Superminaren/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-27 21:46:57 +02:00
Code Issues Packages Projects Releases Wiki Activity

sampling : fix outputs and device checks

Browse Source
This commit is contained in:
Georgi Gerganov
2025-12-04 19:33:01 +02:00
parent abc19635a3
commit 7864074fdb
2 changed files with 6 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-sampling.cpp
+2 -8
View File
@@ -1018,9 +1018,8 @@ static bool llama_sampler_dist_backend_init(
ggml_tensor * op = ggml_cumsum(ctx, probs);
auto * device = ggml_backend_buft_get_device(buft);
GGML_ASSERT(device);
if (!ggml_backend_dev_supports_op(device, op)) {
if (device && !ggml_backend_dev_supports_op(device, op)) {
res = false;
}
@@ -1099,7 +1098,6 @@ static void llama_sampler_dist_backend_apply(
ggml_set_name(sampled_token, "dist_sampled_token");
}
ggml_set_output(sampled_token);
data->sampled = sampled_token;
}
@@ -1192,9 +1190,8 @@ static bool llama_sampler_top_k_backend_init(
ggml_tensor * op = ggml_top_k(ctx, logits, sctx->k);
auto * device = ggml_backend_buft_get_device(buft);
GGML_ASSERT(device);
if (!ggml_backend_dev_supports_op(device, op)) {
if (device && !ggml_backend_dev_supports_op(device, op)) {
res = false;
}
@@ -1411,9 +1408,6 @@ static void llama_sampler_top_p_backend_apply(
data->logits = ggml_add(ctx, sorted_logits, top_p_bias);
ggml_set_name(data->logits, "top_p_logits");
ggml_set_output(data->candidates);
ggml_set_output(data->logits);
GGML_UNUSED(gf);
}