refine python API folder structure

seperate client and server: we may support multiple version of client, but server can be only one version

CMakeLists.txt

@@ -137,4 +137,4 @@ add_subdirectory(python)
 set(PYTHON_INCLUDE_DIR ${PYTHON_INCLUDE})
 set(PYTHON_LIBRARIES ${PYTHON_LIB})
 
-add_subdirectory(examples)
+#add_subdirectory(examples)

examples/demo-serving/conf/model_toolkit.prototxt

-engines {
-  name: "image_classification_resnet"
-  type: "FLUID_CPU_ANALYSIS_DIR"
-  reloadable_meta: "./data/model/paddle/fluid_time_file"
-  reloadable_type: "timestamp_ne"
-  model_data_path: "./data/model/paddle/fluid/SE_ResNeXt50_32x4d"
-  runtime_thread_num: 0
-  batch_infer_size: 0
-  enable_batch_align: 0  
-  enable_memory_optimization: true
-  static_optimization: false
-  force_update_static_cache: false
-}
-
 engines {
   name: "general_model"
   type: "FLUID_CPU_ANALYSIS_DIR"
@@ -22,39 +8,3 @@ engines {
   batch_infer_size: 0
   enable_batch_align: 0
 }
-
-engines {
-  name: "text_classification_bow"
-  type: "FLUID_CPU_ANALYSIS_DIR"
-  reloadable_meta: "./data/model/paddle/fluid_time_file"
-  reloadable_type: "timestamp_ne"
-  model_data_path: "./data/model/paddle/fluid/text_classification_lstm"
-  runtime_thread_num: 0
-  batch_infer_size: 0
-  enable_batch_align: 0
-}
-
-engines {
-  name: "ctr_prediction"
-  type: "FLUID_CPU_ANALYSIS_DIR"
-  reloadable_meta: "./data/model/paddle/fluid_time_file"
-  reloadable_type: "timestamp_ne"
-  model_data_path: "./data/model/paddle/fluid/ctr_prediction"
-  runtime_thread_num: 0
-  batch_infer_size: 0
-  enable_batch_align: 0
-  sparse_param_service_type: REMOTE
-  sparse_param_service_table_name: "test_dict"
-}
-
-engines {
-  name: "bert"
-  type: "FLUID_CPU_ANALYSIS_DIR"
-  reloadable_meta: "./data/model/paddle/fluid_time_file"
-  reloadable_type: "timestamp_ne"
-  model_data_path: "./data/model/paddle/fluid/bert_cased_L-12_H-768_A-12"
-  runtime_thread_num: 0
-  batch_infer_size: 0
-  enable_batch_align: 0
-  enable_memory_optimization: true
-}

examples/demo-serving/conf/service.prototxt

-services {
-  name: "BuiltinDenseFormatService"
-  workflows: "workflow1"
-}
-services {
-  name: "BuiltinSparseFormatService"
-  workflows: "workflow2"
-}
-
-services {
-  name: "BuiltinTestEchoService"
-  workflows: "workflow3"
-}
-
-services {
-  name: "ImageClassifyService"
-  workflows: "workflow4"
-}
-
-services {
-  name: "BuiltinFluidService"
-  workflows: "workflow5"
-}
-services {
-  name: "TextClassificationService"
-  workflows: "workflow6"
-}
-
-services {
-  name: "EchoKVDBService"
-  workflows: "workflow7"
-}
-
-services {
-  name: "CTRPredictionService"
-  workflows: "workflow8"
-}
-services {
-  name: "BertService"
-  workflows: "workflow9"
-}
+port: 9292
 services {
   name: "GeneralModelService"
   workflows: "workflow11"

examples/demo-serving/conf/workflow.prototxt

-workflows {
-  name: "workflow1"
-  workflow_type: "Sequence"
-  nodes {
-    name: "dense_echo_op"
-    type: "DenseEchoOp"
-  }
-}
-
-workflows {
-  name: "workflow2"
-  workflow_type: "Sequence"
-  nodes {
-    name: "sparse_echo_op"
-    type: "SparseEchoOp"
-    dependencies {
-      name: "startup_op"
-      mode: "RO"
-    }
-  }
-}
-workflows {
-  name: "workflow3"
-  workflow_type: "Sequence"
-  nodes {
-    name: "echo_op"
-    type: "CommonEchoOp"
-  }
-}
-workflows {
-  name: "workflow4"
-  workflow_type: "Sequence"
-  nodes {
-    name: "image_reader_op"
-    type: "ReaderOp"
-  }
-  nodes {
-    name: "image_classify_op"
-    type: "ClassifyOp"
-    dependencies {
-      name: "image_reader_op"
-      mode: "RO"
-    }
-  }
-  nodes {
-    name: "write_json_op"
-    type: "WriteJsonOp"
-    dependencies {
-      name: "image_classify_op"
-      mode: "RO"
-    }
-  }
-}
-workflows {
-  name: "workflow5"
-  workflow_type: "Sequence"
-  nodes {
-    name: "int64tensor_echo_op"
-    type: "Int64TensorEchoOp"
-  }
-}
-workflows {
-  name: "workflow6"
-  workflow_type: "Sequence"
-  nodes {
-    name: "text_classify_op"
-    type: "TextClassificationOp"
-  }
-}
-workflows {
-  name: "workflow7"
-  workflow_type: "Sequence"
-  nodes {
-    name: "echo_kvdb_service_op"
-    type: "KVDBEchoOp"
-  } 
-}
-workflows {
-  name: "workflow8"
-  workflow_type: "Sequence"
-  nodes {
-    name: "ctr_prediction_service_op"
-    type: "CTRPredictionOp"
-  } 
-}
-workflows {
-  name: "workflow9"
-  workflow_type: "Sequence"
-  nodes {
-    name: "bert_service_op"
-    type: "BertServiceOp"
-  }
-}
-workflows {
-  name: "workflow10"
-  workflow_type: "Sequence"
-  nodes {
-    name: "general_model_op"
-    type: "GeneralModelOp"
-  }
-}
-
 workflows {
   name: "workflow11"
   workflow_type: "Sequence"

examples/demo-serving/kvdb/CURRENT

-MANIFEST-000020

examples/demo-serving/kvdb/IDENTITY

-13d73780-de4f-4b8c-9040-34e5adc9f9ae

examples/demo-serving/kvdb/OPTIONS-000020

-# This is a RocksDB option file.
-#
-# For detailed file format spec, please refer to the example file
-# in examples/rocksdb_option_file_example.ini
-#
-
-[Version]
-  rocksdb_version=6.2.4
-  options_file_version=1.1
-
-[DBOptions]
-  avoid_unnecessary_blocking_io=false
-  allow_mmap_reads=false
-  allow_fallocate=true
-  WAL_size_limit_MB=0
-  writable_file_max_buffer_size=1048576
-  allow_mmap_writes=false
-  allow_concurrent_memtable_write=true
-  use_direct_reads=false
-  max_open_files=-1
-  strict_bytes_per_sync=false
-  db_write_buffer_size=0
-  max_background_jobs=2
-  WAL_ttl_seconds=0
-  enable_thread_tracking=false
-  error_if_exists=false
-  is_fd_close_on_exec=true
-  recycle_log_file_num=0
-  max_manifest_file_size=1073741824
-  skip_log_error_on_recovery=false
-  skip_stats_update_on_db_open=false
-  max_total_wal_size=0
-  new_table_reader_for_compaction_inputs=false
-  manual_wal_flush=false
-  compaction_readahead_size=0
-  atomic_flush=false
-  random_access_max_buffer_size=1048576
-  create_missing_column_families=false
-  wal_bytes_per_sync=0
-  use_adaptive_mutex=false
-  use_direct_io_for_flush_and_compaction=false
-  max_background_compactions=-1
-  advise_random_on_open=true
-  base_background_compactions=-1
-  max_background_flushes=-1
-  two_write_queues=false
-  table_cache_numshardbits=6
-  keep_log_file_num=1000
-  write_thread_slow_yield_usec=3
-  stats_dump_period_sec=600
-  avoid_flush_during_recovery=false
-  log_file_time_to_roll=0
-  delayed_write_rate=16777216
-  manifest_preallocation_size=4194304
-  paranoid_checks=true
-  max_log_file_size=0
-  allow_2pc=false
-  wal_dir=kvdb
-  db_log_dir=
-  max_subcompactions=1
-  create_if_missing=true
-  enable_pipelined_write=false
-  bytes_per_sync=0
-  stats_persist_period_sec=600
-  stats_history_buffer_size=1048576
-  fail_if_options_file_error=false
-  use_fsync=false
-  wal_recovery_mode=kPointInTimeRecovery
-  delete_obsolete_files_period_micros=21600000000
-  enable_write_thread_adaptive_yield=true
-  avoid_flush_during_shutdown=false
-  write_thread_max_yield_usec=100
-  info_log_level=INFO_LEVEL
-  max_file_opening_threads=16
-  dump_malloc_stats=false
-  allow_ingest_behind=false
-  access_hint_on_compaction_start=NORMAL
-  preserve_deletes=false
-  
-
-[CFOptions "default"]
-  sample_for_compression=0
-  compaction_pri=kMinOverlappingRatio
-  merge_operator=nullptr
-  compaction_filter_factory=nullptr
-  memtable_factory=SkipListFactory
-  memtable_insert_with_hint_prefix_extractor=nullptr
-  comparator=leveldb.BytewiseComparator
-  target_file_size_base=67108864
-  max_sequential_skip_in_iterations=8
-  compaction_style=kCompactionStyleLevel
-  max_bytes_for_level_base=268435456
-  bloom_locality=0
-  write_buffer_size=67108864
-  compression_per_level=
-  memtable_huge_page_size=0
-  max_successive_merges=0
-  arena_block_size=8388608
-  memtable_whole_key_filtering=false
-  target_file_size_multiplier=1
-  max_bytes_for_level_multiplier_additional=1:1:1:1:1:1:1
-  snap_refresh_nanos=0
-  num_levels=7
-  min_write_buffer_number_to_merge=1
-  max_write_buffer_number_to_maintain=0
-  max_write_buffer_number=2
-  compression=kNoCompression
-  level0_stop_writes_trigger=36
-  level0_slowdown_writes_trigger=20
-  compaction_filter=nullptr
-  level0_file_num_compaction_trigger=4
-  max_compaction_bytes=1677721600
-  compaction_options_universal={allow_trivial_move=false;size_ratio=1;min_merge_width=2;max_size_amplification_percent=200;max_merge_width=4294967295;compression_size_percent=-1;stop_style=kCompactionStopStyleTotalSize;}
-  memtable_prefix_bloom_size_ratio=0.000000
-  hard_pending_compaction_bytes_limit=274877906944
-  ttl=0
-  table_factory=BlockBasedTable
-  soft_pending_compaction_bytes_limit=68719476736
-  prefix_extractor=nullptr
-  bottommost_compression=kDisableCompressionOption
-  force_consistency_checks=false
-  paranoid_file_checks=false
-  compaction_options_fifo={allow_compaction=false;max_table_files_size=1073741824;}
-  max_bytes_for_level_multiplier=10.000000
-  optimize_filters_for_hits=false
-  level_compaction_dynamic_level_bytes=false
-  inplace_update_num_locks=10000
-  inplace_update_support=false
-  periodic_compaction_seconds=0
-  disable_auto_compactions=false
-  report_bg_io_stats=false
-  
-[TableOptions/BlockBasedTable "default"]
-  pin_top_level_index_and_filter=true
-  enable_index_compression=true
-  read_amp_bytes_per_bit=8589934592
-  format_version=2
-  block_align=false
-  metadata_block_size=4096
-  block_size_deviation=10
-  partition_filters=false
-  block_size=4096
-  index_block_restart_interval=1
-  no_block_cache=false
-  checksum=kCRC32c
-  whole_key_filtering=true
-  index_shortening=kShortenSeparators
-  data_block_index_type=kDataBlockBinarySearch
-  index_type=kBinarySearch
-  verify_compression=false
-  filter_policy=nullptr
-  data_block_hash_table_util_ratio=0.750000
-  pin_l0_filter_and_index_blocks_in_cache=false
-  block_restart_interval=16
-  cache_index_and_filter_blocks_with_high_priority=false
-  cache_index_and_filter_blocks=false
-  hash_index_allow_collision=true
-  flush_block_policy_factory=FlushBlockBySizePolicyFactory
-  

examples/demo-serving/kvdb/OPTIONS-000023

-# This is a RocksDB option file.
-#
-# For detailed file format spec, please refer to the example file
-# in examples/rocksdb_option_file_example.ini
-#
-
-[Version]
-  rocksdb_version=6.2.4
-  options_file_version=1.1
-
-[DBOptions]
-  avoid_unnecessary_blocking_io=false
-  allow_mmap_reads=false
-  allow_fallocate=true
-  WAL_size_limit_MB=0
-  writable_file_max_buffer_size=1048576
-  allow_mmap_writes=false
-  allow_concurrent_memtable_write=true
-  use_direct_reads=false
-  max_open_files=-1
-  strict_bytes_per_sync=false
-  db_write_buffer_size=0
-  max_background_jobs=2
-  WAL_ttl_seconds=0
-  enable_thread_tracking=false
-  error_if_exists=false
-  is_fd_close_on_exec=true
-  recycle_log_file_num=0
-  max_manifest_file_size=1073741824
-  skip_log_error_on_recovery=false
-  skip_stats_update_on_db_open=false
-  max_total_wal_size=0
-  new_table_reader_for_compaction_inputs=false
-  manual_wal_flush=false
-  compaction_readahead_size=0
-  atomic_flush=false
-  random_access_max_buffer_size=1048576
-  create_missing_column_families=false
-  wal_bytes_per_sync=0
-  use_adaptive_mutex=false
-  use_direct_io_for_flush_and_compaction=false
-  max_background_compactions=-1
-  advise_random_on_open=true
-  base_background_compactions=-1
-  max_background_flushes=-1
-  two_write_queues=false
-  table_cache_numshardbits=6
-  keep_log_file_num=1000
-  write_thread_slow_yield_usec=3
-  stats_dump_period_sec=600
-  avoid_flush_during_recovery=false
-  log_file_time_to_roll=0
-  delayed_write_rate=16777216
-  manifest_preallocation_size=4194304
-  paranoid_checks=true
-  max_log_file_size=0
-  allow_2pc=false
-  wal_dir=kvdb
-  db_log_dir=
-  max_subcompactions=1
-  create_if_missing=true
-  enable_pipelined_write=false
-  bytes_per_sync=0
-  stats_persist_period_sec=600
-  stats_history_buffer_size=1048576
-  fail_if_options_file_error=false
-  use_fsync=false
-  wal_recovery_mode=kPointInTimeRecovery
-  delete_obsolete_files_period_micros=21600000000
-  enable_write_thread_adaptive_yield=true
-  avoid_flush_during_shutdown=false
-  write_thread_max_yield_usec=100
-  info_log_level=INFO_LEVEL
-  max_file_opening_threads=16
-  dump_malloc_stats=false
-  allow_ingest_behind=false
-  access_hint_on_compaction_start=NORMAL
-  preserve_deletes=false
-  
-
-[CFOptions "default"]
-  sample_for_compression=0
-  compaction_pri=kMinOverlappingRatio
-  merge_operator=nullptr
-  compaction_filter_factory=nullptr
-  memtable_factory=SkipListFactory
-  memtable_insert_with_hint_prefix_extractor=nullptr
-  comparator=leveldb.BytewiseComparator
-  target_file_size_base=67108864
-  max_sequential_skip_in_iterations=8
-  compaction_style=kCompactionStyleLevel
-  max_bytes_for_level_base=268435456
-  bloom_locality=0
-  write_buffer_size=67108864
-  compression_per_level=
-  memtable_huge_page_size=0
-  max_successive_merges=0
-  arena_block_size=8388608
-  memtable_whole_key_filtering=false
-  target_file_size_multiplier=1
-  max_bytes_for_level_multiplier_additional=1:1:1:1:1:1:1
-  snap_refresh_nanos=0
-  num_levels=7
-  min_write_buffer_number_to_merge=1
-  max_write_buffer_number_to_maintain=0
-  max_write_buffer_number=2
-  compression=kNoCompression
-  level0_stop_writes_trigger=36
-  level0_slowdown_writes_trigger=20
-  compaction_filter=nullptr
-  level0_file_num_compaction_trigger=4
-  max_compaction_bytes=1677721600
-  compaction_options_universal={allow_trivial_move=false;size_ratio=1;min_merge_width=2;max_size_amplification_percent=200;max_merge_width=4294967295;compression_size_percent=-1;stop_style=kCompactionStopStyleTotalSize;}
-  memtable_prefix_bloom_size_ratio=0.000000
-  hard_pending_compaction_bytes_limit=274877906944
-  ttl=0
-  table_factory=BlockBasedTable
-  soft_pending_compaction_bytes_limit=68719476736
-  prefix_extractor=nullptr
-  bottommost_compression=kDisableCompressionOption
-  force_consistency_checks=false
-  paranoid_file_checks=false
-  compaction_options_fifo={allow_compaction=false;max_table_files_size=1073741824;}
-  max_bytes_for_level_multiplier=10.000000
-  optimize_filters_for_hits=false
-  level_compaction_dynamic_level_bytes=false
-  inplace_update_num_locks=10000
-  inplace_update_support=false
-  periodic_compaction_seconds=0
-  disable_auto_compactions=false
-  report_bg_io_stats=false
-  
-[TableOptions/BlockBasedTable "default"]
-  pin_top_level_index_and_filter=true
-  enable_index_compression=true
-  read_amp_bytes_per_bit=8589934592
-  format_version=2
-  block_align=false
-  metadata_block_size=4096
-  block_size_deviation=10
-  partition_filters=false
-  block_size=4096
-  index_block_restart_interval=1
-  no_block_cache=false
-  checksum=kCRC32c
-  whole_key_filtering=true
-  index_shortening=kShortenSeparators
-  data_block_index_type=kDataBlockBinarySearch
-  index_type=kBinarySearch
-  verify_compression=false
-  filter_policy=nullptr
-  data_block_hash_table_util_ratio=0.750000
-  pin_l0_filter_and_index_blocks_in_cache=false
-  block_restart_interval=16
-  cache_index_and_filter_blocks_with_high_priority=false
-  cache_index_and_filter_blocks=false
-  hash_index_allow_collision=true
-  flush_block_policy_factory=FlushBlockBySizePolicyFactory
-  

examples/demo-serving/op/general_model_op.cpp

@@ -51,14 +51,14 @@ int GeneralModelOp::inference() {
   // infer
   if (batch_size > 0) {
     int var_num = req->insts(0).tensor_array_size();
-    VLOG(3) << "var num: " << var_num;
+    VLOG(2) << "var num: " << var_num;
     elem_type.resize(var_num);
     elem_size.resize(var_num);
     capacity.resize(var_num);
     paddle::PaddleTensor lod_tensor;
     for (int i = 0; i < var_num; ++i) {
       elem_type[i] = req->insts(0).tensor_array(i).elem_type();
-      VLOG(3) << "var[" << i << "] has elem type: " << elem_type[i];
+      VLOG(2) << "var[" << i << "] has elem type: " << elem_type[i];
       if (elem_type[i] == 0) {  // int64
         elem_size[i] = sizeof(int64_t);
         lod_tensor.dtype = paddle::PaddleDType::INT64;
@@ -70,17 +70,17 @@ int GeneralModelOp::inference() {
       if (req->insts(0).tensor_array(i).shape(0) == -1) {
         lod_tensor.lod.resize(1);
         lod_tensor.lod[0].push_back(0);
-        VLOG(3) << "var[" << i << "] is lod_tensor";
+        VLOG(2) << "var[" << i << "] is lod_tensor";
       } else {
         lod_tensor.shape.push_back(batch_size);
         capacity[i] = 1;
         for (int k = 0; k < req->insts(0).tensor_array(i).shape_size(); ++k) {
           int dim = req->insts(0).tensor_array(i).shape(k);
-          VLOG(3) << "shape for var[" << i << "]: " << dim;
+          VLOG(2) << "shape for var[" << i << "]: " << dim;
           capacity[i] *= dim;
           lod_tensor.shape.push_back(dim);
         }
-        VLOG(3) << "var[" << i << "] is tensor, capacity: " << capacity[i];
+        VLOG(2) << "var[" << i << "] is tensor, capacity: " << capacity[i];
       }
       if (i == 0) {
         lod_tensor.name = "words";
@@ -95,19 +95,19 @@ int GeneralModelOp::inference() {
         for (int j = 0; j < batch_size; ++j) {
           const Tensor &tensor = req->insts(j).tensor_array(i);
           int data_len = tensor.data_size();
-          VLOG(3) << "tensor size for var[" << i << "]: " << tensor.data_size();
+          VLOG(2) << "tensor size for var[" << i << "]: " << tensor.data_size();
           int cur_len = in->at(i).lod[0].back();
-          VLOG(3) << "current len: " << cur_len;
+          VLOG(2) << "current len: " << cur_len;
           in->at(i).lod[0].push_back(cur_len + data_len);
-          VLOG(3) << "new len: " << cur_len + data_len;
+          VLOG(2) << "new len: " << cur_len + data_len;
         }
         in->at(i).data.Resize(in->at(i).lod[0].back() * elem_size[i]);
         in->at(i).shape = {in->at(i).lod[0].back(), 1};
-        VLOG(3) << "var[" << i
+        VLOG(2) << "var[" << i
                 << "] is lod_tensor and len=" << in->at(i).lod[0].back();
       } else {
         in->at(i).data.Resize(batch_size * capacity[i] * elem_size[i]);
-        VLOG(3) << "var[" << i
+        VLOG(2) << "var[" << i
                 << "] is tensor and capacity=" << batch_size * capacity[i];
       }
     }
@@ -144,7 +144,7 @@ int GeneralModelOp::inference() {
       }
     }
 
-    VLOG(3) << "going to infer";
+    VLOG(2) << "going to infer";
     TensorVector *out = butil::get_object<TensorVector>();
     if (!out) {
       LOG(ERROR) << "Failed get tls output object";
@@ -157,7 +157,7 @@ int GeneralModelOp::inference() {
     for (uint32_t i = 0; i < 10; i++) {
       oss << *(example + i) << " ";
     }
-    VLOG(3) << "msg: " << oss.str();
+    VLOG(2) << "msg: " << oss.str();
 
     // infer
     if (predictor::InferManager::instance().infer(
@@ -167,7 +167,7 @@ int GeneralModelOp::inference() {
     }
     // print response
     float *example_1 = reinterpret_cast<float *>((*out)[0].data.data());
-    VLOG(3) << "result: " << *example_1;
+    VLOG(2) << "result: " << *example_1;
 
     Response *res = mutable_data<Response>();
 

examples/demo-serving/op/general_reader_op.cpp

@@ -33,16 +33,22 @@ using baidu::paddle_serving::predictor::PaddleGeneralModelConfig;
 int conf_check(const Request *req,
                const std::shared_ptr<PaddleGeneralModelConfig> &model_config) {
   int var_num = req->insts(0).tensor_array_size();
+  VLOG(2) << "var num: " << var_num;
   if (var_num != model_config->_feed_type.size()) {
     LOG(ERROR) << "feed var number not match.";
     return -1;
   }
+
+  VLOG(2) << "begin to checkout feed type";
   for (int i = 0; i < var_num; ++i) {
+    VLOG(2) << "feed type[" << i << "]: " <<
+        model_config->_feed_type[i];
     if (model_config->_feed_type[i] !=
         req->insts(0).tensor_array(i).elem_type()) {
       LOG(ERROR) << "feed type not match.";
       return -1;
     }
+    VLOG(2) << "feed shape size: " << model_config->_feed_shape[i].size();
     if (model_config->_feed_shape[i].size() ==
         req->insts(0).tensor_array(i).shape_size()) {
       for (int j = 0; j < model_config->_feed_shape[i].size(); ++j) {
@@ -84,7 +90,7 @@ int GeneralReaderOp::inference() {
   }
 
   int var_num = req->insts(0).tensor_array_size();
-  VLOG(3) << "var num: " << var_num;
+  VLOG(2) << "var num: " << var_num;
   // read config
 
   LOG(INFO) << "start to call load general model_conf op";
@@ -112,7 +118,7 @@ int GeneralReaderOp::inference() {
   paddle::PaddleTensor lod_tensor;
   for (int i = 0; i < var_num; ++i) {
     elem_type[i] = req->insts(0).tensor_array(i).elem_type();
-    VLOG(3) << "var[" << i << "] has elem type: " << elem_type[i];
+    VLOG(2) << "var[" << i << "] has elem type: " << elem_type[i];
     if (elem_type[i] == 0) {  // int64
       elem_size[i] = sizeof(int64_t);
       lod_tensor.dtype = paddle::PaddleDType::INT64;
@@ -124,17 +130,17 @@ int GeneralReaderOp::inference() {
     if (req->insts(0).tensor_array(i).shape(0) == -1) {
       lod_tensor.lod.resize(1);
       lod_tensor.lod[0].push_back(0);
-      VLOG(3) << "var[" << i << "] is lod_tensor";
+      VLOG(2) << "var[" << i << "] is lod_tensor";
     } else {
       lod_tensor.shape.push_back(batch_size);
       capacity[i] = 1;
       for (int k = 0; k < req->insts(0).tensor_array(i).shape_size(); ++k) {
         int dim = req->insts(0).tensor_array(i).shape(k);
-        VLOG(3) << "shape for var[" << i << "]: " << dim;
+        VLOG(2) << "shape for var[" << i << "]: " << dim;
         capacity[i] *= dim;
         lod_tensor.shape.push_back(dim);
       }
-      VLOG(3) << "var[" << i << "] is tensor, capacity: " << capacity[i];
+      VLOG(2) << "var[" << i << "] is tensor, capacity: " << capacity[i];
     }
     if (i == 0) {
       lod_tensor.name = "words";
@@ -149,19 +155,19 @@ int GeneralReaderOp::inference() {
       for (int j = 0; j < batch_size; ++j) {
         const Tensor &tensor = req->insts(j).tensor_array(i);
         int data_len = tensor.data_size();
-        VLOG(3) << "tensor size for var[" << i << "]: " << tensor.data_size();
+        VLOG(2) << "tensor size for var[" << i << "]: " << tensor.data_size();
         int cur_len = in->at(i).lod[0].back();
-        VLOG(3) << "current len: " << cur_len;
+        VLOG(2) << "current len: " << cur_len;
         in->at(i).lod[0].push_back(cur_len + data_len);
-        VLOG(3) << "new len: " << cur_len + data_len;
+        VLOG(2) << "new len: " << cur_len + data_len;
       }
       in->at(i).data.Resize(in->at(i).lod[0].back() * elem_size[i]);
       in->at(i).shape = {in->at(i).lod[0].back(), 1};
-      VLOG(3) << "var[" << i
+      VLOG(2) << "var[" << i
               << "] is lod_tensor and len=" << in->at(i).lod[0].back();
     } else {
       in->at(i).data.Resize(batch_size * capacity[i] * elem_size[i]);
-      VLOG(3) << "var[" << i
+      VLOG(2) << "var[" << i
               << "] is tensor and capacity=" << batch_size * capacity[i];
     }
   }
@@ -198,14 +204,14 @@ int GeneralReaderOp::inference() {
     }
   }
 
-  VLOG(3) << "read data from client success";
+  VLOG(2) << "read data from client success";
   // print request
   std::ostringstream oss;
   int64_t *example = reinterpret_cast<int64_t *>((*in)[0].data.data());
   for (int i = 0; i < 10; i++) {
     oss << *(example + i) << " ";
   }
-  VLOG(3) << "head element of first feed var : " << oss.str();
+  VLOG(2) << "head element of first feed var : " << oss.str();
   //
   return 0;
 }

python/CMakeLists.txt

 if (CLIENT_ONLY)
-file(GLOB_RECURSE SERVING_CLIENT_PY_FILES serving_client/*.py)
+file(GLOB_RECURSE SERVING_CLIENT_PY_FILES paddle_serving_client/*.py)
 set(PY_FILES ${SERVING_CLIENT_PY_FILES})
 SET(PACKAGE_NAME "serving_client")
 set(SETUP_LOG_FILE "setup.py.client.log")
 endif()
 
 if (NOT CLIENT_ONLY)
-file(GLOB_RECURSE SERVING_SERVER_PY_FILES serving_server/*.py)
+file(GLOB_RECURSE SERVING_SERVER_PY_FILES paddle_serving_server/*.py)
 set(PY_FILES ${SERVING_SERVER_PY_FILES})
 SET(PACKAGE_NAME "serving_server")
 set(SETUP_LOG_FILE "setup.py.server.log")
@@ -29,8 +29,8 @@ message("python env: " ${py_env})
 if (CLIENT_ONLY)
 add_custom_command(
 	OUTPUT ${PADDLE_SERVING_BINARY_DIR}/.timestamp
-	COMMAND cp -r ${CMAKE_CURRENT_SOURCE_DIR}/paddle_serving/ ${PADDLE_SERVING_BINARY_DIR}/python/
-	COMMAND ${CMAKE_COMMAND} -E copy ${PADDLE_SERVING_BINARY_DIR}/core/general-client/serving_client.so ${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving/serving_client/
+	COMMAND cp -r ${CMAKE_CURRENT_SOURCE_DIR}/paddle_serving_client/ ${PADDLE_SERVING_BINARY_DIR}/python/
+	COMMAND ${CMAKE_COMMAND} -E copy ${PADDLE_SERVING_BINARY_DIR}/core/general-client/serving_client.so ${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_client/
 	COMMAND env ${py_env} ${PYTHON_EXECUTABLE} setup.py bdist_wheel
 	DEPENDS ${SERVING_CLIENT_CORE} sdk_configure_py_proto ${PY_FILES})
 add_custom_target(paddle_python ALL DEPENDS serving_client ${PADDLE_SERVING_BINARY_DIR}/.timestamp)
@@ -39,7 +39,7 @@ endif()
 if (NOT CLIENT_ONLY)
 add_custom_command(
    	OUTPUT ${PADDLE_SERVING_BINARY_DIR}/.timestamp
-	COMMAND cp -r ${CMAKE_CURRENT_SOURCE_DIR}/paddle_serving/ ${PADDLE_SERVING_BINARY_DIR}/python/
+	COMMAND cp -r ${CMAKE_CURRENT_SOURCE_DIR}/paddle_serving_server/ ${PADDLE_SERVING_BINARY_DIR}/python/
 	COMMAND env ${py_env} ${PYTHON_EXECUTABLE} setup.py bdist_wheel
 	DEPENDS ${SERVING_SERVER_CORE} server_config_py_proto ${PY_FILES})
 add_custom_target(paddle_python ALL DEPENDS ${PADDLE_SERVING_BINARY_DIR}/.timestamp)

python/paddle_serving/serving_client/__init__.py → python/paddle_serving_client/__init__.py

@@ -13,7 +13,9 @@
 # limitations under the License.
 
 from .serving_client import PredictorClient
-from ..proto import sdk_configure_pb2 as sdk
+from .proto import sdk_configure_pb2 as sdk
+from .proto import general_model_config_pb2 as m_config
+import google.protobuf.text_format
 import time
 
 int_type = 0
@@ -74,34 +76,25 @@ class Client(object):
         self.feed_names_to_idx_ = {}
 
     def load_client_config(self, path):
+        model_conf = m_config.GeneralModelConfig()
+        f = open(path, 'r')
+        model_conf = google.protobuf.text_format.Merge(
+            str(f.read()), model_conf)
+
         # load configuraion here
         # get feed vars, fetch vars
         # get feed shapes, feed types
         # map feed names to index
         self.client_handle_ = PredictorClient()
         self.client_handle_.init(path)
-        self.feed_names_ = []
-        self.fetch_names_ = []
-        self.feed_shapes_ = []
-        self.feed_types_ = {}
+        self.feed_names_ = [var.alias_name for var in model_conf.feed_var]
+        self.fetch_names_ = [var.alias_name for var in model_conf.fetch_var]
+        self.feed_shapes_ = [var.shape for var in model_conf.feed_var]
         self.feed_names_to_idx_ = {}
+        for i, var in enumerate(model_conf.feed_var):
+            self.feed_names_to_idx_[var.alias_name] = i
+            self.feed_types_[var.alias_name] = var.feed_type
 
-        with open(path) as fin:
-            group = fin.readline().strip().split()
-            feed_num = int(group[0])
-            fetch_num = int(group[1])
-            for i in range(feed_num):
-                group = fin.readline().strip().split()
-                self.feed_names_.append(group[0])
-                tmp_shape = []
-                for s in group[2:-1]:
-                    tmp_shape.append(int(s))
-                self.feed_shapes_.append(tmp_shape)
-                self.feed_types_[group[0]] = int(group[-1])
-                self.feed_names_to_idx_[group[0]] = i
-            for i in range(fetch_num):
-                group = fin.readline().strip().split()
-                self.fetch_names_.append(group[0])
         return
 
     def connect(self, endpoints):

python/paddle_serving/io/__init__.py → python/paddle_serving_client/io/__init__.py

@@ -12,13 +12,14 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-
 from paddle.fluid import Executor
 from paddle.fluid.compiler import CompiledProgram
-from paddle.fluid.framework import Program
+from paddle.fluid.framework import core
 from paddle.fluid.framework import default_main_program
+from paddle.fluid.framework import Program
 from paddle.fluid import CPUPlace
 from paddle.fluid.io import save_persistables
+from ..proto import general_model_config_pb2 as model_conf
 import os
 
 def save_model(server_model_folder,
@@ -40,25 +41,42 @@ def save_model(server_model_folder,
     save_persistables(executor, server_model_folder,
                       main_program)
 
+    config = model_conf.GeneralModelConfig()
+
+    for key in feed_var_dict:
+        feed_var = model_conf.FeedVar()
+        feed_var.alias_name = key
+        feed_var.name = feed_var_dict[key].name
+        feed_var.is_lod_tensor = feed_var_dict[key].lod_level == 1
+        if feed_var_dict[key].dtype == core.VarDesc.VarType.INT32 or \
+           feed_var_dict[key].dtype == core.VarDesc.VarType.INT64:
+            feed_var.feed_type = 0
+        if feed_var_dict[key].dtype == core.VarDesc.VarType.FP32:
+            feed_var.feed_type = 1
+        if feed_var.is_lod_tensor:
+            feed_var.shape.extend([-1])
+        else:
+            tmp_shape = []
+            for v in feed_var_dict[key].shape:
+                if v >= 0:
+                    tmp_shape.append(v)
+            feed_var.shape.extend(tmp_shape)
+        config.feed_var.extend([feed_var])
+
+    for key in fetch_var_dict:
+        fetch_var = model_conf.FetchVar()
+        fetch_var.alias_name = key
+        fetch_var.name = fetch_var_dict[key].name
+        fetch_var.shape.extend(fetch_var_dict[key].shape)
+        config.fetch_var.extend([fetch_var])
+
     cmd = "mkdir -p {}".format(client_config_folder)
     os.system(cmd)
-    with open("{}/client.conf".format(client_config_folder), "w") as fout:
-        fout.write("{} {}\n".format(len(feed_var_dict), len(fetch_var_dict)))
-        for key in feed_var_dict:
-            fout.write("{}".format(key))
-            if feed_var_dict[key].lod_level == 1:
-                fout.write(" 1 -1\n")
-            elif feed_var_dict[key].lod_level == 0:
-                fout.write(" {}".format(len(feed_var_dict[key].shape)))
-                for dim in feed_var_dict[key].shape:
-                    fout.write(" {}".format(dim))
-                fout.write("\n")
-        for key in fetch_var_dict:
-            fout.write("{} {}\n".format(key, fetch_var_dict[key].name))
+    with open("{}/serving_client_conf.prototxt", "w") as fout:
+        fout.write(str(config))
+    with open("{}/serving_server_conf.prototxt", "w") as fout:
+        fout.write(str(config))
 
-    cmd = "cp {}/client.conf {}/server.conf".format(
-        client_config_folder, server_model_folder)
-    os.system(cmd)
 
     
 

python/paddle_serving/serving_server/__init__.py → python/paddle_serving_server/__init__.py

@@ -13,7 +13,9 @@
 # limitations under the License.
 
 import os
-from ..proto import server_configure_pb2 as server_sdk
+from .proto import server_configure_pb2 as server_sdk
+from .proto import general_model_config_pb2 as m_config
+import google.protobuf.text_format
 
 class OpMaker(object):
     def __init__(self):
@@ -58,10 +60,13 @@ class Server(object):
         self.model_toolkit_conf = None
         self.resource_conf = None
         self.engine = None
+        self.memory_optimization = False
+        self.model_conf = None
         self.workflow_fn = "workflow.prototxt"
         self.resource_fn = "resource.prototxt"
         self.infer_service_fn = "infer_service.prototxt"
         self.model_toolkit_fn = "model_toolkit.prototxt"
+        self.general_model_config_fn = "general_model.prototxt"
         self.workdir = ""
         self.max_concurrency = 0
         self.num_threads = 0
@@ -83,11 +88,17 @@ class Server(object):
     def set_op_sequence(self, op_seq):
         self.workflow_conf = op_seq
 
+    def set_memory_optimize(self, flag=False):
+        self.memory_optimization = flag
+
     def _prepare_engine(self, model_config_path, device):
         if self.model_toolkit_conf == None:
             self.model_toolkit_conf = server_sdk.ModelToolkitConf()
+
         if self.engine == None:
             self.engine = server_sdk.EngineDesc()
+
+        self.model_config_path = model_config_path
         self.engine.name = "general_model"
         self.engine.reloadable_meta = model_config_path + "/fluid_time_file"
         self.engine.reloadable_type = "timestamp_ne"
@@ -95,13 +106,15 @@ class Server(object):
         self.engine.batch_infer_size = 0
         self.engine.enable_batch_align = 0
         self.engine.model_data_path = model_config_path
-        self.engine.enable_memory_optimization = True
+        self.engine.enable_memory_optimization = self.memory_optimization
         self.engine.static_optimization = False
         self.engine.force_update_static_cache = False
+
         if device == "cpu":
             self.engine.type = "FLUID_CPU_ANALYSIS_DIR"
         elif device == "gpu":
             self.engine.type = "FLUID_GPU_ANALYSIS_DIR"
+
         self.model_toolkit_conf.engines.extend([self.engine])
 
     def _prepare_infer_service(self, port):
@@ -115,18 +128,26 @@ class Server(object):
 
     def _prepare_resource(self, workdir):
         if self.resource_conf == None:
+            with open("{}/{}".format(workdir, self.general_model_config_fn), "w") as fout:
+                fout.write(str(self.model_conf))
             self.resource_conf = server_sdk.ResourceConf()
             self.resource_conf.model_toolkit_path = workdir
             self.resource_conf.model_toolkit_file = self.model_toolkit_fn
+            self.resource_conf.general_model_path = workdir
+            self.resource_conf.general_model_file = self.general_model_config_fn
 
     def _write_pb_str(self, filepath, pb_obj):
         with open(filepath, "w") as fout:
             fout.write(str(pb_obj))
 
     def load_model_config(self, path):
-        self.config_file = "{}/inference.conf".format(path)
         self.model_config_path = path
+        self.model_conf = m_config.GeneralModelConfig()
+        f = open("{}/serving_server_conf.prototxt".format(path), 'r')
+        self.model_conf = google.protobuf.text_format.Merge(
+            str(f.read()), self.model_conf)
         # check config here
+        # print config here
 
     def prepare_server(self, workdir=None, port=9292, device="cpu"):
         if workdir == None:
@@ -154,7 +175,7 @@ class Server(object):
     def run_server(self):
         # just run server with system command
         # currently we do not load cube
-        command = "/home/users/dongdaxiang/github_develop/Serving/examples/demo-serving/serving" \
+        command = "/home/users/dongdaxiang/github_develop/Serving/build_server/core/general-server" \
                   " -enable_model_toolkit " \
                   "-inferservice_path {} " \
                   "-inferservice_file {} " \

python/paddle_serving/__init__.py → python/paddle_serving_server/version.py

-#   Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+# Copyright (c) 2020  PaddlePaddle Authors. All Rights Reserved.
 #
-# Licensed under the Apache License, Version 2.0 (the "License");
+# Licensed under the Apache License, Version 2.0 (the "License"
 # you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at
 #
@@ -11,4 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from .io import save_model
+""" Paddle Serving Client version string """
+serving_client_version = "0.1.0"
+serving_server_version = "0.1.0"
+module_proto_version = "0.1.0"

python/setup.py.client.in

@@ -21,7 +21,7 @@ import platform
 from setuptools import setup, Distribution, Extension
 from setuptools import find_packages
 from setuptools import setup
-from paddle_serving.version import serving_client_version
+from paddle_serving_client.version import serving_client_version
 
 def python_version():
     return [int(v) for v in platform.python_version().split(".")]
@@ -32,17 +32,16 @@ REQUIRED_PACKAGES = [
     'six >= 1.10.0', 'protobuf >= 3.1.0','paddlepaddle'
 ]
 
-packages=['paddle_serving',
-          'paddle_serving.serving_client',
-          'paddle_serving.proto',
-          'paddle_serving.io']
-package_data={'paddle_serving.serving_client': ['serving_client.so']}
-package_dir={'paddle_serving.serving_client':
-             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving/serving_client',
-             'paddle_serving.proto':
-             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving/proto',
-             'paddle_serving.io':
-             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving/io'}
+packages=['paddle_serving_client',
+          'paddle_serving_client.proto',
+          'paddle_serving_client.io']
+package_data={'paddle_serving_client': ['serving_client.so']}
+package_dir={'paddle_serving_client':
+             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_client',
+             'paddle_serving_client.proto':
+             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_client/proto',
+             'paddle_serving_client.io':
+             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_client/io'}
 
 setup(
     name='paddle-serving-client',

python/setup.py.server.in

@@ -21,8 +21,7 @@ import platform
 from setuptools import setup, Distribution, Extension
 from setuptools import find_packages
 from setuptools import setup
-from paddle_serving.version import serving_client_version
-from paddle_serving.version import serving_server_version
+from paddle_serving_server.version import serving_server_version
 
 def python_version():
     return [int(v) for v in platform.python_version().split(".")]
@@ -33,17 +32,13 @@ REQUIRED_PACKAGES = [
     'six >= 1.10.0', 'protobuf >= 3.1.0','paddlepaddle'
 ]
 
-packages=['paddle_serving',
-          'paddle_serving.serving_server',
-          'paddle_serving.proto',
-          'paddle_serving.io']
-package_data={'paddle_serving.serving_server': ['serving_server.so']}
-package_dir={'paddle_serving.serving_server':
-             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving/serving_server',
-             'paddle_serving.proto':
-             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving/proto',
-             'paddle_serving.io':
-             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving/io'}
+packages=['paddle_serving_server',
+          'paddle_serving_server.proto']
+
+package_dir={'paddle_serving_server':
+             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_server',
+             'paddle_serving_server.proto':
+             '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_server/proto'}
 
 setup(
     name='paddle-serving-server',
@@ -55,7 +50,6 @@ setup(
     author_email='guru4elephant@gmail.com',
     install_requires=REQUIRED_PACKAGES,
     packages=packages,
-    package_data=package_data,
     package_dir=package_dir,
     # PyPI package information.
     classifiers=[