bigdl-core-npu 2.6.0b20241207__cp310-cp310-win_amd64.whl → 2.6.0b20241212__cp310-cp310-win_amd64.whl

bigdl-core-npu/include/common.h

@@ -80,6 +80,12 @@ struct tokenizer_params {
     std::vector<int32_t> eos_token_id;
 };
 
+struct npu_generation_params {
+    // may add more later when dealing with more cases
+    float repetition_penalty;
+    int32_t max_new_token;
+};
+
 
 #ifdef __cplusplus
 extern "C" {

bigdl-core-npu/include/npu_llm.h

@@ -51,16 +51,20 @@ extern "C" {
 
     EXPORT_API void load_config_from_file(npu_model_params &model_params, const char* model_path);
 
+    EXPORT_API void load_generation_config_from_file(npu_generation_params &generation_params, const char* model_path);
+
     EXPORT_API std::string add_chat_template(npu_model_params model_params, std::string input_prompt);
 
-    EXPORT_API float* run_prefill(void* void_model, int32_t* embd_inp_ptr, int32_t embd_inp_size);
+    EXPORT_API float* run_prefill(void* void_model, int32_t* embd_inp_ptr, int32_t embd_inp_size, float repetition_penalty);
 
-    EXPORT_API float* run_decode(void* void_model, int32_t input_token);
+    EXPORT_API float* run_decode(void* void_model, int32_t input_token, float repetition_penalty);
 
     EXPORT_API void run_prefill_with_logits(void* void_model, int32_t* embd_inp_ptr, int32_t embd_inp_size, float* logits, int32_t vocab_size);
 
     EXPORT_API void run_decode_with_logits(void* void_model, int32_t input_token, float* logits, int32_t vocab_size);
 
+    EXPORT_API float* process_logits(float* logits, int32_t vocab_size, int32_t* p_updated_input_ids, int32_t updated_input_id_size, float repetition_penalty);
+
     EXPORT_API int32_t llm_sample_token(float* logits, bool greedy_search, int32_t vocab_size);
 
     EXPORT_API void reset(void* void_model);

{bigdl_core_npu-2.6.0b20241207.dist-info → bigdl_core_npu-2.6.0b20241212.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: bigdl-core-npu
-Version: 2.6.0b20241207
+Version: 2.6.0b20241212
 Summary: Intel® NPU Acceleration Library
 Home-page: https://github.com/intel/intel-npu-acceleration-library
 Author: Alessandro Palla

{bigdl_core_npu-2.6.0b20241207.dist-info → bigdl_core_npu-2.6.0b20241212.dist-info}/RECORD

@@ -1,10 +1,10 @@
 bigdl-core-npu/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-bigdl-core-npu/npu_llm.dll,sha256=S79IS3oBbMwutHYkAUg2NNFlXBAGs8friIpw3Pdd8sA,3918336
-bigdl-core-npu/npu_llm.lib,sha256=V8WB7fKSKDwBopflxjFk0460SyoAS3LNbkYgFJrgqMQ,43846
-bigdl-core-npu/include/common.h,sha256=rQ7aSNAthGaImWtvs2ZetTZoZHN8iayiuc02M5QP-Pw,2666
-bigdl-core-npu/include/npu_llm.h,sha256=dS7_Esxw9Nxz3x07zTei2GEhTgxhS5IGBnyOVf3OS0k,2171
+bigdl-core-npu/npu_llm.dll,sha256=UMqprDkqF5iFB3A1kcVTjHPoTfL6QmmmnCmkexnV1hw,3923456
+bigdl-core-npu/npu_llm.lib,sha256=uG0enTHg0GNdEYMN36DPEKQXuYoioueKYpKXCKR3wyk,44328
+bigdl-core-npu/include/common.h,sha256=AWpPmCEAz6-oGUtq_ONHlELtOzSMqy4CBTMIppJeOII,2819
+bigdl-core-npu/include/npu_llm.h,sha256=Ar35OKx1qVbT3Q8TGTx_TQa6IlwIYBpn6oSo9q-t9YM,2509
 intel_npu_acceleration_library/__init__.py,sha256=ZKTIhGMDjF7P6pF-yX8KWcSXbeHWRk24AO_orsa18f8,536
-intel_npu_acceleration_library/_version.py,sha256=gxQ2ZnCFvE6oV4_1IdijvHBsoLbU8v6S6uIL28FmIYg,112
+intel_npu_acceleration_library/_version.py,sha256=SAbcsL8DKJ2CKOwNc76MqXwyYql3QHQhTPW2Hs3xW5c,112
 intel_npu_acceleration_library/compiler.py,sha256=3IdgqjamSC8MLexDBJypIeZRiWIcTFnvQSU1LPXUr7Y,6225
 intel_npu_acceleration_library/device.py,sha256=9bn8eVXJa5cXIqgfLsQAdkMVtVUQABb8z0-mQik5jRg,7424
 intel_npu_acceleration_library/dtypes.py,sha256=gdd06Wsc9zIZFHlauUEx4xcK9WGTn1Mu6GkuYDJeA-E,4683
@@ -198,7 +198,7 @@ intel_npu_acceleration_library/external/openvino/torch/__init__.py,sha256=RXLzsf
 intel_npu_acceleration_library/functional/__init__.py,sha256=WWKwKOh6Sgovv7mKctA872TbLP98Pg5m5-MREvUmlAA,204
 intel_npu_acceleration_library/functional/scaled_dot_product_attention.py,sha256=yGUcg4tDQOLuUnP1g74cl-ec8TRr2SuAMcNLlN6qLvE,1620
 intel_npu_acceleration_library/lib/Release/cache.json,sha256=CyrSqZUWo0Ec4_7ydOiuKIC0Gm8AybrGdozUqUuHxBw,8840377
-intel_npu_acceleration_library/lib/Release/intel_npu_acceleration_library.dll,sha256=Sss6F7c34ZXCjdIj7rPM8i_GoNjGnfkJjtr6sFbulz0,340480
+intel_npu_acceleration_library/lib/Release/intel_npu_acceleration_library.dll,sha256=T_mV9lEtRh9PD6rtQJRN4BzX0v3WIBO7xXCwuZTTGi0,339968
 intel_npu_acceleration_library/lib/Release/openvino.dll,sha256=m7M119p3JBq2YYJJ2zzCaBDz6XivKK3nNykb8L1cvDU,13244768
 intel_npu_acceleration_library/lib/Release/openvino_auto_batch_plugin.dll,sha256=2v_I9P3Qo0St1bQZMEZscnFOUVvgZQQ0HvQlG3HtTd0,203104
 intel_npu_acceleration_library/lib/Release/openvino_auto_plugin.dll,sha256=e3Aj9CDRHN30dBEdPSk7OCWe52tWfhI4xeXgyFjuDHg,475488
@@ -228,7 +228,7 @@ intel_npu_acceleration_library/nn/functional.py,sha256=UfAKBc0u6RtyaMo14ldH2GpEn
 intel_npu_acceleration_library/nn/linear.py,sha256=Q06SoGQeLaI86nA_ky2GnFC6H2Fw1zyMDILKnpYC2eo,5739
 intel_npu_acceleration_library/nn/llm.py,sha256=P6dz36Yf6BHtzWcftaghC6QaMI_WeRfQwrCbO7fD6hk,15002
 intel_npu_acceleration_library/nn/module.py,sha256=EYxoTq6I_YgBDgTF76GPDxHrT8SupOTDGMzQaomBeq8,12667
-bigdl_core_npu-2.6.0b20241207.dist-info/METADATA,sha256=PC8aIlivmoSyjBJNDtqJ9FzCVlp9BAj5apSfx7cgYGM,1541
-bigdl_core_npu-2.6.0b20241207.dist-info/WHEEL,sha256=tcd-HDpskugT8GYYKyyid0lOlzoZtZdWwcrj5ormtfo,101
-bigdl_core_npu-2.6.0b20241207.dist-info/top_level.txt,sha256=iMQZlTsFPJjlD-Y0MqZEP_9ifI0LlbNCJIOTaMoGMjk,46
-bigdl_core_npu-2.6.0b20241207.dist-info/RECORD,,
+bigdl_core_npu-2.6.0b20241212.dist-info/METADATA,sha256=WjfT1msnAHDOr21DLxWpjCi6jS-ujbY4Q1F-yI-10pA,1541
+bigdl_core_npu-2.6.0b20241212.dist-info/WHEEL,sha256=tcd-HDpskugT8GYYKyyid0lOlzoZtZdWwcrj5ormtfo,101
+bigdl_core_npu-2.6.0b20241212.dist-info/top_level.txt,sha256=iMQZlTsFPJjlD-Y0MqZEP_9ifI0LlbNCJIOTaMoGMjk,46
+bigdl_core_npu-2.6.0b20241212.dist-info/RECORD,,

intel_npu_acceleration_library/_version.py

@@ -3,4 +3,4 @@
 # SPDX-License-Identifier: Apache 2.0
 #
 
-__version__ = "2.6.0b20241207"
+__version__ = "2.6.0b20241212"