[XPU] Dispatch nn.LayerNorm to fused SYCL kernel via CustomOp (#57172)
Signed-off-by: Michal Ganczarenko <michal.ganczarenko@intel.com> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com> Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
co-authored by
Claude Sonnet 5
Harry Mellor
parent
a4c4bb418e
commit
df5668e5c9
@@ -35,6 +35,7 @@ from vllm.distributed.parallel_state import (
|
||||
)
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention import Attention
|
||||
from vllm.model_executor.layers.layernorm import LayerNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
QKVParallelLinear,
|
||||
@@ -156,11 +157,11 @@ class GPT2Block(nn.Module):
|
||||
hidden_size = config.hidden_size
|
||||
inner_dim = config.n_inner if config.n_inner is not None else 4 * hidden_size
|
||||
|
||||
self.ln_1 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
|
||||
self.ln_1 = LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
|
||||
self.attn = GPT2Attention(
|
||||
config, cache_config, quant_config, prefix=f"{prefix}.attn"
|
||||
)
|
||||
self.ln_2 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
|
||||
self.ln_2 = LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
|
||||
self.mlp = GPT2MLP(inner_dim, config, quant_config, prefix=f"{prefix}.mlp")
|
||||
|
||||
def forward(
|
||||
@@ -212,7 +213,7 @@ class GPT2Model(nn.Module):
|
||||
lambda prefix: GPT2Block(config, cache_config, quant_config, prefix=prefix),
|
||||
prefix=f"{prefix}.h",
|
||||
)
|
||||
self.ln_f = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_epsilon)
|
||||
self.ln_f = LayerNorm(self.embed_dim, eps=config.layer_norm_epsilon)
|
||||
self.make_empty_intermediate_tensors = make_empty_intermediate_tensors_factory(
|
||||
["hidden_states"], config.n_embd
|
||||
)
|
||||
|
||||
Reference in New Issue
Block a user