-
```
What steps will reproduce the problem?
1. Set "buff_size" to 9000 bytes
echo 9000 > /sys/module/netmap_lin/parameters/buf_size
2. Launch tcpdump using netmap-libpcap
sudo taskset -c 0 sudo tcpdu…
-
```
What steps will reproduce the problem?
1. Set "buff_size" to 9000 bytes
echo 9000 > /sys/module/netmap_lin/parameters/buf_size
2. Launch tcpdump using netmap-libpcap
sudo taskset -c 0 sudo tcpdu…
-
```
What steps will reproduce the problem?
1. Set "buff_size" to 9000 bytes
echo 9000 > /sys/module/netmap_lin/parameters/buf_size
2. Launch tcpdump using netmap-libpcap
sudo taskset -c 0 sudo tcpdu…
-
```
What steps will reproduce the problem?
1. Set "buff_size" to 9000 bytes
echo 9000 > /sys/module/netmap_lin/parameters/buf_size
2. Launch tcpdump using netmap-libpcap
sudo taskset -c 0 sudo tcpdu…
-
```
What steps will reproduce the problem?
1. Set "buff_size" to 9000 bytes
echo 9000 > /sys/module/netmap_lin/parameters/buf_size
2. Launch tcpdump using netmap-libpcap
sudo taskset -c 0 sudo tcpdu…
-
```
What steps will reproduce the problem?
1. Set "buff_size" to 9000 bytes
echo 9000 > /sys/module/netmap_lin/parameters/buf_size
2. Launch tcpdump using netmap-libpcap
sudo taskset -c 0 sudo tcpdu…
-
```
What steps will reproduce the problem?
1. Set "buff_size" to 9000 bytes
echo 9000 > /sys/module/netmap_lin/parameters/buf_size
2. Launch tcpdump using netmap-libpcap
sudo taskset -c 0 sudo tcpdu…
-
### 🐛 Describe the bug
**i'm working with a GCN model using Pytorch_geometric , i created my costom dataset with this shape**
Data(x=[16, 7], edge_index=[2, 28], edge_attr=[28, 3], y=[1280], z=[4…
-
Hello.
First time user of fsharp/mono.
Installed on my stock raspbian(debian stretch on Pi 3) system via the instructions on
https://www.mono-project.com/download/stable/#download-lin-debian
and…
-
After getting help from [this](https://github.com/TensorSpeech/TensorFlowASR/issues/168), I successfully trained a stream-ransducer model and converted the model to tflite.
I tested the trained mod…