Upload fine-tuned Chronos model

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+forecast_example_1.png filter=lfs diff=lfs merge=lfs -text
+forecast_example_2.png filter=lfs diff=lfs merge=lfs -text
+forecast_example_3.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,93 @@
+---
+language: en
+license: apache-2.0
+library_name: chronos
+tags:
+- chronos
+- time-series
+- forecasting
+- finance
+- cryptocurrency
+datasets:
+- time-series
+---
+
+# chronos-t5-small-btc-m1
+
+This is a Chronos model fine-tuned on financial time series data. The model is based on the T5 architecture and is designed for time series forecasting.
+
+## Model Description
+
+- **Model Type:** Chronos (T5-based time series forecasting model)
+- **Fine-tuned from:** amazon/chronos-t5-small
+- **Uploaded by:** mainmagic
+- **Date:** 2025-04-06
+
+Chronos model fine-tuned on BTC/USD M1 data for time series forecasting
+
+## Performance Metrics
+
+| Metric | Value |
+|--------|-------|
+| mse | 1.0823 |
+| mae | 0.8172 |
+| mape | 16552.9256 |
+
+
+
+## Usage
+
+```python
+# Import the Chronos pipeline
+# Note: You may need to adjust the import path based on your installation
+import sys
+sys.path.append('/path/to/chronos-forecasting/src')  # Adjust this path
+from chronos.chronos import ChronosPipeline
+import torch
+
+# Load the model
+pipeline = ChronosPipeline.from_pretrained("mainmagic/chronos-t5-small-btc-m1")
+
+# Create input data (example)
+context = torch.randn(1, 512)  # Batch size 1, context length 512
+
+# Generate forecast
+forecast = pipeline.predict(
+    context,
+    prediction_length=60,  # Predict 60 steps ahead
+    num_samples=20  # Generate 20 different forecast trajectories
+)
+
+# Use median as point forecast
+median_forecast = torch.median(forecast, dim=1)[0]
+```
+
+## Training Details
+
+This model was fine-tuned using the Chronos native training scripts. The model was trained on financial time series data with the following parameters:
+
+- Context length: 512
+- Prediction length: 60
+- Optimizer: adamw_torch
+- Learning rate: 0.0001
+- Batch size: 16
+- Gradient accumulation steps: 4
+
+## Limitations
+
+This model is specifically trained for financial time series forecasting and may not perform well on other types of time series data. The model's performance may also vary depending on market conditions and the specific financial instrument being forecasted.
+
+## Citation
+
+If you use this model, please cite:
+
+```bibtex
+@misc{chronos-forecasting,
+  author = {Amazon Science},
+  title = {Chronos: Learning the Language of Time Series},
+  year = {2024},
+  publisher = {GitHub},
+  journal = {GitHub repository},
+  howpublished = {\url{https://github.com/amazon-science/chronos-forecasting}}
+}
+```

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": 60,
+    "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,
+  "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,
+  "torch_dtype": "float32",
+  "transformers_version": "4.51.0",
+  "use_cache": true,
+  "vocab_size": 4096
+}

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": "4.51.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6be3b1f11a7f5b0aac1048e4b4ca77416240f6005677b5b24e741935a1a59b73
+size 184632360

normalization_params.json

@@ -0,0 +1 @@
+{"min_vals": {"open": 3134.9, "high": 3134.9, "low": 3134.9, "close": 3134.9, "volume": 1.0}, "max_vals": {"open": 109288.19, "high": 109288.19, "low": 109288.19, "close": 109288.19, "volume": 1044.0}}

training_info.json

@@ -0,0 +1,50 @@
+{
+    "training_config": {
+        "training_data_paths": "['./chronos_training_data.arrow']",
+        "probability": "[1.0]",
+        "context_length": 512,
+        "prediction_length": 60,
+        "min_past": 60,
+        "max_steps": 100,
+        "save_steps": 50,
+        "log_steps": 100,
+        "per_device_train_batch_size": 16,
+        "learning_rate": 0.0001,
+        "optim": "adamw_torch",
+        "shuffle_buffer_length": 10000,
+        "gradient_accumulation_steps": 4,
+        "model_id": "amazon/chronos-t5-small",
+        "model_type": "seq2seq",
+        "random_init": false,
+        "tie_embeddings": true,
+        "output_dir": "./chronos-native-fine-tuned-model",
+        "tf32": false,
+        "torch_compile": false,
+        "tokenizer_class": "MeanScaleUniformBins",
+        "tokenizer_kwargs": "{'low_limit': -15.0, 'high_limit': 15.0}",
+        "n_tokens": 4096,
+        "n_special_tokens": 2,
+        "pad_token_id": 0,
+        "eos_token_id": 1,
+        "use_eos_token": true,
+        "lr_scheduler_type": "linear",
+        "warmup_ratio": 0.1,
+        "dataloader_num_workers": 0,
+        "max_missing_prop": 0.9,
+        "num_samples": 20,
+        "temperature": 1.0,
+        "top_k": 50,
+        "top_p": 1.0,
+        "seed": 1907066225
+    },
+    "job_info": {
+        "cuda_available": false,
+        "torchelastic_launched": false,
+        "python_version": "3.13.2 (main, Feb  4 2025, 14:51:09) [Clang 16.0.0 (clang-1600.0.26.6)]",
+        "torch_version": "2.6.0",
+        "numpy_version": "1.26.4",
+        "gluonts_version": "0.16.0",
+        "transformers_version": "4.51.0",
+        "accelerate_version": "0.34.2"
+    }
+}