Abstract: This disclosure proposes a data caching method and a host system. The host system includes a rewritable non-volatile memory module which includes multiple channels. The data caching method includes: obtaining a current token during an inference phase; reading a previous key matrix and a previous value matrix from the rewritable non-volatile memory module; generating a current key matrix according to the current token and the previous key matrix; generating a current value matrix according to the current token and the previous value matrix; writing multiple portions of the current key matrix into different channels; and writing multiple portions of the current value matrix into different channels.

Abstract: A memory management method, including: acquiring multiple architecture parameters of an artificial intelligence model; obtaining a memory allocation size used in a training process of the artificial intelligence model according to the multiple architecture parameters; configuring a target memory region in random access memory according to the memory allocation size, wherein train processing stage in the training process of the artificial intelligence model shares the target memory region; and temporarily storing intermediate data corresponding to each train processing stage of the training process of the artificial intelligence model in the target memory region, wherein the intermediate data is transferred between accelerator processor module, the random access memory, and a storage device. Thereby, the present invention can automatically calculate the optimized memory allocation size, and through the design of shared memory space, effectively improve memory fragmentation issues.

Abstract: A training method for a machine learning model and a host system are provided. The host system includes a rewritable non-volatile memory module. The training method includes: executing a training process of the machine learning model, which includes, in an iteration at an epoch of the training process, storing transient data and backtracking data generated by the iteration in the rewritable non-volatile memory module; and in response to an abnormality occurring in the host system which causes an interruption in the iteration, reading the transient data and the backtracking data from the rewritable non-volatile memory module, determining a stage of the iteration based on the backtracking data, and resuming the stage according to the transient data.

Abstract: A data storage method, a host system, and a data storage system are disclosed. The method includes the following. An artificial intelligence (AI) model is executed. First data to be stored to a memory storage device is obtained. In response to the first data being generated by the AI model, second data is generated according to the first data, in which the second data includes the first data, and a data amount of the second data is greater than a data amount of the first data. A first write command is sent to the memory storage device according to the second data, so as to instruct the memory storage device to store the second data.

Abstract: A read voltage calibration method, a memory storage device, a memory control circuit unit are provided, including: reading, according to a first read command, a first physical unit based on a first read voltage level to obtain first data, and the first read voltage level is a default read voltage level corresponding to the first physical unit or a first voltage difference exists between the first read voltage level and the default read voltage level; decoding the first data to obtain first error bit information; reading, according to a second read command, the first physical unit based on a second read voltage level to obtain second data, and a second voltage difference exists between the second read voltage level and the default read voltage level; decoding the second data to obtain second error bit information; calibrating the default read voltage level according to the first and second error bit information.

Abstract: A memory operation method, a memory storage device, and a memory control circuit unit are disclosed. The memory operation includes following steps. First data is received from a host system. The first data is stored into a first physical unit which is mapped to a first logical unit. In a first operation mode, a target calculation is performed based on the first data and second data stored in a second physical unit to obtain third data, and the third data is different from the first data. The third data is stored into a third physical unit which is also mapped to the first logical unit. The third data is transmitted to the host system.

Abstract: A memory operation method, a memory storage device, and a memory control circuit unit are disclosed. The memory operation includes following steps. First data is received from a host system. The first data is stored into a first physical unit which is mapped to a first logical unit. In a first operation mode, a target calculation is performed based on the first data and second data stored in a second physical unit to obtain third data, and the third data is different from the first data. The third data is stored into a third physical unit which is also mapped to the first logical unit. The third data is transmitted to the host system.

Abstract: A data storage method, a host system, and a data storage system are disclosed. The method includes the following. An artificial intelligence (AI) model is executed. First data to be stored to a memory storage device is obtained. In response to the first data being generated by the AI model, second data is generated according to the first data, in which the second data includes the first data, and a data amount of the second data is greater than a data amount of the first data. A first write command is sent to the memory storage device according to the second data, so as to instruct the memory storage device to store the second data.

Abstract: A read voltage calibration method, a memory storage device, a memory control circuit unit are provided, including: reading, according to a first read command, a first physical unit based on a first read voltage level to obtain first data, and the first read voltage level is a default read voltage level corresponding to the first physical unit or a first voltage difference exists between the first read voltage level and the default read voltage level; decoding the first data to obtain first error bit information; reading, according to a second read command, the first physical unit based on a second read voltage level to obtain second data, and a second voltage difference exists between the second read voltage level and the default read voltage level; decoding the second data to obtain second error bit information; calibrating the default read voltage level according to the first and second error bit information.

Abstract: A read voltage calibration method, a memory storage device, and a memory control circuit unit are provided. The read voltage calibration method includes: reading data from a first physical unit by using multiple read voltage levels; decoding the data to obtain multiple error evaluation parameters; determining a first vector distance parameter according to a first error evaluation parameter; determining multiple candidate read voltage levels according to the first vector distance parameter and a first read voltage level; determining a target read voltage level according to one of the candidate read voltage levels; and reading the data again from the first physical unit by using the target read voltage level.

Abstract: A read voltage calibration method, a memory storage device, and a memory control circuit unit are provided. The read voltage calibration method includes: reading data from a first physical unit by using multiple read voltage levels; decoding the data to obtain multiple error evaluation parameters; determining a first vector distance parameter according to a first error evaluation parameter; determining multiple candidate read voltage levels according to the first vector distance parameter and a first read voltage level; determining a target read voltage level according to one of the candidate read voltage levels; and reading the data again from the first physical unit by using the target read voltage level.

Abstract: A memory control method, a memory storage device and a memory control circuit unit are provided. The method includes: determining a retry threshold value according to decoding history information which includes information related to at least one first decoding operation previously performed; and determining whether to enter a second decoding mode according to the retry threshold value after at least one second decoding operation performed based on a first decoding mode is failed. A decoding ability of the second decoding mode is higher than a decoding ability of the first decoding mode.

Abstract: A method for training an asymmetric generative adversarial network to generate an image and an electronic apparatus using the same are provided. The method includes the following. A first real image belonging to a first category, a second real image belonging to a second category and a third real image belonging to a third category are input to an asymmetric generative adversarial network for training the asymmetric generative adversarial network, and the asymmetric generative adversarial network includes a first generator, a second generator, a first discriminator and a second discriminator. A fourth real image belonging to the second category is input to the first generator in the trained asymmetric generative adversarial network to generate a defect image.

Abstract: A memory control method, a memory storage device and a memory control circuit unit are provided. The method includes: determining a retry threshold value according to decoding history information which includes information related to at least one first decoding operation previously performed; and determining whether to enter a second decoding mode according to the retry threshold value after at least one second decoding operation performed based on a first decoding mode is failed. A decoding ability of the second decoding mode is higher than a decoding ability of the first decoding mode.

Abstract: A memory control method for a rewritable non-volatile memory module is provided according to an exemplary embodiment of the disclosure. The method includes: reading a first physical unit based on a first read voltage level to obtain first data; reading the first physical unit based on a second read voltage level to obtain second data; reading the first physical unit based on a third read voltage level to obtain third data; obtaining a first reference value which reflects a data variation status between the first data and the second data; obtaining a second reference value which reflects a data variation status between the first data and the third data; reading the first physical unit based on a fourth read voltage level to obtain fourth data according to the first reference value and the second reference value; and decoding the fourth data by a decoding circuit.

Abstract: A voltage identifying method, a memory controlling circuit unit, and a memory storage device are provided. The method includes: reading a plurality of first memory cells according to a first read voltage group in a plurality of read voltage groups and performing a first decoding operation to generate first verification information; identifying a plurality of second read voltage groups in the plurality of read voltage groups corresponding to a first interval in a plurality of intervals according to the first interval in which the first verification information is located; and reading the plurality of first memory cells by using a third read voltage group in the plurality of second read voltage groups and performing the first decoding operation.

Abstract: A voltage identifying method, a memory controlling circuit unit, and a memory storage device are provided. The method includes: reading a plurality of first memory cells according to a first read voltage group in a plurality of read voltage groups and performing a first decoding operation to generate first verification information; identifying a plurality of second read voltage groups in the plurality of read voltage groups corresponding to a first interval in a plurality of intervals according to the first interval in which the first verification information is located; and reading the plurality of first memory cells by using a third read voltage group in the plurality of second read voltage groups and performing the first decoding operation.

Abstract: A memory management method for a memory storage device including a rewritable non-volatile memory module is provided according to an exemplary embodiment of the disclosure. The method includes: programming first data into a plurality of first memory cells in the rewritable non-volatile memory module, such that the programmed first memory cells have a plurality of states; sending a first single-stage read command sequence which indicates to read the programmed first memory cells by using a first read voltage level; obtaining first count information corresponding to the first read voltage level according to a read result corresponding to the first single-stage read command sequence; and adjusting the first read voltage level according to the first count information and default count information corresponding to the first read voltage level.

Abstract: A decoding method, a memory controlling circuit unit, and a memory storage device are provided. The method includes: reading a first physical programming unit by using a first read voltage to obtain first data; determining whether a first ratio of a first quantity of a first bit value and a second quantity of a second bit value in the first data is greater than a threshold; when the first ratio is not greater than the threshold, performing a decoding operation according to the first data to generate first decoded data and outputting the first decoded data; and when the first ratio is greater than the threshold, not performing the decoding operation according to the first data.

Abstract: A memory control method for a rewritable non-volatile memory module is provided according to an exemplary embodiment of the disclosure. The method includes: reading first data from a first memory cell of the rewritable non-volatile memory module by a first read voltage level; decoding the first data by a decoding circuit; reading second data from the first memory cell by a second read voltage level; obtaining reliability information according to a first data status of the first data and a second data status of the second data, and the first data status and the second data status reflect that a first bit value of the first data is different from a second bit value of the second data; and decoding the second data by the decoding circuit according to the reliability information.