Ritvik Kapila

icon indicating a website link https://sites.google.com/view/ritvik-kapila/home?authuser=0

icon company University of California San Diego icon location San Diego, United States icon location Strong research professional skilled in Machine Learning, Big Data Analysis and Software Development with experience in both academia and industry.

Results 5 comments of Ritvik Kapila

IMU Input Dimensions are Unclear - Missing Information on Data Prep

Hi, I had a question similar to that of @beitong95, how is the IMU input preprocessed and/or normalized before being fed as an input to the model? Is there a...

final answer extraction

[This](https://github.com/EleutherAI/lm-evaluation-harness/blob/c0fc717240032aec738c3199ae344887b5f34c23/lm_eval/tasks/aime/utils.py#L5) is the function used to extract the answer. It checks for the answer in both the `$$` and `\boxed{}` formats. You can also use few_shot prompting to set the...

GRPO Qwen3 megatron training script

I don't know if you've already seen them but these examples might be helpful - https://github.com/volcengine/verl/tree/main/examples/grpo_trainer You'll find some Qwen3 examples at the end.

unable to run evals on anthropic-completions

Quick update, the [anthropic-sdk-python](https://github.com/anthropics/anthropic-sdk-python/blob/v0.39.0/src/anthropic/_client.py) had the `get_tokenizer` command removed from `v0.38.0` to `v0.39.0`. However, pinning `anthropic` pypi project to `v0.38.0` also throws an error - `TypeError: __init__() got an unexpected...

[Bug] Unable to create Qwen3 MoE model

@casper-hansen this is super relevant for me. Can you point me to some reference where I can read more about this? I want to post-train a Qwen MOE model as...