CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-12 10:47:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into gg/llama-kv-cache

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2025-02-06 10:04:33 +02:00
parent e0d913fccb 2c6c8df56d
commit 0f1c1cab2c
83 changed files with 3593 additions and 1410 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama-grammar.cpp
View File

@@ -1213,5 +1213,7 @@ void llama_grammar_accept_str(struct llama_grammar & grammar, const std::string
}
grammar.partial_utf8 = decoded.second;
GGML_ASSERT(!grammar.stacks.empty());
if (grammar.stacks.empty()) {
throw std::runtime_error("Unexpected empty grammar stack after accepting piece: " + piece);
}
}

6
src/llama.cpp
View File

@@ -3868,7 +3868,8 @@ struct llm_build_context {
ggml_row_size(kv_pe_compresseed->type, kv_lora_rank));
cb(k_pe, "k_pe", il);
kv_compressed = ggml_cont(ctx0, kv_compressed); // TODO: the CUDA backend does not support non-contiguous norm
// TODO: the CUDA backend used to not support non-cont. (RMS) norm, investigate removing ggml_cont
kv_compressed = ggml_cont(ctx0, kv_compressed);
kv_compressed = build_norm(kv_compressed,
model.layers[il].attn_kv_a_norm, NULL,
LLM_NORM_RMS, il);
@@ -5702,7 +5703,8 @@ struct llm_build_context {
ggml_row_size(kv_pe_compresseed->type, kv_lora_rank));
cb(k_pe, "k_pe", il);
kv_compressed = ggml_cont(ctx0, kv_compressed); // TODO: the CUDA backend does not support non-contiguous norm
// TODO: the CUDA backend used to not support non-cont. (RMS) norm, investigate removing ggml_cont
kv_compressed = ggml_cont(ctx0, kv_compressed);
kv_compressed = build_norm(kv_compressed,
model.layers[il].attn_kv_a_norm, NULL,
LLM_NORM_RMS, il);