Skip to content

Commit

Permalink
numbers matching through kernels and bias
Browse files Browse the repository at this point in the history
  • Loading branch information
@balisujohn
balisujohn committed Jun 17, 2024
1 parent 8748b21 commit fb587ba
Showing 1 changed file with 66 additions and 18 deletions.
84 changes: 66 additions & 18 deletions main.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -385,6 +385,12 @@ struct residual_conv_block{

};

struct conv_block{
struct ggml_tensor * conv_block_1_bias;
struct ggml_tensor * conv_block_1_weight;

};


struct vocoder_residual_block{

Expand All @@ -404,7 +410,10 @@ struct vocoder_residual_block{
struct ggml_tensor * convolution_t_pre_bias;


std::vector<conv_block> conv_blocks;


/*
struct ggml_tensor * conv_blocks_0_1_bias;
struct ggml_tensor * conv_blocks_0_1_weight;
Expand All @@ -415,7 +424,7 @@ struct vocoder_residual_block{
struct ggml_tensor * conv_blocks_2_1_weight;
struct ggml_tensor * conv_blocks_3_1_bias;
struct ggml_tensor * conv_blocks_3_1_weight;
struct ggml_tensor * conv_blocks_3_1_weight;*/

};

Expand Down Expand Up @@ -1671,14 +1680,6 @@ bool vocoder_model_load(const std::string & fname, vocoder_model & model)
model.residual_stack[i].convolution_t_pre_weight = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, 8, 32, 32);
}
model.residual_stack[i].convolution_t_pre_bias = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 32);
model.residual_stack[i].conv_blocks_0_1_weight = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, 3, 32, 32);
model.residual_stack[i].conv_blocks_0_1_bias = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 32);
model.residual_stack[i].conv_blocks_1_1_weight = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, 3, 32, 32);
model.residual_stack[i].conv_blocks_1_1_bias = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 32);
model.residual_stack[i].conv_blocks_2_1_weight = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, 3, 32, 32);
model.residual_stack[i].conv_blocks_2_1_bias = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 32);
model.residual_stack[i].conv_blocks_3_1_weight = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, 3, 32, 32);
model.residual_stack[i].conv_blocks_3_1_bias = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 32);

model.tensors["res_stack." + std::to_string(i) + ".kernel_predictor.input_conv.0.bias"] = model.residual_stack[i].kernel_predictor_input_convolution_bias;
model.tensors["res_stack." + std::to_string(i) + ".kernel_predictor.input_conv.0.weight"] = model.residual_stack[i].kernel_predictor_input_convolution_weight;
Expand All @@ -1688,14 +1689,18 @@ bool vocoder_model_load(const std::string & fname, vocoder_model & model)
model.tensors["res_stack." + std::to_string(i) + ".kernel_predictor.bias_conv.weight"] = model.residual_stack[i].kernel_predictor_bias_convolution_weight;
model.tensors["res_stack." + std::to_string(i) + ".convt_pre.1.weight"] = model.residual_stack[i].convolution_t_pre_weight;
model.tensors["res_stack." + std::to_string(i) + ".convt_pre.1.bias"] = model.residual_stack[i].convolution_t_pre_bias;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.0.1.weight"] = model.residual_stack[i].conv_blocks_0_1_weight;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.0.1.bias"] = model.residual_stack[i].conv_blocks_0_1_bias;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.1.1.weight"] = model.residual_stack[i].conv_blocks_1_1_weight;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.1.1.bias"] = model.residual_stack[i].conv_blocks_1_1_bias;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.2.1.weight"] = model.residual_stack[i].conv_blocks_2_1_weight;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.2.1.bias"] = model.residual_stack[i].conv_blocks_2_1_bias;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.3.1.weight"] = model.residual_stack[i].conv_blocks_3_1_weight;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks.3.1.bias"] = model.residual_stack[i].conv_blocks_3_1_bias;

model.residual_stack[i].conv_blocks.resize(4);

for(int c = 0; c < 4; c++)
{
model.residual_stack[i].conv_blocks[c].conv_block_1_weight = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, 3, 32, 32);
model.residual_stack[i].conv_blocks[c].conv_block_1_bias = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 32);

model.tensors["res_stack." + std::to_string(i) + ".conv_blocks."+std::to_string(c) +".1.weight"] = model.residual_stack[i].conv_blocks[c].conv_block_1_weight;
model.tensors["res_stack." + std::to_string(i) + ".conv_blocks."+std::to_string(c) +".1.bias"] = model.residual_stack[i].conv_blocks[c].conv_block_1_bias;

}
}

model.convolution_post_bias = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 1);
Expand Down Expand Up @@ -3830,6 +3835,7 @@ struct ggml_cgraph * vocoder_graph(
int paddings[] = {4,2,2};



struct ggml_tensor * conditioning;

//graph tether
Expand Down Expand Up @@ -3932,8 +3938,50 @@ struct ggml_cgraph * vocoder_graph(
bias = ggml_cpy(ctx0, bias, ggml_new_tensor(ctx0, GGML_TYPE_F32,4,bias->ne));




kernels = ggml_reshape_3d(ctx0, kernels, kernels->ne[0], 6144, 4);
bias = ggml_reshape_3d(ctx0, bias, bias->ne[0], 64, 4);


int conv_block_dilations[] = {1,3,9,27};
int conv_block_paddings[] = {1,3,9,27};

ggml_tensor * output;
ggml_tensor * k;
ggml_tensor * b;

for (int c = 0; c < 1; c++)
{
output = ggml_leaky_relu(ctx0, cur, 0.2, false);

float_16_conv_1d_weight= ggml_cpy(ctx0, model.residual_stack[i].conv_blocks[c].conv_block_1_weight, ggml_new_tensor(ctx0, GGML_TYPE_F16,4,model.residual_stack[i].conv_blocks[c].conv_block_1_weight->ne));

output = ggml_cont(ctx0,ggml_conv_1d(ctx0, float_16_conv_1d_weight, output, 1,conv_block_paddings[c],conv_block_dilations[c]));


output = ggml_cpy(ctx0, output, ggml_new_tensor(ctx0, GGML_TYPE_F32,4,output->ne));


output = ggml_cont(ctx0,ggml_transpose(ctx0,ggml_add(ctx0, ggml_cont(ctx0, ggml_transpose(ctx0, output)), model.residual_stack[i].conv_blocks[c].conv_block_1_bias)));


output = ggml_cpy(ctx0, output, ggml_new_tensor(ctx0, GGML_TYPE_F32,4,output->ne));

output = ggml_leaky_relu(ctx0, output, 0.2, false);

k = ggml_cont(ctx0,ggml_view_3d(ctx0, kernels, kernels->ne[0], 6144, 1, kernels->nb[1], kernels->nb[2], c * kernels->ne[0] * 6144 * sizeof(float)));
k = ggml_reshape_4d(ctx0,k,kernels->ne[0],3,64,32);

b = ggml_cont(ctx0,ggml_view_3d(ctx0, bias, bias->ne[0], 64, 1, bias->nb[1], bias->nb[2], c * bias->ne[0] * 64 * sizeof(float)));
b = ggml_reshape_2d(ctx0,b,bias->ne[0],64);

output = ggml_pad(ctx0, output, 1,0,0,0);


}

cur = bias;
cur = output;

}

Expand Down

0 comments on commit fb587ba

Please sign in to comment.