Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[bugfix] fix aria model and add torch.compile #10645

Merged
merged 1 commit into from
Nov 26, 2024
Merged

[bugfix] fix aria model and add torch.compile #10645

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 4 additions & 22 deletions vllm/model_executor/models/aria.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@
LlamaModel)
from vllm.model_executor.models.utils import (AutoWeightsLoader, WeightsMapper,
is_pp_missing_parameter,
make_layers, maybe_prefix,
maybe_prefix,
merge_multimodal_embeddings)
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.base import MultiModalInputs
Expand Down Expand Up @@ -363,27 +363,9 @@ class AriaMoELMModel(LlamaModel):
"""

def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
super().__init__(vllm_config=vllm_config, prefix=prefix)

config = vllm_config.model_config.hf_config
cache_config = vllm_config.cache_config
quant_config = vllm_config.quant_config

# FIXME: this is a hack to disable the compilation of the model
self.do_not_compile = True
Copy link
Member Author

Choose a reason for hiding this comment

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

@xffxff why do you want to disable it?

Copy link
Member Author

Choose a reason for hiding this comment

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

I tested it, and it can use torch.compile .

Copy link
Contributor

Choose a reason for hiding this comment

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

I encountered an error at the start of my PR: 'AriaMoELMModel' object has no attribute 'do_not_compile'. I didn’t investigate it further, as it might not be necessary right now. It’s great to see that we can now use torch.compile!


self.layers = None

self.start_layer, self.end_layer, self.layers = make_layers(
config.num_hidden_layers,
lambda prefix: MoEDecoderLayer(
config=config,
cache_config=cache_config,
quant_config=quant_config,
prefix=prefix,
),
prefix=f"{prefix}.layers",
)
super().__init__(vllm_config=vllm_config,
prefix=prefix,
layer_type=MoEDecoderLayer)

# Adapted from LlamaModel.load_weights with the modification of adding
# the expert weights mapping to `stacked_params_mapping`
Expand Down
16 changes: 10 additions & 6 deletions vllm/model_executor/models/llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
"""Inference-only LLaMA model compatible with HuggingFace weights."""
from typing import Any, Dict, Iterable, List, Optional, Set, Tuple, Union
from typing import Any, Dict, Iterable, List, Optional, Set, Tuple, Type, Union

import torch
from torch import nn
Expand Down Expand Up @@ -273,7 +273,11 @@ def forward(
@support_torch_compile
class LlamaModel(nn.Module):

def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
def __init__(self,
*,
vllm_config: VllmConfig,
prefix: str = "",
layer_type: Type[LlamaDecoderLayer] = LlamaDecoderLayer):
super().__init__()

config = vllm_config.model_config.hf_config
Expand All @@ -299,10 +303,10 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
self.embed_tokens = PPMissingLayer()
self.start_layer, self.end_layer, self.layers = make_layers(
config.num_hidden_layers,
lambda prefix: LlamaDecoderLayer(config=config,
cache_config=cache_config,
quant_config=quant_config,
prefix=prefix),
lambda prefix: layer_type(config=config,
cache_config=cache_config,
quant_config=quant_config,
prefix=prefix),
prefix=f"{prefix}.layers",
)
if get_pp_group().is_last_rank:
Expand Down