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

llama : propagate the results of graph_compute (#9525)

Browse source 
* llama: propagating the results of `graph_compute` to the user interface

* llama: reverting kv_cache in case of failed compute

* llama: `llama_kv_cache_state` was removed, only the result of `llama_graph_compute` is returned

* llama: restore a kv_cache in case of failed computation

* llama: correct reverting of the entire batch.
also updates `llama_kv_cache_find_slot`, will correctly count the number of `used` cells for recurrent models

* llama: updated comments

* llama : add comments about KV cache state after error

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Michael Podvitskiy 2024-11-13 20:00:35 +02:00 • committed by GitHub
parent 5ea926dad7
commit fb4a0ec083
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 109 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

4
include/llama.h
View file

@ -797,7 +797,7 @@ extern "C" {
// Processes a batch of tokens with the ecoder part of the encoder-decoder model.
// Stores the encoder output internally for later use by the decoder cross-attention layers.
// 0 - success
// < 0 - error
// < 0 - error. the KV cache state is restored to the state before this call
LLAMA_API int32_t llama_encode(
struct llama_context * ctx,
struct llama_batch batch);
@ -805,7 +805,7 @@ extern "C" {
// Positive return values does not mean a fatal error, but rather a warning.
// 0 - success
// 1 - could not find a KV slot for the batch (try reducing the size of the batch or increase the context)
// < 0 - error
// < 0 - error. the KV cache state is restored to the state before this call
LLAMA_API int32_t llama_decode(
struct llama_context * ctx,
struct llama_batch batch);