Shashank Sonkar

icon indicating a website link http://ss164.blogs.rice.edu/

icon company Rice University icon location Houston icon location Graduate student @ Rice University Interested in understanding, training, and optimizing Large Language Models (LLMs)

Results 7 comments of Shashank Sonkar

ModuleNotFoundError: No module named 'utility'

i used python setup.py install, and i get the same error !

ModuleNotFoundError: No module named 'utility'

conda env create -f conda_env.yml conda activate colbert-v0.2 are these the commands to install colbert?

ModuleNotFoundError: No module named 'utility'

I was interested in training v2 from scratch ! I think it is addressed in the other issue :) Thanks a lot for the help and prompt reply :)

How is the file "cloze_test_test__spring2016 - cloze_test_ALL_test.csv" created?

Thanks ! So, the model is not trained on the entire dataset "ROCStories__spring2016 - ROCStories_spring2016.csv"?

Filling zero after `optimizer.step()`

hello, i was wondering if the issue is fixed ? thank you for the code :)

Not all parameters were saved completely.

if you are using: https://github.com/lm-sys/FastChat/blob/main/fastchat/train/train.py, then in line 246: replace `safe_save_model_for_hf_trainer(trainer=trainer, output_dir=training_args.output_dir)` with ``` checkpoint_dir = os.path.join(training_args.output_dir, "checkpoint-final") trainer.deepspeed.save_checkpoint(checkpoint_dir) ``` then, checkpoint-final will contains zero_to_fp32.py after the training is done....

safe_save_model_for_hf_trainer function save mode is very small

if you are using: https://github.com/lm-sys/FastChat/blob/main/fastchat/train/train.py, then in line 246: replace `safe_save_model_for_hf_trainer(trainer=trainer, output_dir=training_args.output_dir)` with ``` checkpoint_dir = os.path.join(training_args.output_dir, "checkpoint-final") trainer.deepspeed.save_checkpoint(checkpoint_dir) ``` then, checkpoint-final will contains zero_to_fp32.py after the training is done....