Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-01-26 03:12:23 +01:00
Code Issues Packages Projects Releases Wiki Activity

llama : fix MPI threads (close #2827)

Browse Source
This commit is contained in:
Georgi Gerganov 2023-08-27 18:55:41 +03:00
parent 230d46c723
commit c10704d01e
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
llama.cpp
View File

@ -2845,7 +2845,6 @@ static bool llama_eval_internal(
GGML_ASSERT(n_tokens > 0); GGML_ASSERT(n_tokens > 0);
GGML_ASSERT(n_past >= 0); GGML_ASSERT(n_past >= 0);
GGML_ASSERT(n_threads > 0);
// TODO: keep the values of n_batch and n_ctx // TODO: keep the values of n_batch and n_ctx
// GGML_ASSERT(n_tokens <= n_batch); // GGML_ASSERT(n_tokens <= n_batch);
// GGML_ASSERT(n_past + n_tokens <= n_ctx); // 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); ggml_mpi_eval_init(lctx.ctx_mpi, &n_tokens, &n_past, &n_threads);
#endif #endif
GGML_ASSERT(n_threads > 0);
const int N = n_tokens; const int N = n_tokens;
const auto & model = lctx.model; const auto & model = lctx.model;