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

imatrix: add --no-ppl option to skip PPL calculations altogether

Browse source
This commit is contained in:
Iwan Kawrakow 2024-01-20 19:04:09 +02:00
parent cdeac23ef5
commit 3aa56562c0
1 changed files with 39 additions and 25 deletions
Download patch file Download diff file Expand all files Collapse all files

26
examples/imatrix/imatrix.cpp
View file

@ -248,7 +248,7 @@ static void process_logits(
}
}
static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
static bool compute_imatrix(llama_context * ctx, const gpt_params & params, bool compute_ppl) {
const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
const int n_ctx = llama_n_ctx(ctx);
@ -269,10 +269,12 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
}
std::vector<float> logit_history;
logit_history.resize(tokens.size());
std::vector<float> prob_history;
if (compute_ppl) {
logit_history.resize(tokens.size());
prob_history.resize(tokens.size());
}
const int n_chunk_max = tokens.size() / n_ctx;
@ -291,7 +293,7 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
const int num_batches = (n_ctx + n_batch - 1) / n_batch;
std::vector<float> logits;
if (num_batches > 1) {
if (compute_ppl && num_batches > 1) {
logits.reserve((size_t)n_ctx * n_vocab);
}
@ -326,7 +328,7 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
// restore the original token in case it was set to BOS
tokens[batch_start] = token_org;
if (num_batches > 1) {
if (compute_ppl && num_batches > 1) {
const auto * batch_logits = llama_get_logits(ctx);
logits.insert(logits.end(), batch_logits, batch_logits + batch_size * n_vocab);
}
@ -345,6 +347,7 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
fprintf(stderr, "%.2f minutes\n", total_seconds / 60.0);
}
if (compute_ppl) {
const int first = n_ctx/2;
const auto all_logits = num_batches > 1 ? logits.data() : llama_get_logits(ctx);
process_logits(n_vocab, all_logits + first*n_vocab, tokens.data() + start + first, n_ctx - 1 - first,
@ -356,8 +359,10 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
logits.clear();
}
}
printf("\n");
if (compute_ppl) {
nll2 /= count;
nll /= count;
const double ppl = exp(nll);
@ -368,6 +373,7 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
} else {
printf("Unexpected negative standard deviation of log(prob)\n");
}
}
return true;
}
@ -375,6 +381,7 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
int main(int argc, char ** argv) {
StatParams sparams;
bool compute_ppl = true;
std::vector<char*> args;
args.push_back(argv[0]);
int iarg = 1;
@ -391,13 +398,20 @@ int main(int argc, char ** argv) {
}
else if (arg == "--verbosity") {
sparams.verbosity = std::stoi(argv[++iarg]);
} else if (arg == "--no-ppl") {
compute_ppl = false;
} else {
args.push_back(argv[iarg]);
}
}
if (iarg < argc) {
std::string arg{argv[iarg]};
if (arg == "--no-ppl") {
compute_ppl = false;
} else {
args.push_back(argv[iarg]);
}
}
gpt_params params;
params.n_batch = 512;
@ -458,7 +472,7 @@ int main(int argc, char ** argv) {
fprintf(stderr, "%s\n", get_system_info(params).c_str());
}
bool OK = compute_imatrix(ctx, params);
bool OK = compute_imatrix(ctx, params, compute_ppl);
if (!OK) {
return 1;
}