mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-10-05 16:48:03 +08:00
fix typos (#3684)
This commit is contained in:
@@ -63,7 +63,7 @@ __global__ void ComputeOrderKernel(
|
||||
position_map[in_offset++] = out_offset++;
|
||||
}
|
||||
in_offset += cur_base_model_seq_lens_this_time - accept_num;
|
||||
// (liuzichang): Temperary Reserved for debug
|
||||
// (liuzichang): Temporary Reserved for debug
|
||||
// if (accept_num <= actual_draft_token_num) /*Accept partial draft tokens*/ {
|
||||
// #ifdef DEBUG_EAGLE_KERNEL
|
||||
// printf("batch %d: accept_num <= actual_draft_token_num \n", i);
|
||||
|
@@ -35,7 +35,7 @@ __global__ void speculate_set_value_by_flag_and_id(int64_t *pre_ids_all,
|
||||
accept_tokens + tid * max_draft_tokens;
|
||||
const int seq_len_dec = seq_lens_decoder[tid];
|
||||
const int seq_len_enc = seq_lens_encoder[tid];
|
||||
if (seq_len_dec == 0 && seq_len_enc == 0) return; // stoped
|
||||
if (seq_len_dec == 0 && seq_len_enc == 0) return; // stopped
|
||||
// printf("step_idx[tid] %d\n", step_idx[tid]);
|
||||
if (step_idx[tid] >= 0) {
|
||||
for (int i = 0; i < accept_num[tid]; i++) {
|
||||
|
@@ -295,7 +295,7 @@ void SpeculateStepSchedule(const paddle::Tensor &stop_flags,
|
||||
std::string inference_msg_id_env_str(inference_msg_id_env_p);
|
||||
inference_msg_id_from_env = std::stoi(inference_msg_id_env_str);
|
||||
if (inference_msg_id_from_env == 2) {
|
||||
// 2 and -2 is perserve for no-output indication.
|
||||
// 2 and -2 is preserve for no-output indication.
|
||||
throw std::runtime_error(
|
||||
" INFERENCE_MSG_ID cannot be 2, please use other number.");
|
||||
}
|
||||
|
Reference in New Issue
Block a user