Upload Distance-Aware Chronos model

README.md

@@ -0,0 +1,73 @@
+---
+license: apache-2.0
+tags:
+- time-series
+- forecasting
+- chronos
+- distance-aware
+library_name: transformers
+---
+
+# Distance-Aware Chronos
+
+This is a distance-aware enhancement of the Chronos time series forecasting model.
+
+## Model Description
+
+This model extends the original [Chronos](https://github.com/amazon-science/chronos-forecasting) 
+architecture with distance-aware loss functions and output layers that explicitly consider the 
+ordinal nature of quantized time series bins.
+
+**Base Model:** amazon/chronos-t5-small  
+**Number of Bins:** 4096  
+**Training Epoch:** 10  
+**Validation Loss:** 2.4703
+
+## Key Features
+
+- **Distance-Aware Loss:** Combines ordinal cross-entropy, smooth label loss, and Earth Mover's Distance
+- **Ordinal Output Layer:** Uses Gaussian kernels and sinusoidal position encodings
+- **Improved Bin Predictions:** Better handling of nearby bin relationships
+
+## Installation
+```bash
+pip install torch transformers chronos
+```
+
+## Usage
+```python
+from distance_aware_chronos import DistanceAwareChronos
+import numpy as np
+
+# Load model
+model = DistanceAwareChronos.from_pretrained("Phoenix21/distance-aware-chronos-t")
+
+# Prepare your time series
+context = np.array([1.0, 2.0, 3.0, 4.0, 5.0])  # Your historical data
+
+# Generate forecasts
+predictions = model.predict(context, horizon=24, num_samples=100)
+
+print(f"Forecast shape: {predictions.shape}")
+```
+
+## Training Data
+
+Trained on the [Chronos datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) 
+from HuggingFace.
+
+## Citation
+
+If you use this model, please cite:
+```bibtex
+@article{chronos2024,
+  title={Chronos: Learning the Language of Time Series},
+  author={Ansari, Abdul Fatir et al.},
+  journal={Transactions on Machine Learning Research},
+  year={2024}
+}
+```
+
+## License
+
+Apache 2.0

base_model/config.json

@@ -0,0 +1,49 @@
+{
+  "architectures": [
+    "T5ForConditionalGeneration"
+  ],
+  "chronos_config": {
+    "context_length": 512,
+    "eos_token_id": 1,
+    "model_type": "seq2seq",
+    "n_special_tokens": 2,
+    "n_tokens": 4096,
+    "num_samples": 20,
+    "pad_token_id": 0,
+    "prediction_length": 64,
+    "temperature": 1.0,
+    "tokenizer_class": "MeanScaleUniformBins",
+    "tokenizer_kwargs": {
+      "high_limit": 15.0,
+      "low_limit": -15.0
+    },
+    "top_k": 50,
+    "top_p": 1.0,
+    "use_eos_token": true
+  },
+  "classifier_dropout": 0.0,
+  "d_ff": 2048,
+  "d_kv": 64,
+  "d_model": 512,
+  "decoder_start_token_id": 0,
+  "dense_act_fn": "relu",
+  "dropout_rate": 0.1,
+  "dtype": "float32",
+  "eos_token_id": 1,
+  "feed_forward_proj": "relu",
+  "initializer_factor": 0.05,
+  "is_encoder_decoder": true,
+  "is_gated_act": false,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "t5",
+  "n_positions": 512,
+  "num_decoder_layers": 6,
+  "num_heads": 8,
+  "num_layers": 6,
+  "pad_token_id": 0,
+  "relative_attention_max_distance": 128,
+  "relative_attention_num_buckets": 32,
+  "transformers_version": "4.57.1",
+  "use_cache": true,
+  "vocab_size": 4096
+}

base_model/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "decoder_start_token_id": 0,
+  "eos_token_id": 1,
+  "pad_token_id": 0,
+  "transformers_version": "4.57.1"
+}

base_model/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:099347ca49e954892070992dae2fa31abecf5e54fcfffe876d69d9ddc3602ed2
+size 184632360

config.json

@@ -0,0 +1,8 @@
+{
+  "model_type": "distance_aware_chronos",
+  "base_model": "amazon/chronos-t5-small",
+  "num_bins": 4096,
+  "training_epoch": 10,
+  "val_loss": 2.470318087331065,
+  "timestamp": "2025-11-21T04:18:03.413552"
+}

distance_output.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cc68ee4f559f59b1aca371fa339dc84269f7649d688f1074328eb6233b321fc
+size 10678922

example_usage.py

@@ -0,0 +1,28 @@
+# Example: Using Distance-Aware Chronos
+
+import numpy as np
+from distance_aware_chronos import DistanceAwareChronos
+
+# Load the model
+model = DistanceAwareChronos.from_pretrained("YOUR_USERNAME/distance-aware-chronos")
+
+# Example 1: Simple forecasting
+context = np.random.randn(100)  # Your time series
+forecast = model.predict(context, horizon=24, num_samples=100)
+
+print(f"Forecast shape: {forecast.shape}")
+print(f"Mean forecast: {forecast.mean()}")
+
+# Example 2: With visualization
+import matplotlib.pyplot as plt
+
+plt.figure(figsize=(12, 6))
+plt.plot(range(len(context)), context, label='Historical', color='blue')
+plt.plot(range(len(context), len(context) + len(forecast)), 
+         forecast, label='Forecast', color='red', linestyle='--')
+plt.xlabel('Time')
+plt.ylabel('Value')
+plt.title('Time Series Forecast')
+plt.legend()
+plt.grid(True, alpha=0.3)
+plt.show()