beviah

This is not applicable to n-grams (subword info) from fasttext .bin files?

1

In other words, one cannot search for OOV words after mapping..

Ezglot has most common words in some languages you are missing: Albanian Belarusian Bosnian Icelandic Javanese Macedonian Norwegian Bokmål Norwegian Nynorsk Serbian Sinhala Tamil Telugu Welsh [most frequently used words](https://www.ezglot.com/most-frequently-used-words.php)...

Why not use xx hash?

7

https://github.com/Cyan4973/xxHash XXH128 seems to be superior in every way to MurmurHash3_x86_32 currently used.. https://github.com/RaRe-Technologies/bounter/blob/8c83e957b072a50be37221dcec1177fbe0128faf/cbounter/cms_common.c I am also a bit surprised by the perfect F1 scores given the high collision probability...

Number of files explodes, compaction does not work..

6

tried numerous settings.. something does not work right.. there are thousands of tiny log and sst files ... not getting merged.

### Describe the bug trace is quite long, but seems many originate here py::enum_(m, "MetricSignature") .value("ArrayArray", metric_punned_signature_t::array_array_k) .value("ArrayArraySize", metric_punned_signature_t::array_array_size_k); ==3173479== HEAP SUMMARY: ==3173479== in use at exit: 2,026,556 bytes in...

Bug: index.keys behaves strangely

1

### Describe the bug ``` >>> for k in index.keys:pass ... keeps running, so i interrupt ^C KeyboardInterrupt >>> print(k) 131959068520736 ``` index has only 100k entries. when i try...

A cross-model LLM analysis of detection of biases in their responses. [GENbAIs](https://genbais.com/)