Arthur Wu

I looked at the llama source code and there is an intermedie layer that palm doesn't have or I didn't find it? Can you point it out clearly? Or how...

memory-efficient attention is default opened? if i dont use flash attn

3

or if i want use memory-efficient attention, i must call scaled_dot_product_attention? PyTorch 2.0 includes an optimized and memory-efficient attention implementation through the [torch.nn.functional.scaled_dot_product_attention](https://pytorch.org/docs/master/generated/torch.nn.functional.scaled_dot_product_attention) function

train your reward model issue

1

can't train reward model with batch seq, prompt_mask, labels = next(train_loader) loss = reward_model(seq, prompt_mask = prompt_mask, labels = labels) accelerator.backward(loss / GRADIENT_ACCUMULATE_EVERY) i set this but i get error...

After the upload is started in the sample code, the dialog box cannot work. How can it work at the same time? Can a new process be started?

there is no text in login page

4


while login， get Unable to sign in.

4

while login， get Unable to sign in. and my code same as example import chainlit as cl @cl.password_auth_callback def auth_callback(username: str, password: str): # Fetch the user matching username from...

i want build login page, How to set css style same as chainlit? any example?

try run code: `./train_all_data_at_once.sh Transformer` then get error ``` Traceback (most recent call last): File "run.py", line 352, in main() File "run.py", line 329, in main train(args, model, tokenizer, device,...

只传入text时，如何固定说话人？

5

payload = { "text": text, "reference_text": None, "reference_audio": None, "max_new_tokens": 0, "chunk_length": 30, "top_k": 0, "top_p": 0.7, "repetition_penalty": 1.5, "temperature": 0.7, "speaker": "纳西妲", "format": "wav" } 这样请求还是会有不同说话人的声音，要求：不传入原始音频，仅传入text，固定某一个说话人(女声)

llama_model_load: error loading model: create_tensor: tensor 'output.weight' not found

1

loaded openchat-3.5-0106-gemma-Q5_K_M.gguf ,then get error. it's newest version of llama.cpp