captum
captum copied to clipboard
pytorch
Model interpretability and understanding for PyTorch
## ❓ Questions and Help We have a set of listed resources available on the website and FAQ: https://captum.ai/ and https://captum.ai/docs/faq . Feel free to open an issue here on...
Hi I am exploring using Captum to get some interpretability for a hybrid model I am building. My basic model calculates an embedding, which is combined with other inputs and...
## 📚 Documentation In https://captum.ai/tutorials/Llama2_LLM_Attribution it seems like Llama2-chat is used but it seems like the chat template is never applied. Do I see this correctly or is it applied...
## ❓ Questions and Help Hi there @sarahtranfb 👋, I would like to use IG for Image and Text to display the results of VLM models such as [LLava-OneVision](https://huggingface.co/llava-hf/llava-onevision-qwen2-0.5b-ov-hf) or...
## ❓ Questions and Help I've been successfull using captum's `LayerIntegratedGradients` class, but none of my attempts trying the same sorts of inputs and targets using `LLMAttribution` seem to work....
## 🐛 Bug ## To Reproduce Steps to reproduce the behavior: 1. Use `inp = TextTemplateInput(template=attr_template, values=input_sentences)` to create an input 1. Use `attr_res = llm_attr.attribute(inp, skip_tokens=[tokenizer.bos_token_id])` to do the...
We've flipped the default for the weights_only argument in torch.load to True in pytorch/pytorch, see [here](https://dev-discuss.pytorch.org/t/bc-breaking-change-torch-load-is-being-flipped-to-use-weights-only-true-by-default-in-the-nightlies-after-137602/2573) for details + [documentation](https://pytorch.org/docs/main/notes/serialization.html#torch-load-with-weights-only-true) and this is coming in torch 2.6. This is expected...
## ❓ Questions and Help I have an 8G NVIDIA GeForce RTX 3050 and it is able to run the [LLama2 demo](https://github.com/pytorch/captum/blob/master/tutorials/Llama2_LLM_Attribution.ipynb) but only until it tries to build the...
I was following this [demonstration](https://captum.ai/tutorials/Bert_SQUAD_Interpret) to interpret my DistilBert-based model. I found that, in a few cases, IG does not converge even with a high `n_steps` value. However, if we...
