virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-10 06:10:52 +02:00
Code Issues Releases Activity

libavfi/dnn: enable LibTorch xpu device option support

Browse Source 
Add xpu device support to libtorch backend.
To enable xpu support you need to add
 "-Wl,--no-as-needed -lintel-ext-pt-gpu -Wl,--as-needed" to
"--extra-libs" when configure ffmpeg.

Signed-off-by: Wenbin Chen <wenbin.chen@intel.com>
This commit is contained in:
Wenbin Chen
2024-06-03 13:09:35 +08:00
committed by Guo Yejun
parent f68f40736f
commit 7560db937d
1 changed files with 15 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
libavfilter/dnn/dnn_backend_torch.cpp
View File

@@ -250,6 +250,10 @@ static int th_start_inference(void *args)
av_log(ctx, AV_LOG_ERROR, "input or output tensor is NULL\n"); av_log(ctx, AV_LOG_ERROR, "input or output tensor is NULL\n");
return DNN_GENERIC_ERROR; return DNN_GENERIC_ERROR;
} }
// Transfer tensor to the same device as model
c10::Device device = (*th_model->jit_model->parameters().begin()).device();
if (infer_request->input_tensor->device() != device)
*infer_request->input_tensor = infer_request->input_tensor->to(device);
inputs.push_back(*infer_request->input_tensor); inputs.push_back(*infer_request->input_tensor);
*infer_request->output = th_model->jit_model->forward(inputs).toTensor(); *infer_request->output = th_model->jit_model->forward(inputs).toTensor();
@@ -285,6 +289,9 @@ static void infer_completion_callback(void *args) {
switch (th_model->model.func_type) { switch (th_model->model.func_type) {
case DFT_PROCESS_FRAME: case DFT_PROCESS_FRAME:
if (task->do_ioproc) { if (task->do_ioproc) {
// Post process can only deal with CPU memory.
if (output->device() != torch::kCPU)
*output = output->to(torch::kCPU);
outputs.scale = 255; outputs.scale = 255;
outputs.data = output->data_ptr(); outputs.data = output->data_ptr();
if (th_model->model.frame_post_proc != NULL) { if (th_model->model.frame_post_proc != NULL) {
@@ -424,7 +431,13 @@ static DNNModel *dnn_load_model_th(DnnContext *ctx, DNNFunctionType func_type, A
th_model->ctx = ctx; th_model->ctx = ctx;
c10::Device device = c10::Device(device_name); c10::Device device = c10::Device(device_name);
if (!device.is_cpu()) { if (device.is_xpu()) {
if (!at::hasXPU()) {
av_log(ctx, AV_LOG_ERROR, "No XPU device found\n");
goto fail;
}
at::detail::getXPUHooks().initXPU();
} else if (!device.is_cpu()) {
av_log(ctx, AV_LOG_ERROR, "Not supported device:\"%s\"\n", device_name); av_log(ctx, AV_LOG_ERROR, "Not supported device:\"%s\"\n", device_name);
goto fail; goto fail;
} }
@@ -432,6 +445,7 @@ static DNNModel *dnn_load_model_th(DnnContext *ctx, DNNFunctionType func_type, A
try { try {
th_model->jit_model = new torch::jit::Module; th_model->jit_model = new torch::jit::Module;
(*th_model->jit_model) = torch::jit::load(ctx->model_filename); (*th_model->jit_model) = torch::jit::load(ctx->model_filename);
th_model->jit_model->to(device);
} catch (const c10::Error& e) { } catch (const c10::Error& e) {
av_log(ctx, AV_LOG_ERROR, "Failed to load torch model\n"); av_log(ctx, AV_LOG_ERROR, "Failed to load torch model\n");
goto fail; goto fail;