diff options
| author | Jared Van Bortel <jared@nomic.ai> | 2024-02-21 10:33:54 -0500 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-02-21 10:33:54 -0500 |
| commit | 89febfed9322c8849520dc63c93ee4f5fd72556e (patch) | |
| tree | 60a680fab03dfd5def2827b7d4290e5db44351fa /examples/server/server.cpp | |
| parent | 5022cf242d689e15defd133f96c4345ad30c5d19 (diff) | |
examples : do not assume BOS when shifting context (#5622)
Diffstat (limited to 'examples/server/server.cpp')
| -rw-r--r-- | examples/server/server.cpp | 13 |
1 files changed, 7 insertions, 6 deletions
diff --git a/examples/server/server.cpp b/examples/server/server.cpp index 1c447951..c84719a0 100644 --- a/examples/server/server.cpp +++ b/examples/server/server.cpp @@ -1487,14 +1487,15 @@ struct llama_server_context if (slot.is_processing() && system_tokens.size() + slot.cache_tokens.size() >= (size_t) slot.n_ctx) { // Shift context - const int n_left = system_tokens.size() + slot.n_past - slot.params.n_keep - 1; + const int n_keep = slot.params.n_keep + add_bos_token; + const int n_left = system_tokens.size() + slot.n_past - n_keep; const int n_discard = n_left / 2; - LOG_TEE("slot %d: context shift - n_keep = %d, n_left = %d, n_discard = %d\n", slot.id, slot.params.n_keep, n_left, n_discard); - llama_kv_cache_seq_rm (ctx, slot.id, slot.params.n_keep + 1 , slot.params.n_keep + n_discard + 1); - llama_kv_cache_seq_shift(ctx, slot.id, slot.params.n_keep + 1 + n_discard, system_tokens.size() + slot.n_past, -n_discard); + LOG_TEE("slot %d: context shift - n_keep = %d, n_left = %d, n_discard = %d\n", slot.id, n_keep, n_left, n_discard); + llama_kv_cache_seq_rm (ctx, slot.id, n_keep , n_keep + n_discard); + llama_kv_cache_seq_shift(ctx, slot.id, n_keep + n_discard, system_tokens.size() + slot.n_past, -n_discard); - for (size_t i = slot.params.n_keep + 1 + n_discard; i < slot.cache_tokens.size(); i++) + for (size_t i = n_keep + n_discard; i < slot.cache_tokens.size(); i++) { slot.cache_tokens[i - n_discard] = slot.cache_tokens[i]; } @@ -1507,7 +1508,7 @@ struct llama_server_context LOG_VERBOSE("context shift", { { "n_ctx", n_ctx }, - { "n_keep", params.n_keep }, + { "n_keep", n_keep }, { "n_left", n_left }, }); } |