Skip to content

[Executorch][llama] Change runner to decouple prompt length from sequence length #9594

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Mar 26, 2025
Merged

[Executorch][llama] Change runner to decouple prompt length from sequence length #9594

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bcfb3a4
[Executorch][llama] Change runner to decouple prompt length from sequ…
kimishpatel Mar 25, 2025
f4f089e
Update on "[Executorch][llama] Change runner to decouple prompt lengt…
kimishpatel Mar 25, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions examples/models/llama/runner/runner.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ static constexpr auto kEnableDynamicShape = "enable_dynamic_shape";
static constexpr auto kBosId = "get_bos_id";
static constexpr auto kEosIds = "get_eos_ids";
static constexpr auto kMaxSeqLen = "get_max_seq_len";
static constexpr auto kMaxContextLen = "get_max_context_len";
static constexpr auto kVocabSize = "get_vocab_size";
static constexpr auto kUseKVCache = "use_kv_cache";
static constexpr auto kUseSDPAWithKVCache = "use_sdpa_with_kv_cache";
Expand All @@ -49,6 +50,7 @@ Runner::Runner(
metadata_({
{kEnableDynamicShape, false},
{kMaxSeqLen, 128},
{kMaxContextLen, 128},
{kUseKVCache, true},
{kUseSDPAWithKVCache, false},
}) {
Expand Down Expand Up @@ -201,9 +203,9 @@ Error Runner::generate(
shouldStop_ = false;

// Set the sequence length to the max seq length if not provided
seq_len = (seq_len > 0 && seq_len <= metadata_.at(kMaxSeqLen))
seq_len = (seq_len > 0 && seq_len <= metadata_.at(kMaxContextLen))
? seq_len
: metadata_.at(kMaxSeqLen);
: metadata_.at(kMaxContextLen);

::tokenizers::Result<std::vector<uint64_t>> encode_res = tokenizer_->encode(
prompt,
Expand Down
Loading