Intermediary storage device, computer system and computer command preprocessing method

Abstract

The invention discloses an intermediary storage device, a computer system and a preprocessing method of computer commands. The intermediary storage device comprises a model storage unit, a data storage unit, an AI selection unit and an access control unit. The model storage unit stores multiple groups of language models; the data storage unit stores file data; the AI selection unit selects a language model according to the model operation command, and the language model executes the model operation command and generates a generation result; the access control unit is connected with the data storage unit and the model storage unit, judges whether the input command is a model operation command or a data access command, and forwards the model operation command to the AI selection unit if the input command is the model operation command; the selected language model generates output data and sends the output data to the access control unit.

Description

TECHNICAL FIELD

[0001] The present application relates to a storage device, a computer and a processing method of commands thereof, in particular to an intermediary storage device, a computer system and a preprocessing method of computer commands. BACKGROUND

[0002] The rise of artificial intelligence (AI) has brought many new possibilities to human society. Among them, the development of large language models (LLMs) has laid the foundation for the breakthrough of AI technology. LLM is an AI model that can process and generate natural language, and its training requires a large amount of data and computing resources. Therefore, the storage problem of LLM has become an important challenge for the development of AI technology.

[0003] The current way is to transmit the interaction information between LLM and user through the Internet. Although the transmission process of information can be processed by encryption, there is still a risk of leakage for personal information or business secrets. SUMMARY

[0004] Therefore, in an embodiment, the intermediary storage device includes a model storage unit, a data storage unit, an AI selection unit and an access control unit. The model storage unit stores a plurality of language models; the data storage unit stores a plurality of file data; the AI selection unit selects a language model according to a model operation command, so that the selected language model executes the model operation command and generates output data; the access control unit is connected to the data storage unit and the model storage unit, and receives an input command. The access control unit determines whether the input command is a model operation command or a data access command. If the input command is a model operation command, the access control unit forwards the model operation command to the AI selection unit, so that the selected language model generates output data, and the access control unit generates a generation result according to the output data. If the input command is a data access command, the access control unit accesses the corresponding file data from the data storage unit according to the data access command. The intermediary storage device can directly call the local language model and select the corresponding language model according to different application scenarios.

[0005] In an embodiment, the AI selection unit further includes an instant job system, which transmits the output data to the access control unit.

[0006] In an embodiment, the access control unit sends the output data or the selected file data to the upper job system.

[0007] In an embodiment, the access control unit is connected to a transmission interface, and the transmission interface is an advanced technology attachment (ATA) interface, a serial ATA (SATA) interface, a universal serial bus (USB) interface, a peripheral component interconnect express (PCIe) interface, or a non-volatile memory express (NVMe) interface.

[0008] In an embodiment, an intermediary storage device includes a plurality of model storage units, a data storage unit, an AI selection unit, and an access control unit. Each model storage unit stores a language model; the data storage unit stores a plurality of sets of file data; the AI selection unit is connected to the model storage units, and the AI selection unit selects a language model according to an operation command, the language model executes the model operation command and generates output data; the access control unit is connected to the data storage unit and the AI selection unit, and the access control unit accepts an input command, the access control unit determines whether the input command is a model operation command or a data access command, if the input command is a model operation command, the access control unit forwards the model operation command to the AI selection unit, so that the selected language model generates output data, and the access control unit generates a result according to the output data, if the input command is a data access command, the access control unit accesses corresponding file data from the data storage unit according to the data access command.

[0009] In an embodiment, the AI selection unit further includes a real-time operating system, and the real-time operating system transmits the output data to the access control unit.

[0010] In an embodiment, the access control unit transmits the output data or the selected file data to an upper-layer operating system.

[0011] In an embodiment, the access control unit is connected to a transmission interface, and the transmission interface is an advanced technology attachment (ATA) interface, a serial ATA (SATA) interface, a universal serial bus (USB) interface, a peripheral component interconnect express (PCIe) interface, or a non-volatile memory express (NVMe) interface.

[0012] In an embodiment, a computer includes a processor and an intermediary storage device. The processor executes an upper-layer operating system, and the processor receives an input command through the upper-layer operating system; the intermediary storage device is connected to the processor, and the intermediary storage device has at least one model storage unit, a data storage unit, an AI selection unit, and an access control unit, the access control unit is connected to the data storage unit, the AI selection unit, and each model storage unit, each model storage unit stores a language model, and the data storage unit stores a plurality of sets of file data; wherein the upper-layer operating system transmits the input command to the access control unit, the access control unit determines whether the input command is a model operation command or a data access command, if the input command is a model operation command, the access control unit forwards the model operation command to the AI selection unit, so that the selected language model generates output data, if the input command is a data access command, the access control unit accesses corresponding file data from the data storage unit according to the data access command.

[0013] A pre-processing method of computer command, comprising receiving an input command by an access control unit of a mediation storage device; determining by the access control unit whether the input command is a model operation command or a data access command; if the input command is a model operation command, sending the model operation command to an AI selection unit; selecting by the AI selection unit one of a plurality of language models according to the model operation command, the selected language model being a selected model; generating output data by the selected model according to the model operation command; and generating a generation result by the access control unit according to the output data.

[0014] In an embodiment, the step of selecting by the AI selection unit one of a plurality of language models according to the model operation command, the selected language model being a selected model, comprises selecting by the AI selection unit one of a plurality of model storage units, each model storage unit storing a corresponding language model.

[0015] In an embodiment, the step of selecting by the AI selection unit one of a plurality of language models according to the model operation command, the selected language model being a selected model, comprises storing a language model in a model storage unit; and selecting by the AI selection unit the language model in the model storage unit.

[0016] In an embodiment, if the input command is a data access command, accessing by the access control unit corresponding file data from a data storage unit according to the data access command.

[0017] The mediation storage device, computer system and pre-processing method of computer command provide large language models that can be used by a local computer. In addition to avoiding the risk of information leakage during network communication, the calling speed of large language models can be accelerated, thereby speeding up the response time of information. And different application scenarios of large language models can be deployed in the computer, so that related services of multiple different application scenarios can be provided in the same computer. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A hardware architecture diagram of the mediation storage device of an embodiment.

[0019] Figure 2A A system architecture diagram applied to a computer of an embodiment.

[0020] Figure 2B A diagram of loading a selected model into a data storage unit of an embodiment.

[0021] Figure 3 A flowchart of a pre-processing method of computer command of an embodiment.

[0022] Figure 4 An element diagram of an AI selection unit of an embodiment.

[0023] Figure 5 A system architecture diagram applied to a computer for another embodiment.

[0024] In which, the reference signs are explained as follows:

[0025] 10: computer

[0026] 100: processor

[0027] 200: input device

[0028] 210: input command

[0029] 300: intermediate storage device

[0030] 310: model storage unit

[0031] 311: language model

[0032] 312: selected model

[0033] 320: data storage unit

[0034] 321: file data

[0035] 322: upper-layer operating system

[0036] 330: AI selection unit

[0037] 331: model operation command

[0038] 332: data access command

[0039] 333: real-time operating system

[0040] 334: output data

[0041] 335: generation result

[0042] 340: access control unit

[0043] S310, S320, S330, S340, S350, S360: steps DETAILED DESCRIPTION

[0044] Please refer to Figure 1 and Figure 2AFig. 1 shows a hardware architecture diagram of an embodiment of the intermediary storage device 300 and a system architecture diagram of the intermediary storage device 300 applied to a computer 10. The computer 10 includes a processor 100 and the intermediary storage device 300. The intermediary storage device 300 is connected to the processor 100 and receives an input command 210 sent by the processor 100. The computer 10 can receive the input command 210 through an input device 200 or the processor 100 can send the input command 210 to the intermediary storage device 300 when the processor 100 executes an application program. The computer 10 can be an electronic device such as a personal computer, a notebook computer, a tablet computer, a mobile phone, etc. The computer 10 executes a corresponding upper-layer operating system 322 (OS). The upper-layer operating system 322 can be a Windows operating system of the Microsoft Corporation, an OS X operating system of the Apple Corporation, a Linux operating system, etc.

[0045] The intermediary storage device 300 can be applied to a Serial Advanced Technology Attachment (SATA), a Serial Attached SCSI (SAS), a Peripheral Component Interconnect Express (PCIe), a Non-Volatile Memory Express (NVMe), a Universal Serial Bus (USB), an external Serial Advanced Technology Attachment (eSATA), a Small Computer System Interface (SCSI), an Integrated Drive Electronics (IDE), a Next Generation Form Factor (M.2), a Universal Serial Bus (USB), or a Thunderbolt.

[0046] The intermediary storage device 300 comprises a model storage unit 310, a data storage unit 320, an artificial intelligence selection unit (hereinafter referred to as AI selection unit 330) and an access control unit 340. The access control unit 340 is connected to the model storage unit 310, the data storage unit 320 and the AI selection unit 330. The model storage unit 310 can be in a non-volatile memory (NVM) or other storage medium. The model storage unit 310 stores a plurality of sets of language models 311 (Large Language Model, LLM for short). The types of language models 311 can be, but are not limited to, language models 311 for dialogue applications (Language Model for Dialogue Applications, LaMDA for short), Meta large language models 311 (Large Language Model Meta AI, LLaMA for short), GPT-3 (Generative Pre-trained Transformer 3, GPT-3 for short), BERT (Bidirectional Encoder Representations from Transformers, BERT for short) of Google, GPT-4 (Generative Pre-trained Transformer 4, GPT-4 for short), Codex (OpenAI Codex), DALL·E (OpenAIDALL·E), BART (Bidirectional and Auto-Regressive Transformers, BART for short), RoBERTa (Robustly optimized BERT approach, RoBERTa for short), T5 (Text-to-Text Transfer Transformer, T5 for short) or PaLM (Pathways Language Model, PaLM for short).

[0047] On the basis of the foregoing language models 311, the model storage unit 310 can deploy language models 311 with different requirements according to different scene requirements. For example, LLaMA can be applied to the scene of English teaching, and DALL·E can be applied to the scene of drawing teaching. The foregoing various language models 311 can also be applied in the fields of personal chat scene, personal knowledge base management, file arrangement, file translation, medical care, financial investment, manufacturing, retail, transportation or agriculture, etc.

[0048] The data storage unit 320 stores a plurality of sets of file data 321, which refers to text files, image files, video files, audio files, database files, compressed files, presentation files, or program code files, etc. In some embodiments, an upper layer operating system 322 can be stored in the intermediary storage device 300 or in another storage device.

[0049] The upper layer operating system 322 causes the processor 100 to send an input command 210 to the intermediary storage device 300 according to an input command 210 issued by a user or generated by an application program. The input command 210 can be further classified as a data access command 332 or a model operation command 331. After receiving the input command 210, the intermediary storage device 300 determines whether the input command 210 is a model operation command 331 or a data access command 332.

[0050] The data access command 332 can include reading, writing, appending, updating, deleting, moving, copying, querying, locking, etc. related processing of the file data 321. The model operation command 331 includes a chat interaction instruction (AI Language Model Chat, AI LLM CHAT), a model configuration instruction (AI Language Model Configuration Model, AI LLM CONFIG MODEL), a model fine-tuning instruction (AI Language Model Fine Tune, AI LLM FINE TUNE), a language model 311 authorization instruction (AI Language Model Authorization, AI LLM AUTH), a model deployment instruction (AI Language Model Deploy Model, AI LLM DEPLOY MODEL), a dataset loading instruction (AI Language Model Load Dataset, AI LLM LOAD DATASET), a text generation instruction (AI Language Model Generate Text, AI LLM GENERATE TEXT), a performance evaluation instruction (AI Language Model Evaluate Performance, AI LLM EVALUATE PERFORMANCE), a parameter updating instruction (AI Language Model Update Parameters, AI LLM UPDATE PARAMETERS), a model saving instruction (AI Language Model Save Model, AI LLM SAVE MODEL), a model testing instruction (AI Language Model Test Model, AI LLM TEST MODEL), a model resetting instruction (AI Language Model Reset Model, AI LLM RESET MODEL), a model training instruction (AI Language Model Train Model, AI LLM TRAIN MODEL), or a permission setting instruction (AI Language Model Set Permissions, AI LLM SET PERMISSIONS), etc.

[0051] After the access control unit 340 receives the input command 210, the access control unit 340 performs the following steps according to the type of the input command 210. Please refer to Figure 3As shown:

[0052] Step S310: receiving the input command by the access control unit of the intermediary storage device;

[0053] Step S320: judging by the access control unit whether the input command is a model operation command or a data access command;

[0054] Step S330: if the input command is a model operation command, sending the model operation command to the AI selection unit by the access control unit;

[0055] Step S340: selecting one of the plurality of language models according to the model operation command by the AI selection unit, the selected language model being a selected model;

[0056] Step S350: generating output data according to the model operation command by the selected model;

[0057] Step S360: generating a result according to the output data by the access control unit; and

[0058] Step S370: if the input command is a data access command, sending the data access command to the data storage unit by the access control unit.

[0059] First, the computer 10 is enabled to boot, so that the computer 10 loads the upper-layer operating system 322. The upper-layer operating system 322 receives the input command 210. Generally, the user can issue the input command 210 through an input interface, for example, the user enters the input command 210 through a keyboard, or selects relevant options through a mouse to constitute the input command 210.

[0060] After the upper-layer operating system 322 receives the input command 210, the upper-layer operating system 322 sends the input command 210 to the corresponding intermediary storage device 300 (corresponding to step S310). The number of the intermediary storage devices 300 is at least one, and in the embodiment, one intermediary storage device 300 is taken as an example for illustration. Figure 1 The access control unit 340 of the intermediary storage device 300 judges whether the input command 210 is a data access command 332 or a model operation command 331 (corresponding to step S320). If the access control unit 340 judges that the input command 210 is a model operation command 331, the access control unit 340 sends the model operation command 331 to the AI selection unit 330 (corresponding to step S330).

[0061] Next, the AI ​​selection unit 330 selects from multiple sets of language models 311 in the model storage unit 310, and the selected language model 311 is referred to as the selected model 312 (corresponding to step S340). In this embodiment, the model storage unit 310 has multiple sets of language models 311. Before selecting a language model 311, the AI ​​selection unit 330 determines the corresponding application scenario based on the content of the model operation command 331. The AI ​​selection unit 330 then selects the corresponding language model 311 based on the selected application scenario. Figure 2A In the example, the selected model 312 is illustrated using a thick black outline.

[0062] In one embodiment, the access control unit 340 can load the selected model 312 into the data storage unit 320, such as... Figure 2B As shown. The selected model 312 can execute model operation command 331 and generate output data 334. The data storage unit 320 sends the output data 334 to the access control unit 340, causing the access control unit 340 to generate a result 335. Generally, the access control unit 340 encapsulates the output data 334 into an API (Application Programming Interface) format corresponding to the upper-level application or operating system. The API interface format can be as follows:

[0063] {"prompt":"content","max_tokens":60,"temperature":0.5}

[0064] The AI selection unit 330 sends the model operation command 331 to the selected model 312. The selected model 312 generates the corresponding output data 334 according to the model operation command 331 (corresponding to step S350). For example, if the model operation command 331 is "Translate the following English paragraph into Traditional Chinese", the AI selection unit 330 can select the LLaMA model in the previous example as the language model 311 for translation and teaching. The LLaMA model executes the model operation command 331 and generates the translated output data 334 "Translate the following English paragraph into Traditional Chinese". The access control unit 340 sends the generation result 335 to the upper-layer operating system 322. In addition, the AI selection unit 330 can also call the relevant language model 311 according to the investment-related input command 210, such as GPT-3, GPT-4, or LaMDA. If the input command 210 is a data access command 332, the access control unit 340 sends the data access command 332 to the data storage unit 320 (corresponding to step S370). The access control unit 340 accesses the corresponding file data 321 from the data storage unit 320. The intermediary storage device 300 sends the selected file information to the upper-layer operating system 322.

[0065] In some embodiments, the AI selection unit 330 further includes a real-time operating system (RTOS), please refer to Figure 4 . The real-time operating system 333 provides task management, resource management, communication, or information security processing of the intermediary storage device 300. The real-time operating system 333 transmits the output data 334 to the access control unit 340, so that the access control unit 340 generates the generation result 335.

[0066] The intermediary storage device 300 further includes a transmission interface, please refer to Figure 4 . The access control unit 340 is connected to the transmission interface, and the transmission interface is further connected to the processor 100. The types of the transmission interface are advanced technology attachment (ATA), serial AT attachment (SATA), universal serial bus (USB), peripheral component interconnect express (PCIe), or non-volatile memory express (NVMe).

[0067] In some embodiments, the computer 10 comprises a processor 100 and an intermediary storage device 300, please refer to Figure 5 The intermediary storage device 300 of this embodiment comprises a plurality of model storage units 310, a data storage unit 320, an AI selection unit 330 and an access control unit 340. The access control unit 340 is connected to the data storage unit 320 and the AI selection unit 330. Each model storage unit 310 stores a respective language model 311, and the types of language models 311 can refer to the foregoing. Alternatively, the same language model 311 used in different application scenarios is stored in a respective model storage unit 310. For example, two sets of LLaMA models are applied to English teaching and Python programming language teaching. Therefore, the English teaching LLaMA model can be stored in an independent model storage unit 310, and the Python programming language teaching LLaMA model is stored in another model storage unit 310.

[0068] The access control unit 340 determines whether the received input command 210 is a model operation command 331 or a data access command 332. If the input command 210 is a data access command 332, the access control unit 340 accesses the corresponding file data 321 from the data storage unit 320 according to the data access command 332, and responds to the selected file data 321 to the upper operating system 322. If the data access command 332 is a model operation command 331, the access control unit 340 sends the model operation command 331 to the AI selection unit 330. The AI selection unit 330 selects the corresponding model storage unit 310 and language model 311 according to the application scenario of the model operation command 331. The model operation command 331 is executed by the selected model 312 and generates output data 334. The access control unit 340 sends the generated result 335 to the upper operating system 322.

[0069] The intermediary storage device 300, the computer 10 system and the pre-processing method of the computer 10 command provide large language models 311 that can be used in the local computer 10. In addition to avoiding the risk of information leakage during network communication, it can also speed up the calling speed of the large language model 311, thereby speeding up the information response time. And the computer 10 can deploy large language models 311 of different application scenarios, so that related services of multiple different application scenarios can be provided in the same computer 10.

Claims

1. An intermediary storage device, characterized by, The application comprises: a model storage unit storing a plurality of language models; a data storage unit storing a plurality of file data; an AI selection unit selecting the language model according to a model operation command, causing the selected language model to execute the model operation command and generate an output data; and an access control unit connected to the data storage unit and the model storage unit, the access control unit receiving an input command, the access control unit determining whether the input command is the model operation command or a data access command, if the input command is the model operation command, the access control unit forwarding the model operation command to the AI selection unit, causing the selected language model to generate the output data, and the access control unit generating a generated result according to the output data, if the input command is the data access command, the access control unit accessing the corresponding file data from the data storage unit according to the data access command.

2. The intermediary storage device of claim 1, wherein, The AI selection unit further comprises a real-time operating system.

3. The intermediary storage device of claim 1, wherein, The access control unit sends the generated result or the selected file data to an upper operating system.

4. The intermediary storage device of claim 1, wherein, Further comprising a transmission interface, the access control unit is connected to the transmission interface, the transmission interface is of advanced technology interface, serial ATA interface, universal serial bus, peripheral component interconnect express interface or non-volatile memory extension.

5. An intermediary storage device, characterized by, The application comprises: a plurality of model storage units, each of the model storage units storing a language model; a data storage unit storing a plurality of file data; an AI selection unit connected to the plurality of model storage units, the AI selection unit selecting the language model according to a model operation command, causing the selected language model to execute the model operation command and generate an output data; and an access control unit connected to the data storage unit and the AI selection unit, the access control unit receiving an input command, the access control unit determining whether the input command is the model operation command or a data access command, if the input command is the model operation command, the access control unit forwarding the model operation command to the AI selection unit, causing the selected language model to generate the output data, and the access control unit generating a generated result according to the output data, if the input command is the data access command, the access control unit accessing the corresponding file data from the data storage unit according to the data access command.

6. The intermediary storage device of claim 5, wherein, The AI selection unit further comprises a real-time operating system.

7. The intermediary storage device of claim 5, wherein, The access control unit sends the generated result or the selected file data to an upper operating system.

8. The intermediary storage device of claim 5, wherein, Further comprising a transmission interface, the access control unit is connected to the transmission interface, the transmission interface is of advanced technology interface, serial ATA interface, universal serial bus, peripheral component interconnect express interface or non-volatile memory extension.

9. A computer system, characterized by The application comprises: a processor executing an upper operating system, the processor receiving an input command through the upper operating system; and ​ An intermediary storage device connected to the processor, the intermediary storage device having at least one model storage unit, a data storage unit, an AI selection unit, and an access control unit connected to the data storage unit, the AI selection unit, and each of the model storage units, each of the model storage units storing a language model, the data storage unit storing a plurality of sets of file data; Wherein the upper-layer operating system transmits the input command to the access control unit, the access control unit determines whether the input command is a model operation command or a data access command, if the input command is the model operation command, the access control unit forwards the model operation command to the AI selection unit, so that the selected language model generates an output data, and the access control unit generates a generation result according to the output data, if the input command is the data access command, the access control unit accesses the corresponding file data from the data storage unit according to the data access command.

10. A method of preprocessing computer commands, characterized by, Comprising: receiving, by an access control unit of an intermediary storage device, an input command; determining, by the access control unit, whether the input command is a model operation command or a data access command; if the input command is the model operation command, sending, by the access control unit, the model operation command to an AI selection unit; selecting, by the AI selection unit, one of a plurality of language models according to the model operation command, the selected language model being a selected model; generating, by the selected model, an output data according to the model operation command; and generating, by the access control unit, a generation result according to the output data.

11. The method of pre-processing computer commands of claim 10, wherein, The step of selecting, by the AI selection unit, one of the plurality of language models according to the model operation command, the selected language model being the selected model, comprises: selecting, by the AI selection unit, one of a plurality of model storage units, each of the model storage units storing a corresponding language model.

12. The method of pre-processing computer commands of claim 10, wherein, The step of selecting, by the AI selection unit, one of the plurality of language models according to the model operation command, the selected language model being the selected model, comprises: storing, in a model storage unit, the plurality of language models; and selecting, by the AI selection unit, one of the language models from the model storage unit.

13. The method of pre-processing computer commands of claim 10, wherein, If the input command is the data access command, accessing, by the access control unit, a corresponding file data from a data storage unit according to the data access command.

Images