Update README.md

README.md

@@ -1,3 +1,48 @@
 ---
 license: apache-2.0
+datasets:
+- 0-hero/lj_speech_with_spectogram_conversations
+- 0-hero/lj_speech_with_spectogram
 ---
+## Explanation
+A small experiment insipred by the [Mistral playing DOOM experiment](https://github.com/umuthopeyildirim/DOOM-Mistral/tree/main) from the Mistral Hackathon
+
+**How it works?** 
+``` 
+Audio -> Waveform Visualization -> Waveform ASCII Art -> Finetune Mistral on ASCII Art to predict text from ASCII Art
+```
+
+**Quick video explanation**
+<video width="640" controls src="https://cdn-uploads.huggingface.co/production/uploads/6382255fcae34727b9cc149e/nCB8Qu8QwDbJAKcq9IzPE.mp4"></video>
+
+## Models & Results
+Mistral 7B 0.2 finetunes on ascii art. **As seen in the results, experiment didn't amount to much**
+- 2 epochs - [0-hero/mistral-speech-to-text-preview](https://huggingface.co/0-hero/mistral-speech-to-text-preview/) (this) - Loss pretty much flattened after this epoch
+  - ```
+    Metrics:
+      rouge-1: {'r': 0.12919024091165357, 'p': 0.1163312036605547, 'f': 0.11315199212991178}
+      rouge-2: {'r': 0.013705453572242508, 'p': 0.0137500428446463, 'f': 0.012676757505648992}
+      rouge-l: {'r': 0.11261286554140228, 'p': 0.09921920076529338, 'f': 0.09705621471622536}
+      length_correlation: 0.014470676120233311
+      avg_actual_length: 16.59
+      avg_pred_length: 21.46
+      exact_match_accuracy: 0.0
+    ```
+- 4 epochs - [0-hero/mistral-speech-to-text](https://huggingface.co/0-hero/mistral-speech-to-text/)
+  - ```
+    Metrics:
+      rouge-1: {'r': 0.11869828051815862, 'p': 0.11697319273190071, 'f': 0.11154343875398197}
+      rouge-2: {'r': 0.008572925612399297, 'p': 0.009040061245943597, 'f': 0.008369604666309954}
+      rouge-l: {'r': 0.10780857719316121, 'p': 0.10373665666448233, 'f': 0.09985384905943501}
+      length_correlation: -0.1500200314034927
+      avg_actual_length: 16.59
+      avg_pred_length: 18.32
+      exact_match_accuracy: 0.0
+    ```
+
+## Datasets
+[lj_speech](https://huggingface.co/datasets/lj_speech) dataset used to convert audio waveforms into ASCII Art
+- [0-hero/lj_speech_with_spectogram_conversations](https://huggingface.co/datasets/0-hero/lj_speech_with_spectogram_conversations) - ShareGPT style finetuning dataset with train, test split
+- [0-hero/lj_speech_with_spectogram](https://huggingface.co/datasets/0-hero/lj_speech_with_spectogram) - Raw dataset with ASCII Art
+
+