---
language: 
  - en
tags:
  - NLP
license: mit
datasets:
  - TristanBehrens/metal_cluster
  - TristanBehrens/metal_arrangement
  - TristanBehrens/metal_interleaved
base_model: None
---

# metal_omni_two - An xLSTM Model

![Trained with Helibrunna](banner.jpg)

Trained with [Helibrunna](https://github.com/AI-Guru/helibrunna) by [Dr. Tristan Behrens](https://de.linkedin.com/in/dr-tristan-behrens-734967a2).

## Configuration

```
training:
  model_name: metal_omni_two
  batch_size: 28
  lr: 0.001
  lr_warmup_steps: 1445
  lr_decay_until_steps: 14455
  lr_decay_factor: 0.001
  weight_decay: 0.1
  amp_precision: bfloat16
  weight_precision: float32
  enable_mixed_precision: true
  num_epochs: 5
  output_dir: output/metal_omni_two
  save_every_step: 500
  log_every_step: 10
  wandb_project: tonnetz
  torch_compile: false
model:
  type: llamathree
  context_length: 2048
  emb_dim: 256
  n_heads: 4
  n_layers: 6
  hidden_dim: 128
  hidden_activation: silu
  n_kv_groups: 1
  rope_base: 50000
  rope_freq: null
  dtype: float32
  vocab_size: 269
dataset:
  hugging_face_ids:
  - TristanBehrens/metal_cluster
  - TristanBehrens/metal_arrangement
  - TristanBehrens/metal_interleaved
tokenizer:
  type: whitespace
  fill_token: '[EOS]'

```