User equipment capability number defined in machine learning limit
Certain aspects of the present disclosure provide techniques for a method for wireless communications by a user equipment (UE), comprising generating a message that includes a first parameter indicating a capability of the UE to support machine learning (ML) functions across a group of cells and transmitting the message to a network entity.
1 . A method for wireless communications by a user equipment (UE), comprising:
generating a message, wherein:
when at least one condition is met, the message includes a first parameter indicating a first machine learning (ML) capability of the UE to support ML functions across a group of cells, wherein the first ML capability comprises a number of cells for which the UE is able to support the ML functions; and
the message further includes a second parameter indicating a second ML capability of the UE to support ML functions within a single cell;
transmitting the message to a network entity; and
dropping one or more ML models to ensure that, after the first ML capability of the UE is split across cells in the group of cells, ML computational complexity or memory usage across all the cells of the group of cells does not exceed the first ML capability of the UE, as indicated by the first parameter or that the ML computational complexity or the memory usage in any one cell in the group of cells does not exceed the second ML capability of the UE, as indicated by the second parameter.
2 . The method of claim 1 , wherein the group of cells comprise at least one of:
cells configured for the UE; or
activated cells among cells configured for the UE.
3 . The method of claim 1 , wherein cells within the group of cells are associated with different numerologies.
4 . The method of claim 1 , wherein the at least one condition is that a maximum supported number of cells in the group of cells is less than or equal to a threshold value.
5 . The method of claim 1 , further comprising receiving, from the network entity, an indication of an actual ML capability per cell group.
6 . The method of claim 5 , wherein the actual ML capability per cell group indicates a number of cells for which the UE is expected to support the ML functions.
7 . The method of claim 1 , wherein, if the ML computational complexity or the memory usage across a total number of cells within the group of cells exceeds the first ML capability of the UE across all the cells in the group of cells, as indicated by the first parameter, the first ML capability of the UE is split across all the cells in the group of cells.
8 . The method of claim 1 , wherein the first ML capability of the UE is split across cells in the group of cells, such that each cell in the group of cells is allocated a portion of the first ML capability of the UE within a range.
9 . The method of claim 1 , wherein the first ML capability of the UE is split across the cells in the group of cells, such that a portion of the first ML capability of the UE allocated to any given cell depends, at least in part, on whether that cell is a master cell or a secondary cell.
10 . The method of claim 1 , wherein the first ML capability of the UE is split across the cells in the group of cells, such that a portion of the first ML capability of the UE allocated to any given cell depends, at least in part, on a numerology of that cell.
11 . The method of claim 1 , wherein the UE is configured to allow the ML computational complexity or the memory usage in one or more cells in the group of cells to exceed the second ML capability of the UE, as indicated by the second parameter.
12 . A method for wireless communications by a network entity, comprising:
receiving, from a user equipment (UE), a message, wherein:
when at least one condition is met, the message includes a first parameter indicating a first ML capability of the UE to support ML functions across a group of cells, wherein the first ML capability comprises a number of cells for which the UE is able to support the ML functions; and
the message further includes a second parameter indicating a second ML capability of the UE to support the ML functions within a single cell; and
controlling ML computational complexity or memory usage across a total number of cells based on the first parameter, wherein one or more ML models are dropped to ensure that, after the first ML capability of the UE is split across the cells in the group of cells, the ML computational complexity or memory usage across all cells in the group of cells does not exceed the first ML capability of the UE, as indicated by the first parameter or that the ML computational complexity or the memory usage in any one cell in the group of cells does not exceed the second ML capability of the UE, as indicated by the second parameter.
13 . The method of claim 12 , wherein the group of cells comprise at least one of:
cells configured for the UE; or
activated cells among cells configured for the UE.
14 . The method of claim 12 , wherein cells within the group of cells are associated with different numerologies.
15 . The method of claim 12 , wherein the at least one condition is that a maximum supported number of cells in the group of cells is less than or equal to a threshold value.
16 . The method of claim 15 , wherein the network entity assumes a default value for the first parameter the first parameter is not included in the message.
17 . The method of claim 12 , further comprising transmitting, to the UE, an indication of an actual ML capability per cell group.
18 . The method of claim 17 , wherein the actual ML capability per cell group indicates a number of cells for which the UE is expected to support the ML functions.
19 . The method of claim 12 , wherein, if the ML computational complexity or the memory usage across the total number of cells within the group of cells exceeds the first ML capability of the UE across all the cells in the group of cells, as indicated by the first parameter, the first ML capability of the UE is split across all the cells in the group of cells.
20 . The method of claim 12 , wherein first ML capability of the UE is split across the cells in the group of cells, such that each cell in the group of cells is allocated a portion of the first ML capability of the UE within a range.
21 . The method of claim 12 , wherein the first ML capability of the UE is split across the cells in the group of cells, such that a portion of the first ML capability allocated to any given cell depends, at least in part, on whether that cell is a master cell or a secondary cell.
22 . An apparatus for wireless communications by a user equipment (UE), comprising:
one or more processors configured to execute instructions stored on one or more memories to cause the UE to:
generate a message, wherein:
when at least one condition is met, the message includes a first parameter indicating a first machine learning (ML) capability of the UE to support ML functions across a group of cells, wherein the first ML capability comprises a number of cells for which the UE is able to support the ML functions; and
the message further includes a second parameter indicating a second ML capability of the UE to support the ML functions within a single cell;
transmit the message to a network entity; and
drop one or more ML models to ensure that, after the first ML capability of the UE is split across cells in the group of cells, ML computational complexity or memory usage across all the cells in the group of cells does not exceed the first ML capability of the UE, as indicated by the first parameter or that the ML computational complexity or the memory usage in any one cell in the group of cells does not exceed the second ML capability of the UE, as indicated by the second parameter.
23 . An apparatus for wireless communications by a network entity, comprising:
one or more processors configured to execute instructions stored on one or more memories to cause the network entity to:
receive, from a user equipment (UE), a message, wherein:
when at least one condition is met, the message includes a first parameter indicating a first machine learning (ML) capability of the UE to support ML functions across a group of cells, wherein the first ML capability comprises a number of cells for which the UE is able to support the ML functions; and
the message further includes a second parameter indicating a second ML capability of the UE to support ML functions within a single cell;
control ML computational complexity or memory usage across a total number of cells based on the first parameter, wherein one or more ML models are dropped to ensure that, after the first ML capability of the UE is split across cells in the group of cells, the ML computational complexity or the memory usage across all the cells in the group of cells does not exceed the first ML capability of the UE, as indicated by the first parameter or that the ML computational complexity or the memory usage in any one cell in the group of cells does not exceed the second ML capability of the UE, as indicated by the second parameter.