msra-jqxu

Hi, @Espere-1119-Song, now the openai key is prohibited. So I replace the 'OpenAIEmbeddings()' with Ollama. Could you tell which embedding models you used in your evaluation code?

Hi @williamrs-openai , could you explain how the variables in [L209](https://github.com/openai/automated-interpretability/blob/e4aa21bc8baf26ff141ccf2e6f4444dcf120193f/neuron-explainer/neuron_explainer/explanations/simulator.py#L209) to [L212](https://github.com/openai/automated-interpretability/blob/e4aa21bc8baf26ff141ccf2e6f4444dcf120193f/neuron-explainer/neuron_explainer/explanations/simulator.py#L212) correspond to the logprobs results given in gpt4-1106? Especially choice["logprobs"]["text_offset"] made me confused. Here I set logprobs=True...

Hi @JacksonWuxs , do you know how to calculate activation? This formula 'h{l}.mlp.c_proj.w[:, n, :] @ diag(ln_f.g) @ wte[t, :]' seems to be independent of the sample and seems to...

> > Hi @JacksonWuxs , do you know how to calculate activation? This formula 'h{l}.mlp.c_proj.w[:, n, :] @ diag(ln_f.g) @ wte[t, :]' seems to be independent of the sample and...