Initial MLX-native tokenizer weights

README.md

@@ -0,0 +1,34 @@
+---
+license: mit
+library_name: kronos-mlx
+pipeline_tag: time-series-forecasting
+tags:
+  - mlx
+  - apple-silicon
+  - finance
+  - kronos
+  - tokenizer
+base_model: NeoQuasar/Kronos-Tokenizer-base
+---
+
+# Kronos-Tokenizer-base (MLX)
+
+Apple [MLX](https://github.com/ml-explore/mlx) port of [`NeoQuasar/Kronos-Tokenizer-base`](https://huggingface.co/NeoQuasar/Kronos-Tokenizer-base) — the BSQ (Binary Spherical Quantizer) tokenizer that compresses OHLCV candlestick sequences into hierarchical discrete tokens for the [Kronos](https://github.com/shiyu-coder/Kronos) family of forecasting models.
+
+Use it together with one of the MLX-native Kronos predictors, e.g. [`gxcsoccer/kronos-mlx-small`](https://huggingface.co/gxcsoccer/kronos-mlx-small).
+
+## Usage
+
+```python
+from kronos_mlx import Kronos, KronosTokenizer, KronosPredictor
+
+tokenizer = KronosTokenizer.from_pretrained("gxcsoccer/kronos-mlx-tokenizer-base")
+model     = Kronos.from_pretrained("gxcsoccer/kronos-mlx-small")
+
+predictor = KronosPredictor(model, tokenizer, max_context=512)
+```
+
+## Original
+
+- Upstream: [shiyu-coder/Kronos](https://github.com/shiyu-coder/Kronos)
+- PyTorch weights: [NeoQuasar/Kronos-Tokenizer-base](https://huggingface.co/NeoQuasar/Kronos-Tokenizer-base)

config.json

@@ -0,0 +1,18 @@
+{
+  "attn_dropout_p": 0.0,
+  "beta": 0.05,
+  "d_in": 6,
+  "d_model": 256,
+  "ff_dim": 512,
+  "ffn_dropout_p": 0.0,
+  "gamma": 1.1,
+  "gamma0": 1.0,
+  "group_size": 4,
+  "n_dec_layers": 4,
+  "n_enc_layers": 4,
+  "n_heads": 4,
+  "resid_dropout_p": 0.0,
+  "s1_bits": 10,
+  "s2_bits": 10,
+  "zeta": 0.05
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c6f4ee1d9a96571e8afb3d78a16158c972ae273ded8bd4e78ae9c4d2081049b
+size 15840646