Class Session
public sealed class Session : LiteRTLMNativeHandle
- Inheritance
-
objectSession
- Inherited Members
Methods
CancelProcess()
Cancels the current processing in the session.
public void CancelProcess()
GenerateContent(ReadOnlySpan<InputData>)
Generates content from the input prompt.
public Responses? GenerateContent(ReadOnlySpan<InputData> inputs)
Parameters
inputsReadOnlySpan<InputData>An array of input wrappers representing multimodal input.
Returns
- Responses
The responses wrapper, or null on failure. The caller is responsible for disposing the returned wrapper.
GenerateContentStream(ReadOnlySpan<InputData>, StreamCallback)
Generates content from the input prompt and streams the response through a callback. This is a non-blocking call and invokes the callback from a background thread for each chunk.
public int GenerateContentStream(ReadOnlySpan<InputData> inputs, StreamCallback callback)
Parameters
inputsReadOnlySpan<InputData>An array of input wrappers representing multimodal input.
callbackStreamCallbackThe callback function (StreamCallback) that receives response chunks.
Returns
- int
0 on success, non-zero on failure to start the stream.
GetBenchmarkInfo()
Retrieves benchmark information for the session. The caller is responsible for disposing the returned wrapper.
public BenchmarkInfo? GetBenchmarkInfo()
Returns
- BenchmarkInfo
The benchmark information wrapper, or null on failure.
ReleaseUnmanagedResources()
protected override void ReleaseUnmanagedResources()
RewindToCheckpoint(string)
Rewinds the session to the given checkpoint label.
public int RewindToCheckpoint(string label)
Parameters
labelstringLabel of the checkpoint to rewind to.
Returns
- int
0 on success, non-zero on failure.
RewindToStep(int)
Rewinds the session to a specific step number.
public int RewindToStep(int step)
Parameters
stepintThe step number to rewind to.
Returns
- int
0 on success, non-zero on failure.
RunDecode()
Starts decoding for the model to predict a response based on the input prompt or query added after calling RunPrefill(ReadOnlySpan<InputData>). This is a blocking call and returns when decoding is complete.
public Responses? RunDecode()
Returns
- Responses
The responses wrapper, or null on failure. The caller is responsible for disposing the returned wrapper.
RunDecodeAsync(StreamCallback)
Starts decoding for the model to predict a response based on the input prompt or query added after calling RunPrefill(ReadOnlySpan<InputData>). This is a non-blocking call that streams response chunks through a callback.
public int RunDecodeAsync(StreamCallback callback)
Parameters
callbackStreamCallbackThe callback function (StreamCallback) that receives response chunks.
Returns
- int
0 on success, non-zero on failure.
RunPrefill(ReadOnlySpan<InputData>)
Adds the input prompt or query to the model and starts the prefill process. This is a blocking call and returns when prefill is complete.
public int RunPrefill(ReadOnlySpan<InputData> inputs)
Parameters
inputsReadOnlySpan<InputData>An array of input wrappers representing multimodal input.
Returns
- int
0 on success, non-zero on failure.
RunTextScoring(string[], bool)
Scores the target text after the prefill process is complete.
public Responses? RunTextScoring(string[] targetText, bool storeTokenLengths)
Parameters
targetTextstring[]An array of target text strings to score.
storeTokenLengthsboolWhether to store the token lengths of the target texts in the responses.
Returns
- Responses
The responses wrapper, or null on failure. The caller is responsible for disposing the returned wrapper.
SaveCheckpoint(string)
Saves the current state of the session to a checkpoint with the given label.
public int SaveCheckpoint(string label)
Parameters
labelstringLabel for the checkpoint.
Returns
- int
0 on success, non-zero on failure.