Markus Tretzmüller

Results 7 comments of Markus Tretzmüller

Hi @BurakaKrishna , I'm curious to read your paper! I also had troubles with performance, it would be great to have a faster sampling procedure.

best issue ever!

Try `from biterm.btm import oBTM` instead of `from biterm.cbtm import oBTM`. If you want to use the cythonic version (cbtm) you have to first compile it yourself.

🆙 Adding a `ParentDocumentRetriever.delete` would be useful for compatibility with langchain indexing also.

change `Z = np.zeros(B)` to `Z = np.zeros(B, dtype=np.int8)`

> Testing with larger, real models would be appreciated. Tried [Meta-Llama-3-70B-Instruct](https://huggingface.co/cortecs/Meta-Llama-3-70B-Instruct-GPTQ/tree/8b), worked like a charm.

The update policy is: w(τ+1) = w(τ) − η∇E(w(τ)) where w(τ+1) is the new weight and w(τ) the old one. η is the learning rate and E the cost function....