Create README.md

README.md

@@ -0,0 +1,23 @@
+---
+license: cc-by-nc-4.0
+pipeline_tag: text-generation
+---
+
+### Current State
+
+Well, I've reduced per-step loss by 0.36 (from 1.57 to 1.21) in a third of an epoch. To compare, the (meh) 13b Mistral glue LoRA reduced per-step loss by 0.37 (2.16 to 1.79) over an entire 4 epochs!
+
+EDIT: First 2 evals both came in at < 1, MUCH better than the 13b attempt. Verdict is that Mistral (or 7Bs in general, I'd guess) can't survive more than one cut, meaningfully. Per-step loss reduced by 0.47 @ 1 epoch.
+
+EDIT 2: Halfway there. Eval loss < 0.85 for the last 2 evals, promising. Per-step loss down to ~1.07, a reduction of ~33%!
+
+EDIT 3: 80% done. Curve is greatly flattened, so 3 epochs seems like it was the right call. Eval down to 0.81 and per-step loss down to 0.93. Can't wait to test!
+
+EDIT 4: Done! Testing time.
+
+### Dataset
+
+The 11b glue consists of:
+- The entirety of HF No Robots.
+- The entirety of TinyPixel/orca-mini
+- Enough of the GPT-4 generated Alpaca dataset (randomly chosen) to make it a roughly even three-way split.