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

[Paddle-Inference] fix pass and convert_op for preln_ernie #39733

Merged
merged 4 commits into from
Feb 22, 2022
Merged

[Paddle-Inference] fix pass and convert_op for preln_ernie #39733

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
b2ecdaf
fix pass and convert_op for preln_ernie
Wangzheee Feb 18, 2022
78fcfbf
fix pass and convert_op for preln_ernie
Wangzheee Feb 21, 2022
5955a4e
fix pass and convert_op for preln_ernie and add preln_ernie'flag in pass
Wangzheee Feb 21, 2022
0ed95bc
fix pass and convert_op for preln_ernie and add preln_ernie'flag in pass
Wangzheee Feb 21, 2022
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
10 changes: 7 additions & 3 deletions paddle/fluid/framework/ir/preln_skip_layernorm_fuse_pass.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,6 @@ struct PrelnSkipLayerNorm : public PatternBase {
void operator()(PDNode *x, PDNode *y);

// declare operator node's name
PATTERN_DECL_NODE(fused_skipe_layernorm);
PATTERN_DECL_NODE(elementwise);
PATTERN_DECL_NODE(layer_norm);
// declare variable node's name
Expand All @@ -62,8 +61,13 @@ void PrelnSkipLayerNorm::operator()(PDNode *x, PDNode *y) {
auto *elementwise_out_var = pattern->NewNode(elementwise_out_repr())
->assert_is_op_output("elementwise_add")
->assert_is_op_input("layer_norm", "X")
->assert_is_op_input("elementwise_add", "Y");

->assert_more([](Node *x) {
if (x->outputs.size() == 2) {
return true;
} else {
return false;
}
});
// Add links for elementwise_add op.
elementwise->LinksFrom({x, y}).LinksTo({elementwise_out_var});

Expand Down
8 changes: 4 additions & 4 deletions paddle/fluid/framework/ir/skip_layernorm_fuse_pass.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,6 @@ struct SkipLayerNorm : public PatternBase {
PDNode *operator()(PDNode *x, PDNode *y);

// declare operator node's name
PATTERN_DECL_NODE(fused_skipe_layernorm);
PATTERN_DECL_NODE(elementwise);
PATTERN_DECL_NODE(layer_norm);
// declare variable node's name
Expand All @@ -59,9 +58,10 @@ PDNode *SkipLayerNorm::operator()(PDNode *x, PDNode *y) {
y->assert_is_op_input("elementwise_add", "Y");
auto *elementwise =
pattern->NewNode(elementwise_repr())->assert_is_op("elementwise_add");
auto *elementwise_out_var = pattern->NewNode(elementwise_out_repr())
->AsOutput()
->assert_is_op_output("elementwise_add");
auto *elementwise_out_var =
pattern->NewNode(elementwise_out_repr())
->AsOutput()
->assert_is_only_output_of_op("elementwise_add");

// Add links for elementwise_add op.
elementwise->LinksFrom({x, y}).LinksTo({elementwise_out_var});
Expand Down
18 changes: 4 additions & 14 deletions paddle/fluid/inference/tensorrt/convert/preln_emb_eltwise_layernorm.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,21 +51,11 @@ class PrelnEmbEltwiseLayerNormOpConverter : public OpConverter {
auto pos_emb_name = op_desc.Input("PosEmbedding").front();
auto sent_emb_name = op_desc.Input("SentEmbedding").front();

std::vector<std::string> id_names;
std::vector<std::string> emb_names;

id_names =
std::vector<std::string>{word_id_name, pos_id_name, sent_id_name};
emb_names =
std::vector<std::string>{word_emb_name, pos_emb_name, sent_emb_name};

int input_num = id_names.size();

// Declare inputs
std::vector<nvinfer1::ITensor*> input_ids;
for (int i = 0; i < input_num; i++) {
input_ids.push_back(engine_->GetITensor(id_names[i]));
}
int input_num = emb_names.size();

// input_embs[0]: word_embedding
// input_embs[1]: pos_embedding
Expand Down Expand Up @@ -126,7 +116,7 @@ class PrelnEmbEltwiseLayerNormOpConverter : public OpConverter {
{"bert_embeddings_position_embeddings", input_embs[1],
nvinfer1::PluginFieldType::kFLOAT32,
static_cast<int32_t>(emb_sizes[1])},
{"output_int8", &output_int8, nvinfer1::PluginFieldType::kINT32, 1},
{"output_fp16", &output_int8, nvinfer1::PluginFieldType::kINT32, 1},
};

nvinfer1::PluginFieldCollection* plugin_ptr =
Expand Down Expand Up @@ -156,7 +146,7 @@ class PrelnEmbEltwiseLayerNormOpConverter : public OpConverter {
shuffle_layer->setReshapeDimensions(shape_dim);
shuffle_layer->setName(
("PrelnEmbeltwise_Shuffle_reshape (Output: max_seqlen " +
op_desc.Output("Out")[0] + ")")
op_desc.Output("Out_0")[0] + ")")
.c_str());
engine_->SetTensorDynamicRange(shuffle_layer->getOutput(0), 1.0f);
plugin_inputs.emplace_back(
Expand All @@ -170,7 +160,7 @@ class PrelnEmbEltwiseLayerNormOpConverter : public OpConverter {
auto plugin_layer = engine_->network()->addPluginV2(
plugin_inputs.data(), plugin_inputs.size(), *plugin_obj);
plugin_layer->setName(("CustomPrelnEmbLayerNormPluginDynamic_V3(Output: " +
op_desc.Output("Out")[0] + ")")
op_desc.Output("Out_0")[0] + ")")
.c_str());
free(plugin_ptr);
float out_0_scale =
Expand Down
6 changes: 4 additions & 2 deletions paddle/fluid/inference/tensorrt/convert/preln_skip_layernorm.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,8 +92,10 @@ class PrelnSkipLayerNormOpConverter : public OpConverter {
"fail to add CustomPrelnSkipLayerNormPluginDynamic layer"));
layer = plugin_layer;

auto output_name = op_desc.Output("Out")[0];
RreplenishLayerAndOutput(layer, "preln_skip_layernorm", {output_name},
std::vector<std::string> output_names;
output_names.push_back(op_desc.Output("Out_0")[0]);
output_names.push_back(op_desc.Output("Out_1")[0]);
RreplenishLayerAndOutput(layer, "preln_skip_layernorm", {output_names},
test_mode);
#else
PADDLE_THROW(platform::errors::Fatal(
Expand Down