-
-
First of all, thank you for your amazing work on the nnScaler project. It has been incredibly inspiring, and I’ve been learning and using the contents from this repository in my own work.
I have a fe…
-
## Hardware
* [ ] ESP8266
* [x] ESP32
* [ ] Raspberry Pi
### nRF24L01+ Module
* [x] nRF24L01+ you verified this is a **Plus model** capable of the required 256kBit/s mode
* [ ] **square …
-
Moe workload generated by AICB using the following command cannot be parsed:
```bash
sh scripts/megatron_gpt.sh \
--nnodes 1 --node_rank 0 --nproc_per_node 8 --master_addr localhost --master_port 2…
-
I'm using a sample file in the Examples/Async-Server directory.
I need to send very large response files.
Currently, communication looks like this:
1) receiving the request from the client 1
2…
-
### 🐛 Describe the bug
Hello,
I'm a new user of PyTorch and recently tried to run the Flight Recorder code provided in the tools. But I cannot get the code to execute as expected.
I use ngc 24.10…
-
Since https://github.com/FEniCS/dolfinx/pull/1417/ we need to assemble (parallel comm) the matrix after regular assembly and before setting a local diagonal, set diagonal, and the assemble (parallel c…
-
For example, reward model has 8 GPU cards with TP and DP configuration. Actor model might have TP&PP&DP(just for example) occupying 64 GPUs.
How do you connect Actor's last stage's output to reward m…
-
- To detect correct communication with the MAX31855, a pull-up resistor must be added between the MISO pin and VCC.
Without a pull-up, random noise may appear that might look like real data.
…
-
**Your question**
1. I have a question about creating the `pp` groups when enabling `context_parallel_size > 1` and `encoder_tensor_parallel_size != tensor_parallel_size`.
When enabling `context_para…