wehub/nvidia--tensorrt
1
0
Fork 0
Code Issues Pull Requests Actions 26 Deployments Models Agent Notes Packages Projects Releases Wiki Activity

chore: import upstream snapshot with attribution
Docker Image CI / build-ubuntu2004 (push) Has been cancelled
Details

Browse Source
This commit is contained in:
wehub-resource-sync
2026-07-13 13:36:55 +08:00
commit c8a779b1bb
1887 changed files with 3245738 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
plugin/bertQKVToContextPlugin/CustomQKVToContextPluginDynamic_PluginConfig.yaml
+216
View File
@@ -0,0 +1,216 @@
#
# SPDX-FileCopyrightText: Copyright (c) 2022-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
---
name: CustomQKVToContextPluginDynamic
interface: "IPluginV3"
versions:
"4":
inputs:
- input
outputs:
- output
input_dims:
input: 5
output_dims:
output: "input_0, input_1, hidden_size_0, 1, 1"
attributes:
- type_id
- hidden_size
- num_heads
- has_mask
- dq_probs
attribute_types:
type_id: int32
hidden_size: int32
num_heads: int32
has_mask: int32
dq_probs: float32
attribute_length:
type_id: 1
hidden_size: 1
num_heads: 1
has_mask: 1
dq_probs: 1
attribute_options:
type_id:
- 0
- 1
- 2
hidden_size:
min: "0"
max: "=pinf"
num_heads:
min: "0"
max: "=pinf"
has_mask:
- 0
- 1
dq_probs:
min: "0"
max: "=pinf"
attribute_constraints:
- "num_heads * k == hidden_size"
attributes_required:
- type_id
- hidden_size
- num_heads
- has_mask
golden_io_path: "plugin/CustomQKVToContextPluginDynamic_PluginGoldenIO.json"
abs_tol: 1e-5
rel_tol: 1e-5
fp16_atol: 1e-2
fp16_rtol: 1e-2
configs:
config1:
input_types:
input: float16
attribute_options:
type_id:
value: 1
shape: "1"
hidden_size:
value: 768
shape: "1"
num_heads:
value: 12
shape: "1"
has_mask:
value: 0
shape: "1"
output_types:
output: float16
config2:
input_types:
input: float16
attribute_options:
type_id:
value: 1
shape: "1"
hidden_size:
value: 1024
shape: "1"
num_heads:
value: 16
shape: "1"
has_mask:
value: 0
shape: "1"
output_types:
output: float16
config5:
input_types:
input: float16
attribute_options:
type_id:
value: 1
shape: "1"
hidden_size:
value: 384
shape: "1"
num_heads:
value: 12
shape: "1"
has_mask:
value: 0
shape: "1"
output_types:
output: float16
has_mask:
input_types:
input: float32
input_mask: int32
attribute_options:
type_id:
value: 0
shape: "1"
hidden_size:
value: 3
shape: "1"
num_heads:
value: 3
shape: "1"
has_mask:
value: 1
shape: "1"
output_types:
output: float32
"5":
inputs:
- input
- input_mask
outputs:
- output
input_dims:
input: 5
input_mask: 1
output_dims:
output: "input_0, input_mask_0, hidden_size_0, 1, 1"
attributes:
- type_id
- hidden_size
- num_heads
- has_mask
- dq_probs
- var_seqlen
- use_int8_scale_max
attribute_types:
type_id: int32
hidden_size: int32
num_heads: int32
has_mask: int32
dq_probs: float32
var_seqlen: int32
use_int8_scale_max: int32
attribute_length:
type_id: 1
hidden_size: 1
num_heads: 1
has_mask: 1
dq_probs: 1
var_seqlen: 1
use_int8_scale_max: 1
attribute_options:
type_id:
- 0
- 1
- 2
hidden_size:
min: "0"
max: "=pinf"
num_heads:
min: "0"
max: "=pinf"
has_mask:
- 0
- 1
dq_probs:
min: "0"
max: "=pinf"
use_int8_scale_max:
- 0
- 1
var_seqlen:
- 0
# - 1 # Disabled since SM specific tests are not supported yet
attributes_required:
- type_id
- hidden_size
- num_heads
- has_mask
# "3": #version3 - Not supported yet
...