ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-02-25 07:34:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix CUDA

Browse Source 
There is an issue with quantized GEMV on CUDA when the left operand
(the matrix) is not contiguous. So, for now, we also create wv_b
during model loading and use that instead of the 3D view of wkv_b.
This commit is contained in:
Iwan Kawrakow
2025-03-16 07:40:18 +02:00
parent fc03b9adbc
commit f2fb15de77
1 changed files with 41 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
src/llama.cpp
View File

@@ -2642,6 +2642,7 @@ struct llama_layer {
struct ggml_tensor * ffn_down_scale;
std::unique_ptr<ggml_tensor> computed_wk_b;
std::unique_ptr<ggml_tensor> computed_wv_b;
};
struct llama_kv_cell {
@@ -8207,7 +8208,6 @@ static bool llm_load_tensors(
if (status != GGML_STATUS_SUCCESS) throw std::runtime_error("Failed to compute wk_b");
auto name = std::string{"blk."} + std::to_string(il) + ".attn_k_b.weight";
printf("Computed %s as %ld x %ld x %ld\n", name.c_str(), wk_b->ne[0], wk_b->ne[1], wk_b->ne[2]);
l.computed_wk_b = std::make_unique<ggml_tensor>(*wk_b);
l.computed_wk_b->buffer = ggml_backend_buft_alloc_buffer(ggml_backend_buffer_get_type(l.wkv_b->buffer), ggml_nbytes(wk_b));
@@ -8221,6 +8221,34 @@ static bool llm_load_tensors(
l.wk_b = l.computed_wk_b.get();
ggml_graph_clear(graph);
auto wv_b = ggml_cont(ctx, ggml_view_3d(ctx, &wkv_b, kv_lora_rank, n_embd_head_v, n_head,
l.wkv_b->nb[1], l.wkv_b->nb[1]*(n_embd_head_qk_nope + n_embd_head_v), l.wkv_b->nb[1]*n_embd_head_qk_nope));
wv_b->data = tensor_data.data();
ggml_build_forward_expand(graph, wv_b);
plan = ggml_graph_plan(graph, std::thread::hardware_concurrency()/2);
if (plan.work_size > work_data.size()) work_data.resize(plan.work_size);
plan.work_data = work_data.data();
status = ggml_graph_compute(graph, &plan);
if (status != GGML_STATUS_SUCCESS) throw std::runtime_error("Failed to compute wv_b");
name = std::string{"blk."} + std::to_string(il) + ".attn_v_b.weight";
l.computed_wv_b = std::make_unique<ggml_tensor>(*wv_b);
l.computed_wv_b->buffer = ggml_backend_buft_alloc_buffer(ggml_backend_buffer_get_type(l.wkv_b->buffer), ggml_nbytes(wv_b));
l.computed_wv_b->data = ggml_backend_buffer_get_base(l.computed_wv_b->buffer);
l.computed_wv_b->op = GGML_OP_NONE; // we absolutely need to do this, else the backend will attempt to find the parents
// of wk_b, which no longer exist, and will therefore crash.
for (int j = 0; j < GGML_MAX_SRC; ++j) l.computed_wv_b->src[j] = nullptr;
ggml_set_name(l.computed_wv_b.get(), name.c_str());
ggml_backend_buffer_set_usage(l.computed_wv_b->buffer, GGML_BACKEND_BUFFER_USAGE_WEIGHTS);
ggml_backend_tensor_set(l.computed_wv_b.get(), wv_b->data, 0, ggml_nbytes(wv_b));
l.wv_b = l.computed_wv_b.get();
printf("Computed %s as %ld x %ld x %ld and stored in buffer %s\n", name.c_str(), wk_b->ne[0], wk_b->ne[1], wk_b->ne[2],
ggml_backend_buffer_name(l.computed_wk_b->buffer));
ggml_graph_clear(graph);
}
ggml_free(ctx);
@@ -13928,10 +13956,18 @@ struct llm_build_context {
}
}
auto wv_b = ggml_view_3d(ctx0, wkv_b, kv_lora_rank, n_embd_head_v, n_head,
wkv_b->nb[1], wkv_b->nb[1]*(n_embd_head_v + n_embd_head_qk_nope),
wkv_b->nb[1]*n_embd_head_qk_nope);
cb(wv_b, "wv_b", il);
auto wv_b = model.layers[il].wv_b;
if (wv_b->ne[1] != n_embd_head_v) {
wv_b = ggml_reshape_3d(ctx0, wv_b, kv_lora_rank, n_embd_head_v, n_head);
cb(wv_b, "wv_b", il);
}
// There is an issue with quantized GEMV on CUDA when the left operand (the matrix) is
// not contiguous. So, for now, we create wv_b during model loading and use that
// instead of the commented out 3D view below.
//auto wv_b = ggml_view_3d(ctx0, wkv_b, kv_lora_rank, n_embd_head_v, n_head,
// wkv_b->nb[1], wkv_b->nb[1]*(n_embd_head_v + n_embd_head_qk_nope),
// wkv_b->nb[1]*n_embd_head_qk_nope);
//cb(wv_b, "wv_b", il);
kqv = ggml_mul_mat(ctx0, wv_b, kqv_compressed);
cb(kqv, "kqv", il);