Update _llama_cpp.py based on new contract of llama_get_logits

[czoccolo]

The new documentation for llama_get_logits is:

// Token logits obtained from the last call to llama_decode()
// The logits for which llama_batch.logits[i] != 0 are stored contiguously
// in the order they have appeared in the batch.
// Rows: number of tokens for which llama_batch.logits[i] != 0
// Cols: n_vocab
LLAMA_API float * llama_get_logits(struct llama_context * ctx);

Since we set llama_batch.logits[i] to true only for the last token, llama_get_logits already returns only the desired slice.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 62.50%. Comparing base (8cf29c2) to head (8de01b4).

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #769      +/-   ##
==========================================
- Coverage   68.97%   62.50%   -6.48%     
==========================================
  Files          55       55              
  Lines        4058     4059       +1     
==========================================
- Hits         2799     2537     -262     
- Misses       1259     1522     +263     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[paulbkoch]

Thanks @czoccolo! This was a great find, and would have taken me a long time to figure out myself, so I'm really glad you submitted this as a PR.