saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-29 20:14:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

draft

Browse source
This commit is contained in:
Tianlei Wu 2025-01-24 16:30:39 -08:00
parent 3c3103e5df
commit 9a830b77e3
2 changed files with 35 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
onnxruntime/contrib_ops/cuda/transformers/generation_cuda_impl.cu
View file

@ -434,6 +434,10 @@ void LaunchBeamSearchScorer_Process(BeamScorerState& state_cpu,
gsl::span<const int32_t> next_tokens,
gsl::span<const int32_t> next_indices,
cudaStream_t stream) {
for (size_t i = 0; i < beam_hyps.length(); i++) {
dump_beam_hypotheses(&beam_hyps[i]);
}
BeamSearchScorer_Process<<<1, state_cpu.batch_size_, 0, stream>>>(state_cpu,
state,
sequences.data(),

31
onnxruntime/contrib_ops/cuda/transformers/generation_cuda_impl.h
View file

@ -52,6 +52,37 @@ struct HypothesisScore {
float score;
};
// Function to dump the data in the struct
__device__ void dump_hypothesis_score(const struct HypothesisScore* hs) {
printf("HypothesisScore Dump:\n");
printf(" hypothesis_length: %d\n", hs->hypothesis_length);
printf(" score: %f\n", hs->score);
printf(" hypothesis: ");
if (hs->hypothesis_length > 0 && hs->hypothesis != NULL) {
for (int i = 0; i < hs->hypothesis_length; ++i) {
printf("%d ", hs->hypothesis[i]);
}
} else {
printf("(empty)");
}
printf("\n");
}
__device__ void dump_beam_hypotheses(const struct BeamHypotheses* bh) {
printf("BeamHypotheses Dump:\n");
printf(" beams_count_: %d\n", bh->beams_count_);
printf(" beams_used_: %d\n", bh->beams_used_);
printf(" length_penalty_: %f\n", bh->length_penalty_);
printf(" done_: %s\n", bh->done_ ? "true" : "false");
printf(" beams_:\n");
for (int i = 0; i < bh->beams_used_; ++i) {
printf(" Beam %d:\n", i + 1);
dump_hypothesis_score(&bh->beams_[i]);
}
}
struct BeamHypotheses {
HypothesisScore* beams_; // Beam width sized array of hypotheses, sorted by highest scoring
int beams_count_;