LLamaSharp/LLama/LLamaInstructExecutor.cs

278 lines
10 KiB
C#

using LLama.Abstractions;
using LLama.Common;
using LLama.Native;
using System;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Text.Json;
using System.Text.Json.Serialization;
using System.Threading.Tasks;
using LLama.Extensions;
using Microsoft.Extensions.Logging;
namespace LLama
{
using llama_token = Int32;
/// <summary>
/// The LLama executor for instruct mode.
/// </summary>
public class InstructExecutor
: StatefulExecutorBase
{
private bool _is_prompt_run = true;
private readonly string _instructionPrefix;
private llama_token[] _inp_pfx;
private llama_token[] _inp_sfx;
/// <summary>
///
/// </summary>
/// <param name="context"></param>
/// <param name="instructionPrefix"></param>
/// <param name="instructionSuffix"></param>
/// <param name="logger"></param>
public InstructExecutor(LLamaContext context,
string instructionPrefix = "\n\n### Instruction:\n\n",
string instructionSuffix = "\n\n### Response:\n\n",
ILogger? logger = null)
: base(context, logger)
{
_inp_pfx = Context.Tokenize(instructionPrefix, true);
_inp_sfx = Context.Tokenize(instructionSuffix, false);
_instructionPrefix = instructionPrefix;
}
/// <inheritdoc />
public override ExecutorBaseState GetStateData()
{
InstructExecutorState state = new()
{
ConsumedSessionCount = _n_session_consumed,
EmbedInps = _embed_inps,
IsPromptRun = _is_prompt_run,
ConsumedTokensCount = _consumedTokensCount,
Embeds = _embeds,
LastTokens = _last_n_tokens.ToArray(),
InputPrefixTokens = _inp_pfx,
InputSuffixTokens = _inp_sfx,
MatchingSessionTokensCount = _n_matching_session_tokens,
PastTokensCount = _pastTokensCount,
SessionFilePath = _pathSession,
SessionTokens = _session_tokens,
LastTokensCapacity = _last_n_tokens.Capacity,
MirostatMu = MirostatMu
};
return state;
}
/// <inheritdoc />
public override Task LoadState(ExecutorBaseState data)
{
if(data is InstructExecutorState state)
{
_n_session_consumed = state.ConsumedSessionCount;
_embed_inps = state.EmbedInps;
_is_prompt_run = state.IsPromptRun;
_consumedTokensCount = state.ConsumedTokensCount;
_embeds = state.Embeds;
_last_n_tokens = new FixedSizeQueue<llama_token>(state.LastTokensCapacity, state.LastTokens);
_inp_pfx = state.InputPrefixTokens;
_inp_sfx = state.InputSuffixTokens;
_n_matching_session_tokens = state.MatchingSessionTokensCount;
_pastTokensCount = state.PastTokensCount;
_pathSession = state.SessionFilePath;
_session_tokens = state.SessionTokens;
}
else
{
throw new ArgumentException("Invalid state data type.");
}
return Task.CompletedTask;
}
/// <inheritdoc />
public override async Task SaveState(string filename)
{
var state = (InstructExecutorState)GetStateData();
using (var fs = new FileStream(filename, FileMode.Create, FileAccess.Write))
{
await JsonSerializer.SerializeAsync(fs, state);
}
}
/// <inheritdoc />
public override async Task LoadState(string filename)
{
using (var fs = new FileStream(filename, FileMode.Open, FileAccess.Read))
{
var state = await JsonSerializer.DeserializeAsync<InstructExecutorState>(fs);
await LoadState(state);
}
}
/// <inheritdoc />
protected override Task<bool> GetLoopCondition(InferStateArgs args)
{
return Task.FromResult(args.RemainedTokens != 0 || _is_prompt_run);
}
/// <inheritdoc />
protected override Task PreprocessInputs(string text, InferStateArgs args)
{
args.Antiprompts ??= new List<string>();
args.Antiprompts.Add(_instructionPrefix);
if (_is_prompt_run)
{
// When running the first input (prompt) in inteactive mode, we should specially process it.
_embed_inps = Context.Tokenize(text, true).ToList();
}
else
{
if (!text.EndsWith("\n"))
{
text += "\n";
}
_consumedTokensCount = _embed_inps.Count;
_embed_inps.AddRange(_inp_pfx);
var line_inp = Context.Tokenize(text, false);
_embed_inps.AddRange(line_inp);
_embed_inps.AddRange(_inp_sfx);
args.RemainedTokens -= line_inp.Length;
}
return Task.CompletedTask;
}
/// <inheritdoc />
protected override async Task<(bool, IReadOnlyList<string>)> PostProcess(IInferenceParams inferenceParams, InferStateArgs args)
{
if (_embed_inps.Count <= _consumedTokensCount)
{
if (_last_n_tokens.TokensEndsWithAnyString(args.Antiprompts, Context.NativeHandle.ModelHandle, Context.Encoding))
{
args.WaitForInput = true;
return (true, Array.Empty<string>());
}
if (_pastTokensCount > 0 && args.WaitForInput)
{
return (true, new[] { "\n> " });
}
}
if (_embeds.Count > 0 && _embeds.Last() == NativeApi.llama_token_eos(Context.NativeHandle.ModelHandle))
{
args.WaitForInput = true;
}
if (args.RemainedTokens <= 0 && inferenceParams.MaxTokens != -1)
{
args.RemainedTokens = inferenceParams.MaxTokens;
args.WaitForInput = true;
}
return (false, Array.Empty<string>());
}
/// <inheritdoc />
protected override Task InferInternal(IInferenceParams inferenceParams, InferStateArgs args)
{
if (_embeds.Count > 0)
{
_is_prompt_run = false;
if (_pastTokensCount + _embeds.Count > Context.ContextSize)
{
HandleRunOutOfContext(inferenceParams.TokensKeep);
}
TryReuseMathingPrefix();
_pastTokensCount = Context.Eval(_embeds, _pastTokensCount);
if (_embeds.Count > 0 && !string.IsNullOrEmpty(_pathSession))
{
_session_tokens.AddRange(_embeds);
_n_session_consumed = _session_tokens.Count;
}
}
_embeds.Clear();
if (_embed_inps.Count <= _consumedTokensCount && !args.WaitForInput)
{
var repeat_last_n = inferenceParams.RepeatLastTokensCount < 0 ? Context.ContextSize : inferenceParams.RepeatLastTokensCount;
// optionally save the session on first sample (for faster prompt loading next time)
if (!string.IsNullOrEmpty(_pathSession) && args.NeedToSaveSession)
{
args.NeedToSaveSession = false;
SaveSessionFile(_pathSession);
}
llama_token id;
if (inferenceParams.SamplingPipeline is not null)
{
id = inferenceParams.SamplingPipeline.Sample(Context.NativeHandle, Context.NativeHandle.GetLogits(), _last_n_tokens.ToArray());
}
else
{
var tokenDataArray = Context.ApplyPenalty(_last_n_tokens, inferenceParams.LogitBias, repeat_last_n,
inferenceParams.RepeatPenalty, inferenceParams.FrequencyPenalty, inferenceParams.PresencePenalty, inferenceParams.PenalizeNL);
var mu = MirostatMu;
id = Context.Sample(
tokenDataArray, ref mu, inferenceParams.Temperature, inferenceParams.Mirostat, inferenceParams.MirostatTau,
inferenceParams.MirostatEta, inferenceParams.TopK, inferenceParams.TopP, inferenceParams.TfsZ, inferenceParams.TypicalP, inferenceParams.Grammar,
inferenceParams.MinP
);
MirostatMu = mu;
}
_last_n_tokens.Enqueue(id);
_embeds.Add(id);
args.RemainedTokens--;
args.ReturnValue = true;
}
else
{
while (_embed_inps.Count > _consumedTokensCount)
{
_embeds.Add(_embed_inps[_consumedTokensCount]);
_last_n_tokens.Enqueue(_embed_inps[_consumedTokensCount]);
_consumedTokensCount++;
if (_embeds.Count >= Context.Params.BatchSize)
{
break;
}
}
}
return Task.CompletedTask;
}
/// <summary>
/// The desciptor of the state of the instruct executor.
/// </summary>
public class InstructExecutorState : ExecutorBaseState
{
/// <summary>
/// Whether the executor is running for the first time (running the prompt).
/// </summary>
[JsonPropertyName("is_prompt_run")]
public bool IsPromptRun { get; set; }
/// <summary>
/// Instruction prefix tokens.
/// </summary>
[JsonPropertyName("inp_pfx")]
public llama_token[] InputPrefixTokens { get; set; }
/// <summary>
/// Instruction suffix tokens.
/// </summary>
[JsonPropertyName("inp_sfx")]
public llama_token[] InputSuffixTokens { get; set; }
}
}
}