[FFmpeg-devel] [PATCH v3 1/2] avfilter/dnn/dnn_backend_tf: simplify the code with ff_hex_to_data
Guo, Yejun
yejun.guo at intel.com
Wed Apr 28 15:26:54 EEST 2021
> -----Original Message-----
> From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of
> lance.lmwang at gmail.com
> Sent: 2021年4月28日 18:47
> To: ffmpeg-devel at ffmpeg.org
> Cc: Limin Wang <lance.lmwang at gmail.com>
> Subject: [FFmpeg-devel] [PATCH v3 1/2] avfilter/dnn/dnn_backend_tf:
> simplify the code with ff_hex_to_data
>
> From: Limin Wang <lance.lmwang at gmail.com>
>
> please use tools/python/tf_sess_config.py to get the sess_config after that.
> note the byte order of session config is in normal order.
> bump the MICRO version for the config change.
>
> Signed-off-by: Limin Wang <lance.lmwang at gmail.com>
> ---
> libavfilter/dnn/dnn_backend_tf.c | 42 +++++++------------------------------
> libavfilter/version.h | 2 +-
> tools/python/tf_sess_config.py | 45
> ++++++++++++++++++++++++++++++++++++++++
> 3 files changed, 54 insertions(+), 35 deletions(-)
> create mode 100644 tools/python/tf_sess_config.py
>
> diff --git a/libavfilter/dnn/dnn_backend_tf.c
> b/libavfilter/dnn/dnn_backend_tf.c
> index fb799d2..076dd3d 100644
> --- a/libavfilter/dnn/dnn_backend_tf.c
> +++ b/libavfilter/dnn/dnn_backend_tf.c
> @@ -28,6 +28,7 @@
> #include "dnn_backend_native_layer_conv2d.h"
> #include "dnn_backend_native_layer_depth2space.h"
> #include "libavformat/avio.h"
> +#include "libavformat/internal.h"
> #include "libavutil/avassert.h"
> #include "../internal.h"
> #include "dnn_backend_native_layer_pad.h"
> @@ -206,53 +207,26 @@ static DNNReturnType load_tf_model(TFModel
> *tf_model, const char *model_filename
>
> // prepare the sess config data
> if (tf_model->ctx.options.sess_config != NULL) {
> + const char *config;
> /*
> tf_model->ctx.options.sess_config is hex to present the serialized
> proto
> required by TF_SetConfig below, so we need to first generate the
> serialized
> - proto in a python script, the following is a script example to
> generate
> - serialized proto which specifies one GPU, we can change the script
> to add
> - more options.
> -
> - import tensorflow as tf
> - gpu_options = tf.GPUOptions(visible_device_list='0')
> - config = tf.ConfigProto(gpu_options=gpu_options)
> - s = config.SerializeToString()
> - b = ''.join("%02x" % int(ord(b)) for b in s[::-1])
> - print('0x%s' % b)
> -
> - the script output looks like: 0xab...cd, and then pass 0xab...cd to
> sess_config.
> + proto in a python script, tools/python/tf_sess_config.py is a script
> example
> + to generate the configs of sess_config.
> */
> - char tmp[3];
> - tmp[2] = '\0';
> -
> if (strncmp(tf_model->ctx.options.sess_config, "0x", 2) != 0) {
> av_log(ctx, AV_LOG_ERROR, "sess_config should start with
> '0x'\n");
> return DNN_ERROR;
> }
> + config = tf_model->ctx.options.sess_config + 2;
> + sess_config_length = ff_hex_to_data(NULL, config);
>
> - sess_config_length = strlen(tf_model->ctx.options.sess_config);
> - if (sess_config_length % 2 != 0) {
> - av_log(ctx, AV_LOG_ERROR, "the length of sess_config is not
> even (%s), "
> - "please re-generate the
> config.\n",
> -
> tf_model->ctx.options.sess_config);
> - return DNN_ERROR;
> - }
> -
> - sess_config_length -= 2; //ignore the first '0x'
> - sess_config_length /= 2; //get the data length in byte
> -
> - sess_config = av_malloc(sess_config_length);
> + sess_config = av_mallocz(sess_config_length +
> AV_INPUT_BUFFER_PADDING_SIZE);
just get a concern, why we need to add PADDING_SIZE here.
Will there be potential issue if not add?
> if (!sess_config) {
> av_log(ctx, AV_LOG_ERROR, "failed to allocate memory\n");
> return DNN_ERROR;
> }
> -
> - for (int i = 0; i < sess_config_length; i++) {
> - int index = 2 + (sess_config_length - 1 - i) * 2;
> - tmp[0] = tf_model->ctx.options.sess_config[index];
> - tmp[1] = tf_model->ctx.options.sess_config[index + 1];
> - sess_config[i] = strtol(tmp, NULL, 16);
> - }
> + ff_hex_to_data(sess_config, config);
> }
>
More information about the ffmpeg-devel
mailing list