Tran Thanh Luan

> I also meet this, could anyone answer?  I solved my problem. Add map_location, "cpu" when training, "cuda" when testing.

@ermenkov i resolved the `FaucetDisabled` exception by changing `cargo build --release` to `cargo build --release --features pow-faucet` in `scripts/localrun.sh` but then i encounter another issue when create new subnet: 

@ermenkov i resolved, just comment all rate limit exceptions in source code. it's ok for me cus i just want to run locally

@calclavia same here

@ibraheem-opentensor thanks, i gonna try it

@DarkLight1337 @Nan2018 thanks guys, this is essential pr. Can you guys make it to be supported by OpenAI Compatible Endpoint? So that, it would be convenient to call it by...

> > @DarkLight1337 @Nan2018 thanks guys, this is essential pr. Can you guys make it to be supported by OpenAI Compatible Endpoint? So that, it would be convenient to call...

@Zyvpeng thanks for the response. I see in the paper you mention that loss is under 0.1 So 0.xx here should be 0.0x? I have made progress in training it...

@Kirili4ik There is a trick that you set position identifier for soft-tokens can make model converge quickly https://arxiv.org/pdf/2409.14364v2 Example: - context positions: [1,2,3,4,5,6,7,8] - soft-tokens icae positions: [9, 10, 11,...

@qyy2003 It looks great I assume with this future, i can manually create kv cache tensors: (n_layers, 2, batch_size, seq_len, n_head, head_dim) then save it to disk with random hash...