tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-14 04:17:53 -04:00
Code Issues Packages Projects Releases Wiki Activity

perplexity : update comments/error msg to use decode [no ci] (#15227)

Browse Source 
This commit updates comments and error messages to use "decode" instead
of "eval" in perplexity.cpp.

The motivation for this is that `llama_eval` was renamed to
`llama_decode` a while ago, but the comments and error messages
still referred to "eval". This change ensures consistency and clarity.
This commit is contained in:
Daniel Bevenius
2025-08-11 10:21:24 +02:00
committed by GitHub
parent a3a7874272
commit 1ebbaddff2
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tools/perplexity/perplexity.cpp
View File

@@ -525,7 +525,7 @@ static results_perplexity perplexity(llama_context * ctx, const common_params &
}
// We get the logits for all the tokens in the context window (params.n_ctx)
// from llama_eval above. Now, based on https://huggingface.co/docs/transformers/perplexity,
// from llama_decode below. Now, based on https://huggingface.co/docs/transformers/perplexity,
// calculate the perplexity over the last half of the window (so the model always has
// some context to predict the token).
//
@@ -559,7 +559,7 @@ static results_perplexity perplexity(llama_context * ctx, const common_params &
for (int seq = 0; seq < n_seq_batch; seq++) {
int seq_start = batch_start + seq*n_ctx;
// save original token and restore it after eval
// save original token and restore it after decode
const auto token_org = tokens[seq_start];
// add BOS token for the first batch of each chunk
@@ -584,7 +584,7 @@ static results_perplexity perplexity(llama_context * ctx, const common_params &
}
if (llama_decode(ctx, batch)) {
LOG_INF("%s : failed to eval\n", __func__);
LOG_INF("%s : failed to decode\n", __func__);
return {tokens, -1, logit_history, prob_history};
}