happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

clean up mtp sample typing after rebase

This commit is contained in:
Aaron Lee 2025-12-21 17:53:27 -05:00
parent a8dc54672c
commit d10a5a4a5b
5 changed files with 8 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
common/sampling.cpp
View File

@ -691,4 +691,4 @@ llama_token common_sampler_sample_speculative(struct common_sampler * gsmpl, str
} }
return best_id; return best_id;
} }

2
common/sampling.h
View File

@ -115,3 +115,5 @@ struct common_sampler_deleter {
}; };
typedef std::unique_ptr<common_sampler, common_sampler_deleter> common_sampler_ptr; typedef std::unique_ptr<common_sampler, common_sampler_deleter> common_sampler_ptr;
llama_token common_sampler_sample_speculative(struct common_sampler * gsmpl, struct llama_context * ctx, int idx);

4
common/speculative.cpp
View File

@ -361,8 +361,8 @@ llama_tokens common_speculative_gen_draft(
} }
llama_tokens mtp_speculative_gen_draft( llama_tokens mtp_speculative_gen_draft(
struct common_sampler* smpl, struct common_sampler * smpl,
struct llama_context* ctx, struct llama_context * ctx,
struct common_speculative_params params, struct common_speculative_params params,
llama_token id_last, llama_token id_last,
int32_t n_past, int32_t n_past,

4
common/speculative.h
View File

@ -57,8 +57,8 @@ llama_tokens common_speculative_gen_draft(
* @return std::vector<llama_token> The generated draft tokens. * @return std::vector<llama_token> The generated draft tokens.
*/ */
llama_tokens mtp_speculative_gen_draft( llama_tokens mtp_speculative_gen_draft(
struct common_sampler* smpl, struct common_sampler * smpl,
struct llama_context* ctx, struct llama_context * ctx,
struct common_speculative_params params, struct common_speculative_params params,
llama_token id_last, llama_token id_last,
int32_t n_past, int32_t n_past,

2
tools/server/server-context.cpp
View File

@ -2001,7 +2001,7 @@ struct server_context_impl {
llama_set_draft_input_hidden_state(ctx, llama_get_embeddings_ith(ctx, -1)); llama_set_draft_input_hidden_state(ctx, llama_get_embeddings_ith(ctx, -1));
draft = mtp_speculative_gen_draft( draft = mtp_speculative_gen_draft(
slot.smpl, slot.smpl.get(),
ctx, ctx,
params_spec, params_spec,
slot.sampled, slot.sampled,