Upload README.md with huggingface_hub

README.md

@@ -1,51 +1,40 @@
----
-base_model: google/gemma-4-E4B-it
-library_name: peft
-license: apache-2.0
-tags:
-- gemma4
-- coding
-- qlora
-- kaggle-proof
----
-
-# Gemma 4 E4B IT Coding LoRA
-
-QLoRA adapter for `google/gemma-4-E4B-it`, trained on filtered benign coding instructions.
-
-## Training
-
-- Runtime: Kaggle 2x Tesla T4
-- Data: filtered benign coding instruction data
-- Safe rows used: 1024
-- Steps: 200
-- LoRA: r=16, alpha=32, target_modules=`all-linear`
-- Trainable parameters: 50,499,584
-- Final train loss: 1.1427
-
-## Proof
-
-- HumanEval subset: first 8 tasks
-- Executable pass count before: 5/8
-- Executable pass count after: 7/8
-- Heuristic score before: 0.7688
-- Heuristic score after: 0.7688
-- Relative executable pass-count increase: 40%
-- Absolute executable pass-rate increase: +25 percentage points
-
-The public executable proof is intentionally small because the Kaggle GPU-hour
-budget was exhausted during training, merge preparation, and upload validation.
-`eval_before_after.csv` contains output previews; executable pass/fail proof is
-recorded in `executable_eval.json`.
-
-Artifacts included:
-
-- `eval_before_after.csv`
-- `executable_eval.json`
-- `trainer_log_history.json`
-- `summary.json`
-- `proof_summary.json`
-- `nvidia_smi.txt`
-- `evaluation_scope.json`
-
-This adapter is for benign coding assistance only. It was not trained on malware, phishing, exploit, credential theft, evasion, or destructive automation examples.
+---
+base_model: google/gemma-4-E4B-it
+library_name: peft
+license: apache-2.0
+tags:
+- gemma4
+- coding
+- qlora
+- kaggle-proof
+---
+
+# Gemma 4 E4B IT Coding LoRA
+
+QLoRA adapter for `google/gemma-4-E4B-it`, trained on filtered benign coding instructions.
+
+## Training
+
+- Runtime: Kaggle 2x Tesla T4
+- Data: filtered benign coding instruction data
+- Safe rows used: 1024
+- Steps: 200
+- LoRA: r=16, alpha=32, target_modules=`all-linear`
+- Trainable parameters: 50,499,584
+- Final train loss: 1.1427
+
+## 50-Problem HumanEval Proof
+
+This adapter was merged into `josephmayo/gemma-4-E4B-it-Coder` and evaluated on Kaggle with 2x Tesla T4 GPUs using an executable 50-task HumanEval subset. Full generated before/after code is published in `eval50_before_after_full_code.csv`.
+
+| Metric | Base `google/gemma-4-E4B-it` | Coder merge |
+|---|---:|---:|
+| Pass count | 34 / 50 | 42 / 50 |
+| Absolute lift | - | +16.0 pp |
+| Relative pass-count lift | - | +23.53% |
+
+Proof files included here: `eval50_summary.json`, `eval50_before_after_full_code.csv`, `EVAL50_README.md`, `nvidia_smi.txt`.
+
+Earlier 8-task smoke artifacts are still included for reproducibility (`eval_before_after.csv`, `executable_eval.json`, `trainer_log_history.json`, `summary.json`, `proof_summary.json`, `evaluation_scope.json`), but the headline proof is the 50-task executable run above.
+
+This adapter is for benign coding assistance only. It was not trained on malware, phishing, exploit, credential theft, evasion, or destructive automation examples.