codellama
codellama copied to clipboard
Inference code for CodeLlama models
Does Codellama support code review, comment filling, bug fixing? Or can only do code filling and code generation?
I have a question regarding my fine-tuning pipeline, specifically concerning a memory usage spike when the model saves checkpoint during the training step. This cause sudden CUDA Memory error. I...
as describe in titlle,much appreciate
Is it possible to continue the initial autoregressive pre-training on a custom dataset, as was done for Code Llama - Python? This would in principle allow for the fine-tuning of...
Enhancement on file: download.sh #51
![image](https://github.com/facebookresearch/codellama/assets/57563332/b0854b14-4f25-41d4-9de0-95ac63b37807) The bash command and output information are shown above. I wonder if you have encountered this problem or have any idea to solve it. Thanks!
root@main:/mnt/d/repos/LLAMA2/llama-main# ./setup.py ./setup.py: line 4: from: command not found ./setup.py: line 7: syntax error near unexpected token `(' ./setup.py: line 7: `def get_requirements(path: str):'
code: torchrun --nproc_per_node 2 example_instructions.py --ckpt_dir CodeLlama-13b-Instruct/ --tokenizer_path CodeLlama-13b-Instruct/tokenizer.model --max_seq_len 2048 --max_batch_size 4 checkpoint = torch.load(ckpt_path, map_location="cpu") --> checkpoint = torch.load(ckpt_path, map_location="cuda") 我将cpu改为cuda后,会报下列错误 torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to...
Hi All, when I ran an official example like this : torchrun --nproc_per_node 1 example_instructions.py --ckpt_dir CodeLlama-7b-Instruct/ --tokenizer_path CodeLlama-7b-Instruct/tokenizer.model --max_seq_len 192 --max_batch_size 4 I got a result thant can not...
i found this https://github.com/facebookresearch/codellama/issues/55 but is closed,and no solved. # env: win10+conda(pytorch-gpu+python3.11)+powershell # error: ``` (pytorch-gpu) PS F:\aiProject\codellama> torchrun --nproc_per_node 1 example_completion.py --ckpt_dir .\CodeLlama-34b-Python\ --tokenizer_path .\CodeLlama-34b-Python\tokenizer.model --max_seq_len 512 --max_batch_size 4...