Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
change
sglang
Commits
8616357a
Unverified
Commit
8616357a
authored
Feb 12, 2025
by
Liangsheng Yin
Committed by
GitHub
Feb 12, 2025
Browse files
Fix deepseek awq v3 (#3450)
parent
8adbc78b
Changes
4
Hide whitespace changes
Inline
Side-by-side
Showing
4 changed files
with
69 additions
and
10 deletions
+69
-10
python/sglang/srt/layers/linear.py
python/sglang/srt/layers/linear.py
+12
-5
python/sglang/srt/layers/moe/fused_moe_triton/layer.py
python/sglang/srt/layers/moe/fused_moe_triton/layer.py
+2
-0
python/sglang/srt/layers/quantization/__init__.py
python/sglang/srt/layers/quantization/__init__.py
+51
-5
python/sglang/srt/models/deepseek_v2.py
python/sglang/srt/models/deepseek_v2.py
+4
-0
No files found.
python/sglang/srt/layers/linear.py
View file @
8616357a
...
@@ -421,11 +421,18 @@ class ColumnParallelLinear(LinearBase):
...
@@ -421,11 +421,18 @@ class ColumnParallelLinear(LinearBase):
if
len
(
loaded_weight
.
shape
)
==
0
:
if
len
(
loaded_weight
.
shape
)
==
0
:
assert
loaded_weight
.
numel
()
==
1
assert
loaded_weight
.
numel
()
==
1
loaded_weight
=
loaded_weight
.
reshape
(
1
)
loaded_weight
=
loaded_weight
.
reshape
(
1
)
param
.
load_column_parallel_weight
(
loaded_weight
,
from
sglang.srt.layers.parameter
import
_ColumnvLLMParameter
tp_rank
=
self
.
tp_rank
,
use_presharded_weights
=
self
.
use_presharded_weights
,
if
isinstance
(
param
,
_ColumnvLLMParameter
):
)
# FIXME: why would we need this special case?
param
.
load_column_parallel_weight
(
loaded_weight
,
tp_rank
=
self
.
tp_rank
,
use_presharded_weights
=
self
.
use_presharded_weights
,
)
else
:
param
.
load_column_parallel_weight
(
loaded_weight
)
def
forward
(
self
,
input_
):
def
forward
(
self
,
input_
):
bias
=
self
.
bias
if
not
self
.
skip_bias_add
else
None
bias
=
self
.
bias
if
not
self
.
skip_bias_add
else
None
...
...
python/sglang/srt/layers/moe/fused_moe_triton/layer.py
View file @
8616357a
...
@@ -298,7 +298,9 @@ class FusedMoE(torch.nn.Module):
...
@@ -298,7 +298,9 @@ class FusedMoE(torch.nn.Module):
layer
=
self
,
layer
=
self
,
num_experts
=
num_experts
,
num_experts
=
num_experts
,
hidden_size
=
hidden_size
,
hidden_size
=
hidden_size
,
# FIXME: figure out which intermediate_size to use
intermediate_size
=
self
.
intermediate_size_per_partition
,
intermediate_size
=
self
.
intermediate_size_per_partition
,
intermediate_size_per_partition
=
self
.
intermediate_size_per_partition
,
params_dtype
=
params_dtype
,
params_dtype
=
params_dtype
,
weight_loader
=
self
.
weight_loader
,
weight_loader
=
self
.
weight_loader
,
)
)
...
...
python/sglang/srt/layers/quantization/__init__.py
View file @
8616357a
# Adapted from https://raw.githubusercontent.com/vllm-project/vllm/v0.5.5/vllm/model_executor/layers/quantization/__init__.py
# Adapted from https://raw.githubusercontent.com/vllm-project/vllm/v0.5.5/vllm/model_executor/layers/quantization/__init__.py
from
typing
import
Callable
,
Dict
,
Optional
,
Type
from
typing
import
Dict
,
Type
import
torch
from
vllm.model_executor.layers.quantization.aqlm
import
AQLMConfig
from
vllm.model_executor.layers.quantization.aqlm
import
AQLMConfig
from
vllm.model_executor.layers.quantization.awq
import
AWQConfig
from
vllm.model_executor.layers.quantization.awq
import
AWQConfig
from
vllm.model_executor.layers.quantization.awq_marlin
import
AWQMarlinConfig
from
vllm.model_executor.layers.quantization.awq_marlin
import
(
AWQMarlinConfig
,
AWQMoEMethod
,
)
from
vllm.model_executor.layers.quantization.bitsandbytes
import
BitsAndBytesConfig
from
vllm.model_executor.layers.quantization.bitsandbytes
import
BitsAndBytesConfig
from
vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors
import
(
from
vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors
import
(
CompressedTensorsConfig
,
CompressedTensorsConfig
,
...
@@ -73,21 +76,61 @@ def gptq_get_quant_method(self, layer, prefix):
...
@@ -73,21 +76,61 @@ def gptq_get_quant_method(self, layer, prefix):
def
awq_get_quant_method
(
self
,
layer
,
prefix
):
def
awq_get_quant_method
(
self
,
layer
,
prefix
):
from
vllm.model_executor.layers.quantization.awq
import
is_layer_skipped_awq
from
vllm.model_executor.layers.quantization.awq_marlin
import
(
from
vllm.model_executor.layers.quantization.awq_marlin
import
(
AWQMarlinLinearMethod
,
AWQMarlinLinearMethod
,
AWQMoEMethod
,
AWQMoEMethod
,
)
)
from
sglang.srt.layers.linear
import
LinearBase
from
sglang.srt.layers.linear
import
LinearBase
,
UnquantizedLinearMethod
from
sglang.srt.layers.moe.fused_moe_triton.layer
import
FusedMoE
from
sglang.srt.layers.moe.fused_moe_triton.layer
import
FusedMoE
from
sglang.srt.layers.vocab_parallel_embedding
import
ParallelLMHead
if
isinstance
(
layer
,
LinearBase
):
if
isinstance
(
layer
,
LinearBase
)
or
(
isinstance
(
layer
,
ParallelLMHead
)
and
self
.
lm_head_quantized
):
if
is_layer_skipped_awq
(
prefix
,
self
.
modules_to_not_convert
):
return
UnquantizedLinearMethod
()
return
AWQMarlinLinearMethod
(
self
)
return
AWQMarlinLinearMethod
(
self
)
elif
isinstance
(
layer
,
FusedMoE
):
elif
isinstance
(
layer
,
FusedMoE
):
return
AWQMoEMethod
(
self
)
return
AWQMoEMethod
(
self
)
return
None
return
None
original_awq_moe_method_apply
=
AWQMoEMethod
.
apply
def
awq_moe_method_apply
(
self
,
layer
:
torch
.
nn
.
Module
,
x
:
torch
.
Tensor
,
router_logits
:
torch
.
Tensor
,
top_k
:
int
,
renormalize
:
bool
,
use_grouped_topk
:
bool
=
False
,
topk_group
:
Optional
[
int
]
=
None
,
num_expert_group
:
Optional
[
int
]
=
None
,
custom_routing_function
:
Optional
[
Callable
]
=
None
,
scoring_func
:
str
=
"softmax"
,
e_score_correction_bias
:
Optional
[
torch
.
Tensor
]
=
None
,
**
kwargs
,
):
return
original_awq_moe_method_apply
(
self
,
layer
,
x
,
router_logits
,
top_k
,
renormalize
,
use_grouped_topk
,
topk_group
,
num_expert_group
,
custom_routing_function
,
scoring_func
,
e_score_correction_bias
,
)
def
patch_vllm_linear_base_isinstance
():
def
patch_vllm_linear_base_isinstance
():
import
builtins
import
builtins
...
@@ -107,8 +150,11 @@ def patch_vllm_linear_base_isinstance():
...
@@ -107,8 +150,11 @@ def patch_vllm_linear_base_isinstance():
def
apply_monkey_patches
():
def
apply_monkey_patches
():
"""Apply all monkey patches in one place."""
"""Apply all monkey patches in one place."""
from
vllm.model_executor.layers.quantization.awq_marlin
import
AWQMoEMethod
setattr
(
GPTQMarlinConfig
,
"get_quant_method"
,
gptq_get_quant_method
)
setattr
(
GPTQMarlinConfig
,
"get_quant_method"
,
gptq_get_quant_method
)
setattr
(
AWQMarlinConfig
,
"get_quant_method"
,
awq_get_quant_method
)
setattr
(
AWQMarlinConfig
,
"get_quant_method"
,
awq_get_quant_method
)
setattr
(
AWQMoEMethod
,
"apply"
,
awq_moe_method_apply
)
patch_vllm_linear_base_isinstance
()
patch_vllm_linear_base_isinstance
()
...
...
python/sglang/srt/models/deepseek_v2.py
View file @
8616357a
...
@@ -255,6 +255,8 @@ class DeepseekV2Attention(nn.Module):
...
@@ -255,6 +255,8 @@ class DeepseekV2Attention(nn.Module):
self
.
kv_lora_rank
+
self
.
qk_rope_head_dim
,
self
.
kv_lora_rank
+
self
.
qk_rope_head_dim
,
bias
=
False
,
bias
=
False
,
quant_config
=
quant_config
,
quant_config
=
quant_config
,
# FIXME: quick fix for skip quantization
prefix
=
f
"self_attn.kv_a_proj_with_mqa"
,
)
)
self
.
kv_a_layernorm
=
RMSNorm
(
self
.
kv_lora_rank
,
eps
=
config
.
rms_norm_eps
)
self
.
kv_a_layernorm
=
RMSNorm
(
self
.
kv_lora_rank
,
eps
=
config
.
rms_norm_eps
)
self
.
kv_b_proj
=
ColumnParallelLinear
(
self
.
kv_b_proj
=
ColumnParallelLinear
(
...
@@ -455,6 +457,8 @@ class DeepseekV2AttentionMLA(nn.Module):
...
@@ -455,6 +457,8 @@ class DeepseekV2AttentionMLA(nn.Module):
self
.
kv_lora_rank
+
self
.
qk_rope_head_dim
,
self
.
kv_lora_rank
+
self
.
qk_rope_head_dim
,
bias
=
False
,
bias
=
False
,
quant_config
=
quant_config
,
quant_config
=
quant_config
,
# FIXME: quick fix for skip quantization
prefix
=
f
"self_attn.kv_a_proj_with_mqa"
,
)
)
self
.
kv_a_layernorm
=
RMSNorm
(
self
.
kv_lora_rank
,
eps
=
config
.
rms_norm_eps
)
self
.
kv_a_layernorm
=
RMSNorm
(
self
.
kv_lora_rank
,
eps
=
config
.
rms_norm_eps
)
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment