py", line 217, in new_forward
output = function_reference.forward(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/diffusers/hooks/hooks.py", line 216, in new_forward
args, kwargs = function_reference.pre_forward(module, *args, **kwargs)
[... torch/_dynamo frames ...]
File "/usr/local/lib/python3.12/dist-packages/modelopt/torch/opt/dynamic.py", line 81, in __len__
return len(self._callbacks)
^^^^^^^^^^^^^^^
torch._dynamo.exc.InternalTorchDynamoError: AttributeError: '_FoldedCallback' object has no attribute '_callbacks'
from user code:
File "/usr/local/lib/python3.12/dist-packages/diffusers/hooks/hooks.py", line 217, in torch_dynamo_resume_in_new_forward_at_216
output = function_reference.forward(*args, **kwargs)
File "/usr/local/lib/python3.12/dist-packages/diffusers/models/transformers/transformer_cosmos3.py", line 745, in forward
packed_tokens_vision = self.proj_in(packed_tokens_vision)
File "/usr/local/lib/python3.12/dist-packages/modelopt/torch/quantization/nn/modules/quant_linear.py", line 240, in forward
return super().forward(input, *args, **kwargs)
File "/usr/local/lib/python3.12/dist-packages/modelopt/torch/quantization/nn/modules/quant_module.py", line 240, in forward
return super().forward(input, *args, **kwargs)
File "/usr/local/lib/python3.12/dist-packages/modelopt/torch/quantization/nn/modules/quant_module.py", line 192, in forward
output = super().forward(input, *args, **kwargs)
File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/linear.py", line 134, in forward
return F.linear(input, self.weight, self.bias)
File "/usr/local/lib/python3.12/dist-packages/modelopt/torch/opt/dynamic.py", line 782, in __getattr__
return manager.get_da_cb(name)(self, value)
Suggested fix
Either of these, plus a regression test (resolve(torch.compile(model)) == resolve(model) on a small model whose config declares the ModelOpt FP8 policy):
- Look the loader up on the instance:
loader = getattr(module, "load_config", None).load_configis aConfigMixinclassmethod, so this returns the same bound classmethod, and wrappers that forward attribute access (OptimizedModule.__getattr__→_orig_mod) work too. - Or unwrap first with the existing helper
unwrap_module, as other Diffusers code does for compiled modules. This coverstorch.compileonly.
Optionally, a warning when a ModelOpt FP8 backend is detected but the on-disk policy could not be read, so the fallback to all-W8A8 is visible. Happy to open a PR.
System Info
- diffusers 0.41.0, nvidia-modelopt 0.44.0, transformers 5.20.2, accelerate 1.15.0
- torch 2.9.0a0+145a3a7 (container
nvcr.io/nvidia/pytorch:25.10-py3), CUDA 13.0, Python 3.12 - NVIDIA GeForce RTX 5080 16 GB, driver 620.62
- Lingua principale
- Python
- Stelle
- 34.9k
- Fork
- 7.5k
- Merge medio
- 3g 16h
- PR unite (30g)
- 65
Preparare l'ambiente
- Nessun Dockerfile né file Docker Compose
- Ha un modello di pull request
- Leggi la guida per i contributori
Come iniziare
- Leggi tutta la issue e poi la guida ai contributi del progetto.
- Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
- Fai un fork del repository e lavora su un branch.
- Apri una pull request che faccia riferimento al numero della issue.
Altre issue di huggingface/diffusers
-
Difficoltà 2/5 1-3 ore Idoneità per principianti 84/100
huggingface/diffusers#15363 ·
I maintainer di solito rispondono entro 1 giorno
-
`ZImageImg2ImgPipeline` / `ZImageInpaintPipeline`: explicit height/width different from the input image raises RuntimeErrorForse già presa @chessing234 l’ha presa 1 giorno fa. Apertabug
Difficoltà 2/5 1-3 ore Idoneità per principianti 70/100
huggingface/diffusers#15337 · 2 commenti ·
I maintainer di solito rispondono entro 1 giorno
-
Multi-vector textual inversion uses only the first vector when the token has uppercase letters (transformers v5)Forse già presa @sunhouy l’ha presa 2 giorni fa. Aperta
Difficoltà 2/5 1-3 ore Idoneità per principianti 75/100
huggingface/diffusers#15313 ·
I maintainer di solito rispondono entro 1 giorno
-
TCDScheduler advances the generator on the final step of a strength-truncated custom scheduleForse già presa @SIDDARTHAREDDY8 l’ha presa 5 giorni fa. Apertabug
Difficoltà 2/5 1-3 ore Idoneità per principianti 74/100
huggingface/diffusers#15288 · 2 commenti ·
I maintainer di solito rispondono entro 1 giorno
-
unload_textual_inversion(token) leaves the extra tokens of a multi-vector embedding behind, so it can't be loaded againForse già presa @sohailshk l’ha presa 11 giorni fa. Aperta
Difficoltà 2/5 1-3 ore Idoneità per principianti 72/100
huggingface/diffusers#15211 · 3 commenti ·
I maintainer di solito rispondono entro 1 giorno
Tutte le issue di huggingface/diffusers
Issue simili
-
Difficoltà 2/5 1-3 ore Idoneità per principianti 73/100
I maintainer di solito rispondono entro 1 giorno
-
Unused loot bag group names left in dreamcraft en_US.langForse già presa Una pull request collegata a questa issue è aperta o già unita. ApertaStatus: Triage Suggestion
Difficoltà 1/5 Meno di un'ora Idoneità per principianti 90/100
GTNewHorizons/GT-New-Horizons-Modpack#27789 ·
I maintainer di solito rispondono entro 1 giorno
-
bug frontend sev-5
Difficoltà 2/5 1-3 ore Idoneità per principianti 83/100
-
core good first issue
Difficoltà 2/5 1-3 ore Idoneità per principianti 65/100
-
b4b code health next
Difficoltà 2/5 1-3 ore Idoneità per principianti 68/100
ESCOMP/CTSM#4330 · 1 commento ·
I maintainer di solito rispondono entro 1 giorno