vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix perplexity after c-api refactor by proving a large enough token buffer

Browse source
This commit is contained in:
Green Sky 2023-03-22 12:44:26 +01:00
parent d5850c53ca
commit 827bcb1375
No known key found for this signature in database
1 changed files with 6 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

7
main.cpp
View file

@ -85,7 +85,12 @@ void perplexity(llama_context * ctx, const gpt_params & params) {
// Download: https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip?ref=salesforce-research // Download: https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip?ref=salesforce-research
// Run `./main --perplexity -m models/7B/ggml-model-q4_0.bin -f wiki.test.raw` // Run `./main --perplexity -m models/7B/ggml-model-q4_0.bin -f wiki.test.raw`
// Output: `perplexity: 13.5106 [114/114]` // Output: `perplexity: 13.5106 [114/114]`
auto tokens = ::llama_tokenize(ctx, params.prompt.c_str(), true); std::vector<llama_token> tokens(params.prompt.size()+1); // initialize to prompt numer of chars, since n_tokens <= n_prompt_chars+1
{
const auto res = llama_tokenize(ctx, params.prompt.c_str(), tokens.data(), tokens.size(), true);
assert(res >= 0);
tokens.resize(res);
}
int count = 0; int count = 0;
double nll = 0.0; double nll = 0.0;