mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2024-12-23 21:17:54 +01:00
llama : fix MPI threads (close #2827)
This commit is contained in:
parent
230d46c723
commit
c10704d01e
@ -2845,7 +2845,6 @@ static bool llama_eval_internal(
|
||||
|
||||
GGML_ASSERT(n_tokens > 0);
|
||||
GGML_ASSERT(n_past >= 0);
|
||||
GGML_ASSERT(n_threads > 0);
|
||||
// TODO: keep the values of n_batch and n_ctx
|
||||
// GGML_ASSERT(n_tokens <= n_batch);
|
||||
// GGML_ASSERT(n_past + n_tokens <= n_ctx);
|
||||
@ -2856,6 +2855,8 @@ static bool llama_eval_internal(
|
||||
ggml_mpi_eval_init(lctx.ctx_mpi, &n_tokens, &n_past, &n_threads);
|
||||
#endif
|
||||
|
||||
GGML_ASSERT(n_threads > 0);
|
||||
|
||||
const int N = n_tokens;
|
||||
|
||||
const auto & model = lctx.model;
|
||||
|
Loading…
Reference in New Issue
Block a user