biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

128
vllm/model_executor/models/orion.py
View File

@@ -6,6 +6,7 @@
# Copyright (c) OrionStar Inc.
# LICENSE: https://huggingface.co/OrionStarAI/Orion-14B-Base/blob/main/LICENSE
"""Inference-only Orion-14B model compatible with HuggingFace weights."""
from collections.abc import Iterable
from itertools import islice
from typing import Any, Optional, Union
@@ -19,25 +20,32 @@ from vllm.compilation.decorators import support_torch_compile
from vllm.config import CacheConfig, VllmConfig
from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
from vllm.model_executor.layers.activation import SiluAndMul
from vllm.model_executor.layers.linear import (MergedColumnParallelLinear,
QKVParallelLinear,
RowParallelLinear)
from vllm.model_executor.layers.linear import (
MergedColumnParallelLinear,
QKVParallelLinear,
RowParallelLinear,
)
from vllm.model_executor.layers.logits_processor import LogitsProcessor
from vllm.model_executor.layers.quantization import QuantizationConfig
from vllm.model_executor.layers.rotary_embedding import get_rope
from vllm.model_executor.layers.vocab_parallel_embedding import (
ParallelLMHead, VocabParallelEmbedding)
ParallelLMHead,
VocabParallelEmbedding,
)
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.sequence import IntermediateTensors
from .interfaces import SupportsPP
from .utils import (AutoWeightsLoader, is_pp_missing_parameter,
make_empty_intermediate_tensors_factory, make_layers,
maybe_prefix)
from .utils import (
AutoWeightsLoader,
is_pp_missing_parameter,
make_empty_intermediate_tensors_factory,
make_layers,
maybe_prefix,
)
class OrionMLP(nn.Module):
def __init__(
self,
hidden_size: int,
@@ -47,16 +55,15 @@ class OrionMLP(nn.Module):
) -> None:
super().__init__()
self.gate_up_proj = MergedColumnParallelLinear(
hidden_size, [intermediate_size] * 2,
bias=False,
quant_config=quant_config)
self.down_proj = RowParallelLinear(intermediate_size,
hidden_size,
bias=False,
quant_config=quant_config)
hidden_size, [intermediate_size] * 2, bias=False, quant_config=quant_config
)
self.down_proj = RowParallelLinear(
intermediate_size, hidden_size, bias=False, quant_config=quant_config
)
if hidden_act != "silu":
raise ValueError(f"Unsupported activation: {hidden_act}. "
"Only silu is supported for now.")
raise ValueError(
f"Unsupported activation: {hidden_act}. Only silu is supported for now."
)
self.act_fn = SiluAndMul()
def forward(self, x):
@@ -67,7 +74,6 @@ class OrionMLP(nn.Module):
class OrionAttention(nn.Module):
def __init__(
self,
hidden_size: int,
@@ -125,13 +131,15 @@ class OrionAttention(nn.Module):
base=rope_theta,
rope_scaling=rope_scaling,
)
self.attn = Attention(self.num_heads,
self.head_dim,
self.scaling,
num_kv_heads=self.num_kv_heads,
cache_config=cache_config,
quant_config=quant_config,
prefix=f"{prefix}.attn")
self.attn = Attention(
self.num_heads,
self.head_dim,
self.scaling,
num_kv_heads=self.num_kv_heads,
cache_config=cache_config,
quant_config=quant_config,
prefix=f"{prefix}.attn",
)
def forward(
self,
@@ -147,7 +155,6 @@ class OrionAttention(nn.Module):
class OrionDecoderLayer(nn.Module):
def __init__(
self,
config: PretrainedConfig,
@@ -159,8 +166,7 @@ class OrionDecoderLayer(nn.Module):
self.hidden_size = config.hidden_size
rope_theta = getattr(config, "rope_theta", 10000)
rope_scaling = getattr(config, "rope_scaling", None)
max_position_embeddings = getattr(config, "max_position_embeddings",
8192)
max_position_embeddings = getattr(config, "max_position_embeddings", 8192)
self.self_attn = OrionAttention(
hidden_size=self.hidden_size,
num_heads=config.num_attention_heads,
@@ -179,10 +185,10 @@ class OrionDecoderLayer(nn.Module):
quant_config=quant_config,
)
self.input_layernorm = nn.LayerNorm(config.hidden_size,
eps=config.rms_norm_eps)
self.post_attention_layernorm = nn.LayerNorm(config.hidden_size,
eps=config.rms_norm_eps)
self.input_layernorm = nn.LayerNorm(config.hidden_size, eps=config.rms_norm_eps)
self.post_attention_layernorm = nn.LayerNorm(
config.hidden_size, eps=config.rms_norm_eps
)
def forward(
self,
@@ -209,7 +215,6 @@ class OrionDecoderLayer(nn.Module):
@support_torch_compile
class OrionModel(nn.Module):
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
super().__init__()
@@ -226,13 +231,17 @@ class OrionModel(nn.Module):
self.start_layer, self.end_layer, self.layers = make_layers(
config.num_hidden_layers,
lambda prefix: OrionDecoderLayer(
config, cache_config, quant_config, prefix=prefix),
prefix=f"{prefix}.layers")
config, cache_config, quant_config, prefix=prefix
),
prefix=f"{prefix}.layers",
)
self.norm = nn.LayerNorm(config.hidden_size, eps=config.rms_norm_eps)
self.make_empty_intermediate_tensors = (
make_empty_intermediate_tensors_factory([
self.make_empty_intermediate_tensors = make_empty_intermediate_tensors_factory(
[
"hidden_states",
], config.hidden_size))
],
config.hidden_size,
)
def get_input_embeddings(self, input_ids: torch.Tensor) -> torch.Tensor:
return self.embed_tokens(input_ids)
@@ -255,14 +264,15 @@ class OrionModel(nn.Module):
for layer in islice(self.layers, self.start_layer, self.end_layer):
hidden_states = layer(positions, hidden_states)
if not get_pp_group().is_last_rank:
return IntermediateTensors({
"hidden_states": hidden_states,
})
return IntermediateTensors(
{
"hidden_states": hidden_states,
}
)
hidden_states = self.norm(hidden_states)
return hidden_states
def load_weights(self, weights: Iterable[tuple[str,
torch.Tensor]]) -> set[str]:
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
stacked_params_mapping = [
# (param_name, shard_name, shard_id)
("qkv_proj", "q_proj", "q"),
@@ -274,7 +284,7 @@ class OrionModel(nn.Module):
params_dict = dict(self.named_parameters())
loaded_params: set[str] = set()
for name, loaded_weight in weights:
for (param_name, weight_name, shard_id) in stacked_params_mapping:
for param_name, weight_name, shard_id in stacked_params_mapping:
if weight_name not in name:
continue
name = name.replace(weight_name, param_name)
@@ -294,32 +304,34 @@ class OrionModel(nn.Module):
if is_pp_missing_parameter(name, self):
continue
param = params_dict[name]
weight_loader = getattr(param, "weight_loader",
default_weight_loader)
weight_loader = getattr(param, "weight_loader", default_weight_loader)
weight_loader(param, loaded_weight)
loaded_params.add(name)
return loaded_params
class OrionForCausalLM(nn.Module, SupportsPP):
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
super().__init__()
config = vllm_config.model_config.hf_config
quant_config = vllm_config.quant_config
self.config = config
self.quant_config = quant_config
self.model = OrionModel(vllm_config=vllm_config,
prefix=maybe_prefix(prefix, "model"))
self.lm_head = ParallelLMHead(config.vocab_size,
config.hidden_size,
quant_config=quant_config,
prefix=maybe_prefix(prefix, "lm_head"))
self.model = OrionModel(
vllm_config=vllm_config, prefix=maybe_prefix(prefix, "model")
)
self.lm_head = ParallelLMHead(
config.vocab_size,
config.hidden_size,
quant_config=quant_config,
prefix=maybe_prefix(prefix, "lm_head"),
)
if self.config.tie_word_embeddings:
self.lm_head.weight = self.model.embed_tokens.weight
self.logits_processor = LogitsProcessor(config.vocab_size)
self.make_empty_intermediate_tensors = (
self.model.make_empty_intermediate_tensors)
self.model.make_empty_intermediate_tensors
)
def get_input_embeddings(self, input_ids: torch.Tensor) -> torch.Tensor:
return self.model.get_input_embeddings(input_ids)
@@ -331,8 +343,9 @@ class OrionForCausalLM(nn.Module, SupportsPP):
intermediate_tensors: Optional[IntermediateTensors] = None,
inputs_embeds: Optional[torch.Tensor] = None,
) -> Union[torch.Tensor, IntermediateTensors]:
hidden_states = self.model(input_ids, positions, intermediate_tensors,
inputs_embeds)
hidden_states = self.model(
input_ids, positions, intermediate_tensors, inputs_embeds
)
return hidden_states
def compute_logits(
@@ -342,7 +355,6 @@ class OrionForCausalLM(nn.Module, SupportsPP):
logits = self.logits_processor(self.lm_head, hidden_states)
return logits
def load_weights(self, weights: Iterable[tuple[str,
torch.Tensor]]) -> set[str]:
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
loader = AutoWeightsLoader(self)
return loader.load_weights(weights)