Seungbaek Hong
Seungbaek Hong
add some options to 'generate' function of llm - add naive repetition_penalty option - add bad_words option **Self evaluation:** 1. Build test: [X]Passed [ ]Failed [ ]Skipped 2. Run test:...
Added generate_multiple_tokens function for first generation on llm. This function takes one logits and generates multiple output tokens. To meet the purpose of the target application, even if input are...
I added tensor & function(operation) API structure for supporting autograd. Users can make a model graph using this api. The operators will be supported one-to-one with ONNX. Signed-off-by: Seungbaek Hong
It is a list of embedding layers that need to be implemented to support the diffusion model. It would be nice if anyone who has time could take care of...
We already support operations such as multiply or transpose at the tensor level, but if we support them at the layer level like addition_layer, it will be much more convenient...
When creating a model with multiple input layers, it need to be synchronized based on the order in which the layers are defined and the order in which the data...
When creating an application using MSE Loss, training proceeds even if the shape of the target (label) is different from the shape of the final output of the model. In...
As @lhs8928 left in the [comment](https://github.com/nnstreamer/nntrainer/pull/2068#issuecomment-1359148463) of PR #2068, It seems better to support free_mem option on inference function even if there is no loss layer. Then, the free_mem option...
Currently, developers using API in NNTrainer do not have access to tensor data structures. However, it is inconvenient to handle data without using a tensor data structure. Since other deep...
[Wait for #2571][Application] cuda support for example of pytorch yolo v2 @open sesame 05/10 19:54
- add cuda option to train yolo v2 model backbone (for converting to nntrainer model) - preprocessing for input dataset * unmatched paired dataset * no annotation value Signed-off-by: Seungbaek...