LalchandPandia

Few shot prompting

3

Hi, Which model (either chat or text-completion) should be used for in-context learning using few-shot prompting?

Which data can be used for GPU machine utilization as in section 4.1 for P90 GPU utilization

Greedy sampling gives a warning message

5

### System Info With new version starting from 4.39, performing greedy search gives a warning: You should set do_sample=True or unset temperature. I am loading pretrained llama2-7b-chat-hf model. I understand...

Fine-tuning error in conda environment without docker image

8

## Environment python 3.11.9 cuda 11.8 torch 2.4.0+cu118 PyTorch information ------------------- PyTorch version: 2.4.0+cu118 Is debug build: False CUDA used to build PyTorch: 11.8 ROCM used to build PyTorch: N/A...

Special token gets tokenized while training tokenizer from scratch

1

@ArthurZucker I am trying to train a bytepiece tokenizer on my dataset. I have a list of words which I want to be treated as a single token. But when...

Identical perplexity/loss for instances during inference

1

Hi, I have fine-tuned OPT-125M using the finetune.py in bf16. But when I load the model and try to calculate the perplexity of each instance, I am getting same perplexities...

Gradient of the loss w.r.t sharded parameters

2

**Describe the bug** Getting gradient of loss during inference as None. I am fine-tuning llama 2 using accelerate+deepseed zero3. During evaluation, which is run after every checkpoint step, I need...