You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

IContextParamsExtensions.cs 2.7 kB

1 year ago
April 2024 Binary Update (#662) * Updated binaries, using [this build](https://github.com/SciSharp/LLamaSharp/actions/runs/8654672719/job/23733195669) for llama.cpp commit `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7`. - Added all new functions. - Moved some functions (e.g. `SafeLlamaModelHandle` specific functions) into `SafeLlamaModelHandle.cs` - Exposed tokens on `SafeLlamaModelHandle` and `LLamaWeights` through a `Tokens` property. As new special tokens are added in the future they can be added here. - Changed all token properties to return nullable tokens, to handle some models not having some tokens. - Fixed `DefaultSamplingPipeline` to handle no newline token in some models. * Moved native methods to more specific locations. - Context specific things have been moved into `SafeLLamaContextHandle.cs` and made private - they're exposed through C# properties and methods already. - Checking that GPU layer count is zero if GPU offload is not supported. - Moved methods for creating default structs (`llama_model_quantize_default_params` and `llama_context_default_params`) into relevant structs. * Removed exception if `GpuLayerCount > 0` when GPU is not supported. * - Added low level wrapper methods for new per-sequence state load/save in `SafeLLamaContextHandle` - Added high level wrapper methods (save/load with `State` object or memory mapped file) in `LLamaContext` - Moved native methods for per-sequence state load/save into `SafeLLamaContextHandle` * Added update and defrag methods for KV cache in `SafeLLamaContextHandle` * Updated submodule to `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7` * Passing the sequence ID when saving a single sequence state
1 year ago
April 2024 Binary Update (#662) * Updated binaries, using [this build](https://github.com/SciSharp/LLamaSharp/actions/runs/8654672719/job/23733195669) for llama.cpp commit `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7`. - Added all new functions. - Moved some functions (e.g. `SafeLlamaModelHandle` specific functions) into `SafeLlamaModelHandle.cs` - Exposed tokens on `SafeLlamaModelHandle` and `LLamaWeights` through a `Tokens` property. As new special tokens are added in the future they can be added here. - Changed all token properties to return nullable tokens, to handle some models not having some tokens. - Fixed `DefaultSamplingPipeline` to handle no newline token in some models. * Moved native methods to more specific locations. - Context specific things have been moved into `SafeLLamaContextHandle.cs` and made private - they're exposed through C# properties and methods already. - Checking that GPU layer count is zero if GPU offload is not supported. - Moved methods for creating default structs (`llama_model_quantize_default_params` and `llama_context_default_params`) into relevant structs. * Removed exception if `GpuLayerCount > 0` when GPU is not supported. * - Added low level wrapper methods for new per-sequence state load/save in `SafeLLamaContextHandle` - Added high level wrapper methods (save/load with `State` object or memory mapped file) in `LLamaContext` - Moved native methods for per-sequence state load/save into `SafeLLamaContextHandle` * Added update and defrag methods for KV cache in `SafeLLamaContextHandle` * Updated submodule to `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7` * Passing the sequence ID when saving a single sequence state
1 year ago
April 2024 Binary Update (#662) * Updated binaries, using [this build](https://github.com/SciSharp/LLamaSharp/actions/runs/8654672719/job/23733195669) for llama.cpp commit `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7`. - Added all new functions. - Moved some functions (e.g. `SafeLlamaModelHandle` specific functions) into `SafeLlamaModelHandle.cs` - Exposed tokens on `SafeLlamaModelHandle` and `LLamaWeights` through a `Tokens` property. As new special tokens are added in the future they can be added here. - Changed all token properties to return nullable tokens, to handle some models not having some tokens. - Fixed `DefaultSamplingPipeline` to handle no newline token in some models. * Moved native methods to more specific locations. - Context specific things have been moved into `SafeLLamaContextHandle.cs` and made private - they're exposed through C# properties and methods already. - Checking that GPU layer count is zero if GPU offload is not supported. - Moved methods for creating default structs (`llama_model_quantize_default_params` and `llama_context_default_params`) into relevant structs. * Removed exception if `GpuLayerCount > 0` when GPU is not supported. * - Added low level wrapper methods for new per-sequence state load/save in `SafeLLamaContextHandle` - Added high level wrapper methods (save/load with `State` object or memory mapped file) in `LLamaContext` - Moved native methods for per-sequence state load/save into `SafeLLamaContextHandle` * Added update and defrag methods for KV cache in `SafeLLamaContextHandle` * Updated submodule to `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7` * Passing the sequence ID when saving a single sequence state
1 year ago
April 2024 Binary Update (#662) * Updated binaries, using [this build](https://github.com/SciSharp/LLamaSharp/actions/runs/8654672719/job/23733195669) for llama.cpp commit `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7`. - Added all new functions. - Moved some functions (e.g. `SafeLlamaModelHandle` specific functions) into `SafeLlamaModelHandle.cs` - Exposed tokens on `SafeLlamaModelHandle` and `LLamaWeights` through a `Tokens` property. As new special tokens are added in the future they can be added here. - Changed all token properties to return nullable tokens, to handle some models not having some tokens. - Fixed `DefaultSamplingPipeline` to handle no newline token in some models. * Moved native methods to more specific locations. - Context specific things have been moved into `SafeLLamaContextHandle.cs` and made private - they're exposed through C# properties and methods already. - Checking that GPU layer count is zero if GPU offload is not supported. - Moved methods for creating default structs (`llama_model_quantize_default_params` and `llama_context_default_params`) into relevant structs. * Removed exception if `GpuLayerCount > 0` when GPU is not supported. * - Added low level wrapper methods for new per-sequence state load/save in `SafeLLamaContextHandle` - Added high level wrapper methods (save/load with `State` object or memory mapped file) in `LLamaContext` - Moved native methods for per-sequence state load/save into `SafeLLamaContextHandle` * Added update and defrag methods for KV cache in `SafeLLamaContextHandle` * Updated submodule to `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7` * Passing the sequence ID when saving a single sequence state
1 year ago
April 2024 Binary Update (#662) * Updated binaries, using [this build](https://github.com/SciSharp/LLamaSharp/actions/runs/8654672719/job/23733195669) for llama.cpp commit `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7`. - Added all new functions. - Moved some functions (e.g. `SafeLlamaModelHandle` specific functions) into `SafeLlamaModelHandle.cs` - Exposed tokens on `SafeLlamaModelHandle` and `LLamaWeights` through a `Tokens` property. As new special tokens are added in the future they can be added here. - Changed all token properties to return nullable tokens, to handle some models not having some tokens. - Fixed `DefaultSamplingPipeline` to handle no newline token in some models. * Moved native methods to more specific locations. - Context specific things have been moved into `SafeLLamaContextHandle.cs` and made private - they're exposed through C# properties and methods already. - Checking that GPU layer count is zero if GPU offload is not supported. - Moved methods for creating default structs (`llama_model_quantize_default_params` and `llama_context_default_params`) into relevant structs. * Removed exception if `GpuLayerCount > 0` when GPU is not supported. * - Added low level wrapper methods for new per-sequence state load/save in `SafeLLamaContextHandle` - Added high level wrapper methods (save/load with `State` object or memory mapped file) in `LLamaContext` - Moved native methods for per-sequence state load/save into `SafeLLamaContextHandle` * Added update and defrag methods for KV cache in `SafeLLamaContextHandle` * Updated submodule to `f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7` * Passing the sequence ID when saving a single sequence state
1 year ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667
  1. using System;
  2. using System.IO;
  3. using LLama.Abstractions;
  4. using LLama.Native;
  5. namespace LLama.Extensions
  6. {
  7. /// <summary>
  8. /// Extension methods to the IContextParams interface
  9. /// </summary>
  10. public static class IContextParamsExtensions
  11. {
  12. /// <summary>
  13. /// Convert the given `IModelParams` into a `LLamaContextParams`
  14. /// </summary>
  15. /// <param name="params"></param>
  16. /// <param name="result"></param>
  17. /// <returns></returns>
  18. /// <exception cref="FileNotFoundException"></exception>
  19. /// <exception cref="ArgumentException"></exception>
  20. public static void ToLlamaContextParams(this IContextParams @params, out LLamaContextParams result)
  21. {
  22. result = LLamaContextParams.Default();
  23. result.n_ctx = @params.ContextSize ?? 0;
  24. result.n_batch = @params.BatchSize;
  25. result.n_ubatch = @params.UBatchSize;
  26. result.n_seq_max = @params.SeqMax;
  27. result.seed = @params.Seed;
  28. result.embeddings = @params.Embeddings;
  29. result.rope_freq_base = @params.RopeFrequencyBase ?? 0;
  30. result.rope_freq_scale = @params.RopeFrequencyScale ?? 0;
  31. // Default YaRN values copied from here: https://github.com/ggerganov/llama.cpp/blob/381efbf480959bb6d1e247a8b0c2328f22e350f8/common/common.h#L67
  32. result.yarn_ext_factor = @params.YarnExtrapolationFactor ?? -1f;
  33. result.yarn_attn_factor = @params.YarnAttentionFactor ?? 1f;
  34. result.yarn_beta_fast = @params.YarnBetaFast ?? 32f;
  35. result.yarn_beta_slow = @params.YarnBetaSlow ?? 1f;
  36. result.yarn_orig_ctx = @params.YarnOriginalContext ?? 0;
  37. result.rope_scaling_type = @params.YarnScalingType ?? RopeScalingType.Unspecified;
  38. result.defrag_threshold = @params.DefragThreshold;
  39. result.cb_eval = IntPtr.Zero;
  40. result.cb_eval_user_data = IntPtr.Zero;
  41. result.abort_callback = IntPtr.Zero;
  42. result.abort_callback_user_data = IntPtr.Zero;
  43. result.type_k = @params.TypeK ?? GGMLType.GGML_TYPE_F16;
  44. result.type_k = @params.TypeV ?? GGMLType.GGML_TYPE_F16;
  45. result.offload_kqv = !@params.NoKqvOffload;
  46. result.llama_pooling_type = @params.PoolingType;
  47. result.n_threads = Threads(@params.Threads);
  48. result.n_threads_batch = Threads(@params.BatchThreads);
  49. }
  50. private static uint Threads(uint? value)
  51. {
  52. if (value is > 0)
  53. return (uint)value;
  54. return (uint)Math.Max(Environment.ProcessorCount / 2, 1);
  55. }
  56. }
  57. }