|
Mila
Deep Neural Network Library
|
Abstract base for Mila autoregressive language models. More...
#include <vector>#include <algorithm>#include <span>#include <unordered_set>#include <memory>#include <string>#include <stdexcept>#include <format>#include <functional>#include <stop_token>#include <cstddef>#include <cstdint>#include <filesystem>import Dnn.LanguageModelConfig;import Compute.CpuMemoryResource;import Serialization.WeightsReader;import Compute.Observation;import Dnn.Model;import Dnn.TensorTypes;import Compute.DeviceType;import Dnn.TensorDataType;import Dnn.Samplers.SamplingConfig;import Dnn.Tensor;import Serialization.SafeTensors;import Dnn.RuntimeMode;import Dnn.TensorOps;import Compute.DeviceTypeTraits;import Dnn.TensorDataTypeTraits;import Compute.Device;import Dnn.GenerateParams;import Dnn.SamplingParams;import Dnn.LanguageModelNetwork;import Dnn.GenerateStatus;import Dnn.Samplers.TokenSampler;Classes | |
| class | Mila::Dnn::LanguageModel< TDeviceType, TPrecision > |
Namespaces | |
| namespace | Mila |
| Mila main API namespace. | |
Abstract base for Mila autoregressive language models.
Extends Model with blocking and streaming generation APIs common to all language models. Derived classes implement the prefill + decode loop via the protected onGenerating() hook.