fix: qnn rope file name

UbiquitousLearning · Jan 4, 2024 · 9e9bb8b · 9e9bb8b
1 parent 9b2ab1b
commit 9e9bb8b
Show file tree

Hide file tree

Showing 4 changed files with 79 additions and 98 deletions.
diff --git a/src/backends/QNN/op/QNNLinear.cpp b/src/backends/QNN/op/QNNLinear.cpp
@@ -63,6 +63,7 @@ ErrorCode QNNLinear::setUp(vector<shared_ptr<Tensor>> inputs, vector<shared_ptr<
     for (int i = 0; i < 4; i++) {
         dimensionsWeight[i] = weight_.shape()[i];
     }
+    weight_.fullData(2.f);
     qnnBackend_->modelAddTensor(weight_.name().c_str(), (Qnn_Tensor_t){
                                                             .version = QNN_TENSOR_VERSION_1,
                                                             {.v1 = {

diff --git a/src/backends/QNN/op/QNNRoPE.cpp b/src/backends/QNN/op/QNNRoPE.cpp
@@ -9,11 +9,88 @@ QNNRoPE::QNNRoPE(Backend *bn, string opName) :
 }
 
 ErrorCode QNNRoPE::reshape(vector<shared_ptr<Tensor>> inputs, vector<shared_ptr<Tensor>> outputs) {
+    outputs[0]->reshape(inputs[0]->batch(), inputs[0]->head(), inputs[0]->sequence(), inputs[0]->dimension());
+
     return NO_ERROR;
 }
 
 ErrorCode QNNRoPE::setUp(vector<shared_ptr<Tensor>> inputs, vector<shared_ptr<Tensor>> outputs) {
-    return graphAddNode(name(), "RoPE", inputs, outputs, {}, "LLaMAOpPackageHtp");
+
+    pos_max_ = 16384;
+
+    ishape = inputs[0]->dimension();
+
+    if(!sin_.allocted()) {
+        if (pose_type_ == 1) {
+            sinusoidal_position_embedding_hf(1, 1, pos_max_, ishape, sin_, cos_);
+        } else if (pose_type_ == 2) {
+            sinusoidal_position_embedding(1, 1, pos_max_, ishape, sin_, cos_);
+        } else {
+            sinusoidal_position_embedding_hf(1, 1, pos_max_, ishape/2, sin_, cos_);
+        }
+    }
+
+    // vector<Tensor> qnn_RoPE_inputs;
+    // qnn_RoPE_inputs.push_back(inputs[0])
+    // qnn_RoPE_inputs.push_back(sin_);
+    // qnn_RoPE_inputs.push_back(cos_);
+
+    // return graphAddNode(name(), "RoPE", qnn_RoPE_inputs, outputs);
+
+    return NO_ERROR;
 }
+
+
+void QNNRoPE::sinusoidal_position_embedding(int batch_size, int nums_head, int seq_len, int output_dim, Tensor &sin, Tensor &cos) {
+    sin.reshape(batch_size, nums_head, seq_len, output_dim);
+    cos.reshape(batch_size, nums_head, seq_len, output_dim);
+    sin.alloc();
+    cos.alloc();
+    for (int n = 0; n < batch_size; ++n) {
+        for (int h = 0; h < nums_head; ++h) {
+            for (int s = 0; s < seq_len; ++s) {
+                for (int d = 0; d < output_dim; d += 2) {
+                    int i = (int)d / 2;
+                    float sin_value = std::sin(s / std::pow(10000, 2.0 * i / output_dim));
+                    float cos_value = std::cos(s / std::pow(10000, 2.0 * i / output_dim));
+                    sin.setDataAt<float>(n, h, s, d, sin_value);
+                    cos.setDataAt<float>(n, h, s, d, cos_value);
+                    if (d + 1 < output_dim) {
+                        sin.setDataAt<float>(n, h, s, d + 1, sin_value);
+                        cos.setDataAt<float>(n, h, s, d + 1, cos_value);
+                    }
+                }
+            }
+        }
+    }
+}
+
+void QNNRoPE::sinusoidal_position_embedding_hf(int batch_size, int nums_head, int seq_len, int output_dim, Tensor &sin, Tensor &cos) {
+    sin.reshape(batch_size, nums_head, seq_len, output_dim);
+    cos.reshape(batch_size, nums_head, seq_len, output_dim);
+    sin.alloc();
+    cos.alloc();
+    for (int n = 0; n < batch_size; ++n) {
+        for (int h = 0; h < nums_head; ++h) {
+            for (int s = 0; s < seq_len; ++s) {
+                for (int d = 0; d < output_dim; d += 2) {
+                    int i = (int)d;
+                    if (d >= (int)output_dim / 2) {
+                        i = (int)(d - output_dim / 2);
+                    }
+                    float sin_value = std::sin(s / std::pow(10000, 2.0 * i / output_dim));
+                    float cos_value = std::cos(s / std::pow(10000, 2.0 * i / output_dim));
+                    sin.setDataAt<float>(n, h, s, d, sin_value);
+                    cos.setDataAt<float>(n, h, s, d, cos_value);
+                    if (d + 1 < output_dim) {
+                        sin.setDataAt<float>(n, h, s, d + 1, sin_value);
+                        cos.setDataAt<float>(n, h, s, d + 1, cos_value);
+                    }
+                }
+            }
+        }
+    }
+}
+
 } // namespace mllm
 
diff --git a/src/backends/QNN/op/QNNRope.cpp b/src/backends/QNN/op/QNNRope.cpp
diff --git a/src/backends/QNN/op/QNNSiLU.cpp b/src/backends/QNN/op/QNNSiLU.cpp
@@ -9,7 +9,6 @@ QNNSiLU::QNNSiLU(Backend *bn, string opName) :
 }
 
 ErrorCode QNNSiLU::reshape(vector<shared_ptr<Tensor>> inputs, vector<shared_ptr<Tensor>> outputs) {
-<<<<<<< HEAD
     outputs[0]->reshape(inputs[0]->batch(), inputs[0]->head(), inputs[0]->sequence(), inputs[0]->dimension());
     return Op::reshape(inputs, outputs);
 }