|
Mila
Deep Neural Network Library
|
Exported Modules | |
| module | Compute.DeviceTypeTraits |
| module | Dnn.ActivationType |
| module | Dnn.Components.RmsNorm |
| module | Dnn.ITensor |
| module | Compute.ExecutionContext |
| module | Compute.DeviceType |
| module | Serialization.ModelArchive |
| module | Dnn.Quantization.Weight.Policies |
| module | Compute.GqaState |
| module | Dnn.TensorDataTypeTraits |
| module | Dnn.TensorTypes |
| module | Dnn.RuntimeMode |
| module | Compute.CpuMemoryResource |
| module | Dnn.ModelType |
| module | Dnn.Components.Linear |
| module | Dnn.Tensor |
| module | Compute.DeviceId |
| module | Compute.ExecutionContextFactory |
| module | Dnn.LanguageModelNetwork |
| module | Dnn.Components.Rope |
| module | Serialization.Tensor |
| module | Serialization.WeightsReader |
| module | Dnn.ComponentType |
| module | Dnn.TensorDataType |
| module | Logging.Logger |
| module | Dnn.Quantization.KvCache.Policy |
| module | Compute.DeviceAllocation |
| module | Compute.Device |
| module | Dnn.Components.TokenEmbedding |
| module | Dnn.Component |
| module | Compute.GqaWorkspace |
Classes | |
| class | Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy > |
| class | Mila::Dnn::LlamaConfig |
| Network-level configuration for LLaMA-style transformer networks. More... | |
| class | Mila::Dnn::LlamaTransformer< TDeviceType, TPrecision, TWeightQuantization, TKvCachePolicy > |
| LLaMA-style transformer (decoder-only) for autoregressive token prediction. More... | |
Functions | |
| LlamaConfig | Mila::Dnn::Llama3_2_1B () |
| Usage Examples: | |
| LlamaConfig | Mila::Dnn::Llama3_2_3B () |
| Llama 3.2 3B. | |
| LlamaConfig | Mila::Dnn::Llama3_1_8B () |
| Llama 3.1 8B. | |
| LlamaConfig | Mila::Dnn::Llama3_1_70B () |
| Llama 3.1 70B. | |
| LlamaConfig | Mila::Dnn::Llama3_1_405B () |
| Llama 3.1 405B. | |
| LlamaConfig | Mila::Dnn::Llama3_8B () |
| Llama 3 8B (Original release). | |
| LlamaConfig | Mila::Dnn::Llama3_70B () |
| Llama 3 70B (Original release). | |
| LlamaConfig | Mila::Dnn::Llama2_7B () |
| Llama 2 7B. | |
| LlamaConfig | Mila::Dnn::Llama2_13B () |
| Llama 2 13B. | |
| LlamaConfig | Mila::Dnn::Llama2_70B () |
| Llama 2 70B. | |
Variables | |
| constexpr int64_t | Mila::Dnn::kLlamaPrefillChunkFloor = 128 |
| constexpr int64_t | Mila::Dnn::kLlamaPrefillChunkRungs [] = { 512, 256, 128 } |
Files | |
| file | Mila/Src/Dnn/Components/Transformers/LlaMa/Llama.ixx |
| LLaMA-style decoder-only transformer network. | |
| file | Mila/Src/Dnn/Components/Transformers/LlaMa/Llama.Block.ixx |
| LLaMA transformer block – module partition of LlamaTransformer. | |
| file | Mila/Src/Dnn/Components/Transformers/LlaMa/Llama.Config.ixx |
| LLaMA network-level configuration. | |
| file | Mila/Src/Dnn/Components/Transformers/LlaMa/Llama.Presets.ixx |