ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-06-17 06:18:08 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

sync : ggml (ggml_scale, ggml_row_size, etc.) (#1677)

Browse Source 
* sync : ggml

* sync : llama.cpp

* talk-llama : fix obsolete param

* ggml-alloc : fix ggml_tallocr_is_own

* talk.wasm : update to new ggml

* ggml : fix type punning in ggml_scale

* ggml : cuda jetson + arm quants warnings
This commit is contained in:
Georgi Gerganov
2023-12-22 17:53:39 +02:00
committed by GitHub
parent d2ee117a0a
commit 3a5302108d
18 changed files with 3479 additions and 1537 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
ggml-alloc.c
View File

@ -72,7 +72,7 @@ static void remove_allocated_tensor(ggml_tallocr_t alloc, struct ggml_tensor * t
// check if a tensor is allocated by this buffer
static bool ggml_tallocr_is_own(ggml_tallocr_t alloc, const struct ggml_tensor * tensor) {
return tensor->buffer == alloc->buffer;
return tensor->buffer == alloc->buffer && (!tensor->view_src || tensor->view_src->buffer == alloc->buffer);
}
static bool ggml_is_view(struct ggml_tensor * t) {
@ -449,11 +449,10 @@ static void init_view(ggml_gallocr_t galloc, struct ggml_tensor * view, bool upd
if (update_backend) {
view->backend = view->view_src->backend;
}
view->buffer = view->view_src->buffer;
// views are initialized in the alloc buffer rather than the view_src buffer
view->buffer = alloc->buffer;
view->data = (char *)view->view_src->data + view->view_offs;
// FIXME: the view should be initialized by the owning buffer, but currently this breaks the CUDA backend
// due to the ggml_tensor_extra_gpu ring buffer overwriting the KV cache extras
assert(ggml_tallocr_is_measure(alloc) || !view->buffer || view->buffer->buft == alloc->buffer->buft);
if (!alloc->measure) {
@ -736,6 +735,10 @@ void ggml_allocr_set_parse_seq(ggml_allocr_t alloc, const int * list, int n) {
}
void ggml_allocr_free(ggml_allocr_t alloc) {
if (alloc == NULL) {
return;
}
ggml_gallocr_free(alloc->galloc);
ggml_tallocr_free(alloc->talloc);
free(alloc);
@ -775,7 +778,7 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte
}
if (nbytes == 0) {
fprintf(stderr, "%s: no tensors to allocate\n", __func__);
// all the tensors in the context are already allocated
return NULL;
}
@ -789,6 +792,11 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte
} else {
ggml_backend_view_init(buffer, t);
}
} else {
if (t->view_src != NULL) {
// view of a pre-allocated tensor
ggml_backend_view_init(buffer, t);
}
}
}