apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

init (#6642)

Browse Source 
Co-authored-by: gongweibao <gognweibao@baidu.com>
This commit is contained in:
gongweibao
2026-03-04 21:55:31 +08:00
committed by GitHub
parent 5c8f5184d9
commit ddb06ff83f
306 changed files with 40627 additions and 34418 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/gpu_ops/env.h
+59 -35
View File
@@ -15,59 +15,83 @@
#pragma once
inline uint32_t get_decoder_block_shape_q() {
static const char* decoder_block_shape_q_env = std::getenv("FLAGS_dec_block_shape_q");
static const uint32_t decoder_block_shape_q =
decoder_block_shape_q_env == nullptr ? 16 : std::stoi(std::string(decoder_block_shape_q_env));
return decoder_block_shape_q;
static const char* decoder_block_shape_q_env =
std::getenv("FLAGS_dec_block_shape_q");
static const uint32_t decoder_block_shape_q =
decoder_block_shape_q_env == nullptr
? 16
: std::stoi(std::string(decoder_block_shape_q_env));
return decoder_block_shape_q;
}
inline uint32_t get_encoder_block_shape_q() {
static const char* encoder_block_shape_q_env = std::getenv("FLAGS_enc_block_shape_q");
static const uint32_t encoder_block_shape_q =
encoder_block_shape_q_env == nullptr ? 64 : std::stoi(std::string(encoder_block_shape_q_env));
return encoder_block_shape_q;
static const char* encoder_block_shape_q_env =
std::getenv("FLAGS_enc_block_shape_q");
static const uint32_t encoder_block_shape_q =
encoder_block_shape_q_env == nullptr
? 64
: std::stoi(std::string(encoder_block_shape_q_env));
return encoder_block_shape_q;
}
inline uint32_t get_max_partition_size(int bsz) {
static const char* max_partition_size_env = std::getenv("FLAGS_cascade_attention_max_partition_size");
static const uint32_t max_partition_size =
max_partition_size_env == nullptr ? 32768 : std::stoul(std::string(max_partition_size_env));
return max_partition_size;
static const char* max_partition_size_env =
std::getenv("FLAGS_cascade_attention_max_partition_size");
static const uint32_t max_partition_size =
max_partition_size_env == nullptr
? 32768
: std::stoul(std::string(max_partition_size_env));
return max_partition_size;
}
inline uint32_t get_cascade_attention_deal_each_time() {
static const char* cascade_attention_deal_each_time_env = std::getenv("FLAGS_cascade_attention_deal_each_time");
static const uint32_t cascade_attention_deal_each_time =
cascade_attention_deal_each_time_env == nullptr ? 0 : std::stoul(std::string(cascade_attention_deal_each_time_env));
return (cascade_attention_deal_each_time != 0 ? cascade_attention_deal_each_time : 32);
static const char* cascade_attention_deal_each_time_env =
std::getenv("FLAGS_cascade_attention_deal_each_time");
static const uint32_t cascade_attention_deal_each_time =
cascade_attention_deal_each_time_env == nullptr
? 0
: std::stoul(std::string(cascade_attention_deal_each_time_env));
return (cascade_attention_deal_each_time != 0
? cascade_attention_deal_each_time
: 32);
}
inline uint32_t get_cascade_attention_num_stages() {
static const char* cascade_attention_num_stages_env = std::getenv("FLAGS_cascade_attention_num_stages");
static const uint32_t cascade_attention_num_stages =
cascade_attention_num_stages_env == nullptr ? 0 : std::stoul(std::string(cascade_attention_num_stages_env));
return cascade_attention_num_stages != 0 ? cascade_attention_num_stages : 2;
static const char* cascade_attention_num_stages_env =
std::getenv("FLAGS_cascade_attention_num_stages");
static const uint32_t cascade_attention_num_stages =
cascade_attention_num_stages_env == nullptr
? 0
: std::stoul(std::string(cascade_attention_num_stages_env));
return cascade_attention_num_stages != 0 ? cascade_attention_num_stages : 2;
}
inline uint32_t get_cascade_attention_num_threads() {
static const char* cascade_attention_num_threads_env = std::getenv("FLAGS_cascade_attention_num_threads");
static const uint32_t cascade_attention_num_threads =
cascade_attention_num_threads_env == nullptr ? 0 : std::stoul(std::string(cascade_attention_num_threads_env));
return cascade_attention_num_threads != 0 ? cascade_attention_num_threads : 128;
static const char* cascade_attention_num_threads_env =
std::getenv("FLAGS_cascade_attention_num_threads");
static const uint32_t cascade_attention_num_threads =
cascade_attention_num_threads_env == nullptr
? 0
: std::stoul(std::string(cascade_attention_num_threads_env));
return cascade_attention_num_threads != 0 ? cascade_attention_num_threads
: 128;
}
inline bool get_mla_use_tensorcore() {
static const char* mla_use_tensorcore_env = std::getenv("FLAGS_mla_use_tensorcore");
static const uint32_t mla_use_tensorcore =
mla_use_tensorcore_env == nullptr ? 0 : std::stoul(std::string(mla_use_tensorcore_env));
return mla_use_tensorcore != 0 ? true : false;
static const char* mla_use_tensorcore_env =
std::getenv("FLAGS_mla_use_tensorcore");
static const uint32_t mla_use_tensorcore =
mla_use_tensorcore_env == nullptr
? 0
: std::stoul(std::string(mla_use_tensorcore_env));
return mla_use_tensorcore != 0 ? true : false;
}
inline int get_mla_dec_chunk_size(int bsz) {
static const char* mla_dec_chunk_size_env =
std::getenv("FLAGS_mla_dec_chunk_size");
static const int mla_dec_chunk_size =
mla_dec_chunk_size_env == nullptr
? -1
: std::stoi(std::string(mla_dec_chunk_size_env));
return bsz > 1 ? mla_dec_chunk_size : 64;
static const char* mla_dec_chunk_size_env =
std::getenv("FLAGS_mla_dec_chunk_size");
static const int mla_dec_chunk_size =
mla_dec_chunk_size_env == nullptr
? -1
: std::stoi(std::string(mla_dec_chunk_size_env));
return bsz > 1 ? mla_dec_chunk_size : 64;
}