From 28eb35108306e382fd749e210ebdb404ab0775fe Mon Sep 17 00:00:00 2001
From: Amit Bahree <github@desigeek.com>
Date: Sat, 26 Sep 2026 17:07:11 -0700
Subject: [PATCH] MEAP review feedback - Ch6-9

---
 .gitignore                                    |  21 +-
 CHANGELOG.md                                  |  14 +-
 .../eval/test_split/eval_report_test.log      |  34 ++++
 .../eval/test_split/it_qlora_rerun_train.log  | 192 ++++++++++++++++++
 .../eval/test_split/format_adherence_test.log |  35 ++++
 5 files changed, 293 insertions(+), 3 deletions(-)
 create mode 100644 code/chapter05/eval/test_split/eval_report_test.log
 create mode 100644 code/chapter05/eval/test_split/it_qlora_rerun_train.log
 create mode 100644 code/chapter07/eval/test_split/format_adherence_test.log

diff --git a/.gitignore b/.gitignore
index fc284fe..5dd6d5f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -50,6 +50,9 @@ cover/
 
 # Environments
 .env
+.env.*
+!.env.example
+!code/.env.example
 .envrc
 .venv
 env/
@@ -79,4 +82,20 @@ cython_debug/
 code/**/runs/
 code/**/wandb/
 code/**/__pycache__/
-code/**/.pytest_cache/
+code/**/.pytest_cache/
+
+# SSH keys / credentials (never commit private keys or secrets)
+*.pem
+*.key
+id_rsa*
+id_ed25519*
+id_ecdsa*
+*_rsa
+*_ed25519
+*_ecdsa
+*_key
+*_key.txt
+known_hosts
+*.ppk
+secrets.*
+credentials.*
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 46ad763..15d5132 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -8,6 +8,16 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and
 
 <!-- Changes landing on `main` between releases go here. -->
 
+### Added
+
+- **Chapters 6 through 9** (2026-09-23) — the remaining hands-on code: full supervised fine-tuning (`chapter06`), black-box distillation (`chapter07`), DPO with GRPO/RFT examples (`chapter08`), and the operations toolkit (`chapter09`: JSON model registry, TF-IDF drift detector, rollback demo, safety monitor). All run on the shared IT-support dataset in `data/it_support/` with the held-out test split (`--split test`).
+
+### Changed
+
+- **Chapter 6** (2026-09-26) — `eval/safety/` now holds the run the chapter prints (base 62%, fine-tuned 62%, harmful-request refusal 100% to 50%, regression flagged); the README's safety summary matches it.
+- **Chapter 8** (2026-09-26) — README training-time row reflects the measured runs (about 3.5 min full DPO on three 24 GB cards, about 2.8 min LoRA-DPO on one).
+- **Chapter 9** (2026-09-26) — README safety-monitor compare command writes `safety_report_dpo.json` against `safety_report.json`; `model_registry.py` docstring shows `--registry_dir` before the subcommand; the retired `data/drift_report_same_domain.json` removed.
+
 ## [MEAP-v0.1] — TBD
 
 First public release of the code repository alongside Manning's MEAP launch.
@@ -16,7 +26,7 @@ First public release of the code repository alongside Manning's MEAP launch.
 
 - Initial release of code for Chapters 1 through 5.
 - **Chapter 1** — reproducibility script for the §1.6 sidebar (`run_sidebar_example.py`). Runs the chapter's prompt through base Qwen3-4B, the Chapter 5 LoRA adapter, and the Chapter 6 SFT model side by side; degrades gracefully when later-chapter artifacts are not yet built.
-- **Chapter 2** — Unsloth-based fine-tuning quickstart reproducing the Dragon LLM open-finance recipe on Qwen3-0.6B end to end (data preparation across four HF datasets, LoRA via TRL's `SFTTrainer`, five evaluation tests, model export).
+- **Chapter 2** — five-step LoRA quickstart (`quickstart.py`) on Qwen3-4B-Instruct-2507 with a 40-example slice of the IT-support dataset: prepare data, load the base model with a LoRA config, train 20 steps with TRL's `SFTTrainer`, compare outputs before and after, save the adapter with a manifest. Plus `run_chapter5_adapter.py`, which previews the chapter 5 adapter (local or from the Hub) on the same prompts.
 - **Chapter 3** — data-quality experiment, six-step synthetic-data-generation pipeline using a frontier teacher, and a standalone `DatasetManifest` module for content hashing and lineage tracking.
 - **Chapter 4** — few-shot ticket classifier, many-shot prompt assembly, prompt validator with run-to-run variability measurement, minimal RAG pipeline (50 lines), Precision@k / Recall@k / Hit@1 retrieval evaluator.
 - **Chapter 5** — LoRA and QLoRA training, evaluation, and inference on a 400-example Dolly subset of Qwen3-4B-Instruct-2507; published adapter on Hugging Face Hub at `bahree/qwen3-4b-dolly-lora-ch5`.
@@ -27,7 +37,7 @@ First public release of the code repository alongside Manning's MEAP launch.
 
 ### Notes
 
-- Chapters 6 through 9 (Full SFT, Distillation, DPO, Operations) are written and will be released to this repo as they reach MEAP in subsequent drops.
+- Chapters 6 through 9 (Full SFT, Distillation, DPO, Operations) landed on `main` on 2026-09-23 (see Unreleased above).
 - All hands-on chapters use Qwen3-4B-Instruct-2507 as the base model and Databricks Dolly-15K (filtered subsets) as the dataset, so the chapters compose into a single coherent example pipeline.
 
 ---
diff --git a/code/chapter05/eval/test_split/eval_report_test.log b/code/chapter05/eval/test_split/eval_report_test.log
new file mode 100644
index 0000000..1f037ef
--- /dev/null
+++ b/code/chapter05/eval/test_split/eval_report_test.log
@@ -0,0 +1,34 @@
+
+Step 1/4: Loading base model...
+Failed to load /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_mxfp8.cpython-310-x86_64-linux-gnu.so: Could not load this library: /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_mxfp8.cpython-310-x86_64-linux-gnu.so
+Failed to load /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_cutlass_90a.abi3.so: Could not load this library: /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_cutlass_90a.abi3.so
+/home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torch/cuda/__init__.py:1061: UserWarning: Can't initialize NVML
+  raw_cnt = _raw_device_count_nvml()
+
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:01<00:02,  1.17s/it]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:02<00:01,  1.14s/it]
Loading checkpoint shards: 100%|██████████| 3/3 [00:02<00:00,  1.29it/s]
+✓ Base model loaded
+
+Step 2/4: Evaluating base model...
+The following generation flags are not valid and may be ignored: ['temperature', 'top_p', 'top_k']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+  Evaluating examples... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100%
+  Evaluating toy test set... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100%
+  Running safety checks... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100%
+✓ Base evaluation complete
+
+Step 3/4: Loading adapter from chapter05/runs/it_lora...
+
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:00<00:01,  1.64it/s]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:01<00:00,  1.60it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:01<00:00,  2.37it/s]
+✓ Adapter loaded
+
+Step 4/4: Evaluating fine-tuned model...
+  Evaluating examples... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100%
+  Evaluating toy test set... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100%
+  Running safety checks... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100%
+✓ Fine-tuned evaluation complete
+
+
+Writing evaluation reports...
+
+✓ Evaluation complete!
+✓ JSON report: chapter05/runs/eval_report/report.json
+✓ Markdown summary: chapter05/runs/eval_report/report.md
+
+→ View the markdown report for a human-readable summary
diff --git a/code/chapter05/eval/test_split/it_qlora_rerun_train.log b/code/chapter05/eval/test_split/it_qlora_rerun_train.log
new file mode 100644
index 0000000..afb2db5
--- /dev/null
+++ b/code/chapter05/eval/test_split/it_qlora_rerun_train.log
@@ -0,0 +1,192 @@
+Failed to load /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_mxfp8.cpython-310-x86_64-linux-gnu.so: Could not load this library: /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_mxfp8.cpython-310-x86_64-linux-gnu.so
+Failed to load /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_cutlass_90a.abi3.so: Could not load this library: /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_cutlass_90a.abi3.so
+/home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torch/cuda/__init__.py:1061: UserWarning: Can't initialize NVML
+  raw_cnt = _raw_device_count_nvml()
+Using the `WANDB_DISABLED` environment variable is deprecated and will be removed in v5. Use the --report_to flag to control the integrations used for logging result (for instance --report_to none).
+
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]/home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/bitsandbytes/backends/cuda/ops.py:213: FutureWarning: _check_is_size will be removed in a future PyTorch release along with guard_size_oblivious.     Use _check(i >= 0) instead.
+  torch._check_is_size(blocksize)
+
Loading checkpoint shards:  33%|███▎      | 1/3 [00:13<00:27, 13.72s/it]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:27<00:13, 13.72s/it]
Loading checkpoint shards: 100%|██████████| 3/3 [00:27<00:00,  7.62s/it]
Loading checkpoint shards: 100%|██████████| 3/3 [00:27<00:00,  9.27s/it]
+
Tokenizing train dataset:   0%|          | 0/450 [00:00<?, ? examples/s]
Tokenizing train dataset:  31%|███       | 138/450 [00:00<00:00, 1356.15 examples/s]
Tokenizing train dataset:  67%|██████▋   | 300/450 [00:00<00:00, 1500.48 examples/s]
Tokenizing train dataset: 100%|██████████| 450/450 [00:00<00:00, 1347.33 examples/s]
+
Tokenizing eval dataset:   0%|          | 0/50 [00:00<?, ? examples/s]
Tokenizing eval dataset: 100%|██████████| 50/50 [00:00<00:00, 1294.48 examples/s]
+The tokenizer has new PAD/BOS/EOS tokens that differ from the model config and generation config. The model config and generation config were aligned accordingly, being updated with the tokenizer's values. Updated tokens: {'bos_token_id': None, 'pad_token_id': 151643}.
+
  0%|          | 0/171 [00:00<?, ?it/s]
  1%|          | 1/171 [00:05<16:51,  5.95s/it]
  1%|          | 2/171 [00:11<15:38,  5.55s/it]
  2%|▏         | 3/171 [00:16<14:51,  5.31s/it]
  2%|▏         | 4/171 [00:21<14:44,  5.30s/it]
  3%|▎         | 5/171 [00:26<14:26,  5.22s/it]
  4%|▎         | 6/171 [00:32<14:38,  5.32s/it]
  4%|▍         | 7/171 [00:37<14:08,  5.18s/it]
  5%|▍         | 8/171 [00:41<13:52,  5.11s/it]
  5%|▌         | 9/171 [00:47<14:07,  5.23s/it]
  6%|▌         | 10/171 [00:52<13:55,  5.19s/it]
                                                

  6%|▌         | 10/171 [00:52<13:55,  5.19s/it]
  6%|▋         | 11/171 [00:58<14:05,  5.28s/it]
  7%|▋         | 12/171 [01:03<13:55,  5.25s/it]
  8%|▊         | 13/171 [01:08<14:05,  5.35s/it]
  8%|▊         | 14/171 [01:13<13:45,  5.26s/it]
  9%|▉         | 15/171 [01:19<13:56,  5.36s/it]
  9%|▉         | 16/171 [01:24<13:37,  5.28s/it]
 10%|▉         | 17/171 [01:30<13:55,  5.43s/it]
 11%|█         | 18/171 [01:35<13:36,  5.34s/it]
 11%|█         | 19/171 [01:41<13:42,  5.41s/it]
 12%|█▏        | 20/171 [01:46<13:29,  5.36s/it]
                                                

 12%|█▏        | 20/171 [01:46<13:29,  5.36s/it]
 12%|█▏        | 21/171 [01:52<13:40,  5.47s/it]
 13%|█▎        | 22/171 [01:57<13:25,  5.40s/it]
 13%|█▎        | 23/171 [02:02<13:29,  5.47s/it]
 14%|█▍        | 24/171 [02:07<13:06,  5.35s/it]
 15%|█▍        | 25/171 [02:13<13:05,  5.38s/it]
 15%|█▌        | 26/171 [02:18<12:46,  5.28s/it]
 16%|█▌        | 27/171 [02:23<12:51,  5.36s/it]
 16%|█▋        | 28/171 [02:29<12:34,  5.28s/it]
 17%|█▋        | 29/171 [02:34<12:27,  5.26s/it]
 18%|█▊        | 30/171 [02:39<12:12,  5.19s/it]
                                                

 18%|█▊        | 30/171 [02:39<12:12,  5.19s/it]
 18%|█▊        | 31/171 [02:44<12:23,  5.31s/it]
 19%|█▊        | 32/171 [02:50<12:09,  5.25s/it]
 19%|█▉        | 33/171 [02:55<12:15,  5.33s/it]
 20%|█▉        | 34/171 [03:00<12:03,  5.28s/it]
 20%|██        | 35/171 [03:05<11:49,  5.22s/it]
 21%|██        | 36/171 [03:11<11:45,  5.22s/it]
 22%|██▏       | 37/171 [03:16<11:31,  5.16s/it]
 22%|██▏       | 38/171 [03:21<11:31,  5.20s/it]
 23%|██▎       | 39/171 [03:26<11:19,  5.15s/it]
 23%|██▎       | 40/171 [03:31<11:33,  5.30s/it]
                                                

 23%|██▎       | 40/171 [03:31<11:33,  5.30s/it]
 24%|██▍       | 41/171 [03:37<11:18,  5.22s/it]
 25%|██▍       | 42/171 [03:42<11:20,  5.28s/it]
 25%|██▌       | 43/171 [03:47<11:09,  5.23s/it]
 26%|██▌       | 44/171 [03:53<11:18,  5.35s/it]
 26%|██▋       | 45/171 [03:58<11:01,  5.25s/it]
 27%|██▋       | 46/171 [04:03<11:01,  5.29s/it]
 27%|██▋       | 47/171 [04:08<10:34,  5.11s/it]
 28%|██▊       | 48/171 [04:13<10:34,  5.16s/it]
 29%|██▊       | 49/171 [04:18<10:12,  5.02s/it]
 29%|██▉       | 50/171 [04:23<10:24,  5.16s/it]
                                                

 29%|██▉       | 50/171 [04:23<10:24,  5.16s/it]{'loss': 2.4878, 'grad_norm': 0.478515625, 'learning_rate': 0.00018947368421052632, 'entropy': 1.6272694066166877, 'num_tokens': 19854.0, 'mean_token_accuracy': 0.5242026183754206, 'epoch': 0.18}
+{'loss': 1.9379, 'grad_norm': 0.65234375, 'learning_rate': 0.00017777777777777779, 'entropy': 1.9506154969334601, 'num_tokens': 37793.0, 'mean_token_accuracy': 0.5873597532510757, 'epoch': 0.36}
+{'loss': 1.8181, 'grad_norm': 0.59375, 'learning_rate': 0.00016608187134502925, 'entropy': 1.8098402447998523, 'num_tokens': 57150.0, 'mean_token_accuracy': 0.5992907334119082, 'epoch': 0.53}
+{'loss': 1.7138, 'grad_norm': 0.380859375, 'learning_rate': 0.0001543859649122807, 'entropy': 1.642770117521286, 'num_tokens': 74968.0, 'mean_token_accuracy': 0.6238362595438958, 'epoch': 0.71}
+{'loss': 1.7301, 'grad_norm': 0.416015625, 'learning_rate': 0.00014269005847953217, 'entropy': 1.6729066163301467, 'num_tokens': 93889.0, 'mean_token_accuracy': 0.6098065294325352, 'epoch': 0.89}
+
+
  0%|          | 0/50 [00:00<?, ?it/s]
+
  4%|▍         | 2/50 [00:00<00:03, 13.59it/s]
+
  8%|▊         | 4/50 [00:00<00:05,  8.71it/s]
+
 10%|█         | 5/50 [00:00<00:05,  8.13it/s]
+
 12%|█▏        | 6/50 [00:00<00:05,  7.61it/s]
+
 14%|█▍        | 7/50 [00:00<00:05,  7.42it/s]
+
 16%|█▌        | 8/50 [00:01<00:05,  7.26it/s]
+
 18%|█▊        | 9/50 [00:01<00:05,  6.86it/s]
+
 20%|██        | 10/50 [00:01<00:05,  6.87it/s]
+
 22%|██▏       | 11/50 [00:01<00:05,  6.89it/s]
+
 24%|██▍       | 12/50 [00:01<00:05,  6.91it/s]
+
 26%|██▌       | 13/50 [00:01<00:05,  6.92it/s]
+
 28%|██▊       | 14/50 [00:01<00:05,  6.92it/s]
+
 30%|███       | 15/50 [00:02<00:05,  6.97it/s]
+
 32%|███▏      | 16/50 [00:02<00:04,  7.01it/s]
+
 34%|███▍      | 17/50 [00:02<00:04,  7.01it/s]
+
 36%|███▌      | 18/50 [00:02<00:04,  7.03it/s]
+
 38%|███▊      | 19/50 [00:02<00:04,  7.04it/s]
+
 40%|████      | 20/50 [00:02<00:04,  7.07it/s]
+
 42%|████▏     | 21/50 [00:02<00:04,  7.10it/s]
+
 44%|████▍     | 22/50 [00:03<00:04,  6.80it/s]
+
 46%|████▌     | 23/50 [00:03<00:03,  6.80it/s]
+
 48%|████▊     | 24/50 [00:03<00:03,  6.81it/s]
+
 50%|█████     | 25/50 [00:03<00:03,  6.83it/s]
+
 52%|█████▏    | 26/50 [00:03<00:03,  6.85it/s]
+
 54%|█████▍    | 27/50 [00:03<00:03,  6.86it/s]
+
 56%|█████▌    | 28/50 [00:03<00:03,  6.87it/s]
+
 58%|█████▊    | 29/50 [00:04<00:03,  6.86it/s]
+
 60%|██████    | 30/50 [00:04<00:02,  6.87it/s]
+
 62%|██████▏   | 31/50 [00:04<00:02,  6.91it/s]
+
 64%|██████▍   | 32/50 [00:04<00:02,  6.94it/s]
+
 66%|██████▌   | 33/50 [00:04<00:02,  6.96it/s]
+
 68%|██████▊   | 34/50 [00:04<00:02,  7.01it/s]
+
 70%|███████   | 35/50 [00:04<00:02,  6.98it/s]
+
 72%|███████▏  | 36/50 [00:05<00:02,  6.98it/s]
+
 74%|███████▍  | 37/50 [00:05<00:01,  7.02it/s]
+
 76%|███████▌  | 38/50 [00:05<00:01,  6.99it/s]
+
 78%|███████▊  | 39/50 [00:05<00:01,  6.73it/s]
+
 80%|████████  | 40/50 [00:05<00:01,  6.80it/s]
+
 82%|████████▏ | 41/50 [00:05<00:01,  6.84it/s]
+
 84%|████████▍ | 42/50 [00:05<00:01,  6.87it/s]
+
 86%|████████▌ | 43/50 [00:06<00:01,  6.86it/s]
+
 88%|████████▊ | 44/50 [00:06<00:00,  6.88it/s]
+
 90%|█████████ | 45/50 [00:06<00:00,  6.89it/s]
+
 92%|█████████▏| 46/50 [00:06<00:00,  6.91it/s]
+
 94%|█████████▍| 47/50 [00:06<00:00,  6.95it/s]
+
 96%|█████████▌| 48/50 [00:06<00:00,  7.00it/s]
+
 98%|█████████▊| 49/50 [00:06<00:00,  7.00it/s]
+
100%|██████████| 50/50 [00:07<00:00,  7.01it/s]
                                                
+
                                               

 29%|██▉       | 50/171 [04:31<10:24,  5.16s/it]
+
100%|██████████| 50/50 [00:07<00:00,  7.01it/s]
+
                                               
 30%|██▉       | 51/171 [04:36<14:40,  7.34s/it]
 30%|███       | 52/171 [04:41<13:28,  6.79s/it]
 31%|███       | 53/171 [04:46<12:21,  6.28s/it]
 32%|███▏      | 54/171 [04:52<11:48,  6.05s/it]
 32%|███▏      | 55/171 [04:57<11:09,  5.77s/it]
 33%|███▎      | 56/171 [05:03<10:59,  5.74s/it]
 33%|███▎      | 57/171 [05:04<08:24,  4.42s/it]
 34%|███▍      | 58/171 [05:09<08:44,  4.64s/it]
 35%|███▍      | 59/171 [05:14<08:54,  4.78s/it]
 35%|███▌      | 60/171 [05:19<08:58,  4.85s/it]
                                                

 35%|███▌      | 60/171 [05:19<08:58,  4.85s/it]
 36%|███▌      | 61/171 [05:24<09:05,  4.96s/it]
 36%|███▋      | 62/171 [05:29<08:59,  4.95s/it]
 37%|███▋      | 63/171 [05:35<09:07,  5.07s/it]
 37%|███▋      | 64/171 [05:40<09:01,  5.06s/it]
 38%|███▊      | 65/171 [05:45<09:08,  5.18s/it]
 39%|███▊      | 66/171 [05:50<08:57,  5.11s/it]
 39%|███▉      | 67/171 [05:56<09:00,  5.19s/it]
 40%|███▉      | 68/171 [06:01<08:49,  5.14s/it]
 40%|████      | 69/171 [06:05<08:38,  5.08s/it]
 41%|████      | 70/171 [06:11<08:42,  5.18s/it]
                                                

 41%|████      | 70/171 [06:11<08:42,  5.18s/it]
 42%|████▏     | 71/171 [06:16<08:30,  5.11s/it]
 42%|████▏     | 72/171 [06:21<08:37,  5.22s/it]
 43%|████▎     | 73/171 [06:27<08:31,  5.22s/it]
 43%|████▎     | 74/171 [06:32<08:37,  5.33s/it]
 44%|████▍     | 75/171 [06:37<08:26,  5.27s/it]
 44%|████▍     | 76/171 [06:43<08:30,  5.37s/it]
 45%|████▌     | 77/171 [06:48<08:17,  5.29s/it]
 46%|████▌     | 78/171 [06:54<08:19,  5.37s/it]
 46%|████▌     | 79/171 [06:59<08:08,  5.31s/it]
 47%|████▋     | 80/171 [07:05<08:17,  5.47s/it]
                                                

 47%|████▋     | 80/171 [07:05<08:17,  5.47s/it]
 47%|████▋     | 81/171 [07:10<08:07,  5.42s/it]
 48%|████▊     | 82/171 [07:15<08:04,  5.44s/it]
 49%|████▊     | 83/171 [07:20<07:50,  5.34s/it]
 49%|████▉     | 84/171 [07:26<07:48,  5.38s/it]
 50%|████▉     | 85/171 [07:31<07:34,  5.29s/it]
 50%|█████     | 86/171 [07:36<07:35,  5.36s/it]
 51%|█████     | 87/171 [07:42<07:23,  5.28s/it]
 51%|█████▏    | 88/171 [07:47<07:22,  5.33s/it]
 52%|█████▏    | 89/171 [07:52<07:10,  5.24s/it]
 53%|█████▎    | 90/171 [07:58<07:11,  5.32s/it]
                                                

 53%|█████▎    | 90/171 [07:58<07:11,  5.32s/it]
 53%|█████▎    | 91/171 [08:03<07:00,  5.26s/it]
 54%|█████▍    | 92/171 [08:08<07:04,  5.37s/it]
 54%|█████▍    | 93/171 [08:13<06:50,  5.26s/it]
 55%|█████▍    | 94/171 [08:19<06:54,  5.39s/it]
 56%|█████▌    | 95/171 [08:24<06:39,  5.26s/it]
 56%|█████▌    | 96/171 [08:30<06:42,  5.36s/it]
 57%|█████▋    | 97/171 [08:35<06:29,  5.27s/it]
 57%|█████▋    | 98/171 [08:40<06:35,  5.41s/it]
 58%|█████▊    | 99/171 [08:45<06:20,  5.29s/it]
 58%|█████▊    | 100/171 [08:50<06:11,  5.23s/it]
                                                 

 58%|█████▊    | 100/171 [08:50<06:11,  5.23s/it]{'eval_loss': 1.7385369539260864, 'eval_runtime': 7.2588, 'eval_samples_per_second': 6.888, 'eval_steps_per_second': 6.888, 'eval_entropy': 1.7628279089927674, 'eval_num_tokens': 93889.0, 'eval_mean_token_accuracy': 0.6038225066661834, 'epoch': 0.89}
+{'loss': 1.7381, 'grad_norm': 0.408203125, 'learning_rate': 0.00013099415204678364, 'entropy': 1.7080736683832634, 'num_tokens': 111381.0, 'mean_token_accuracy': 0.6169048342350367, 'epoch': 1.05}
+{'loss': 1.6779, 'grad_norm': 0.36328125, 'learning_rate': 0.00011929824561403509, 'entropy': 1.6585225902497769, 'num_tokens': 130462.0, 'mean_token_accuracy': 0.6282420255243778, 'epoch': 1.23}
+{'loss': 1.6187, 'grad_norm': 0.458984375, 'learning_rate': 0.00010760233918128655, 'entropy': 1.5947254374623299, 'num_tokens': 150737.0, 'mean_token_accuracy': 0.6322602931410074, 'epoch': 1.41}
+{'loss': 1.719, 'grad_norm': 0.50390625, 'learning_rate': 9.590643274853801e-05, 'entropy': 1.6463869243860245, 'num_tokens': 168985.0, 'mean_token_accuracy': 0.6209228429943323, 'epoch': 1.59}
+{'loss': 1.6138, 'grad_norm': 0.5546875, 'learning_rate': 8.421052631578948e-05, 'entropy': 1.5865977682173251, 'num_tokens': 187364.0, 'mean_token_accuracy': 0.6334278259426356, 'epoch': 1.76}
+
+
  0%|          | 0/50 [00:00<?, ?it/s]
+
  4%|▍         | 2/50 [00:00<00:03, 14.42it/s]
+
  8%|▊         | 4/50 [00:00<00:05,  9.13it/s]
+
 12%|█▏        | 6/50 [00:00<00:05,  7.98it/s]
+
 14%|█▍        | 7/50 [00:00<00:05,  7.80it/s]
+
 16%|█▌        | 8/50 [00:00<00:05,  7.65it/s]
+
 18%|█▊        | 9/50 [00:01<00:05,  7.17it/s]
+
 20%|██        | 10/50 [00:01<00:05,  7.21it/s]
+
 22%|██▏       | 11/50 [00:01<00:05,  7.26it/s]
+
 24%|██▍       | 12/50 [00:01<00:05,  7.32it/s]
+
 26%|██▌       | 13/50 [00:01<00:05,  7.34it/s]
+
 28%|██▊       | 14/50 [00:01<00:04,  7.32it/s]
+
 30%|███       | 15/50 [00:01<00:04,  7.35it/s]
+
 32%|███▏      | 16/50 [00:02<00:04,  7.32it/s]
+
 34%|███▍      | 17/50 [00:02<00:04,  7.32it/s]
+
 36%|███▌      | 18/50 [00:02<00:04,  7.34it/s]
+
 38%|███▊      | 19/50 [00:02<00:04,  7.35it/s]
+
 40%|████      | 20/50 [00:02<00:04,  7.34it/s]
+
 42%|████▏     | 21/50 [00:02<00:03,  7.37it/s]
+
 44%|████▍     | 22/50 [00:02<00:03,  7.40it/s]
+
 46%|████▌     | 23/50 [00:03<00:03,  7.37it/s]
+
 48%|████▊     | 24/50 [00:03<00:03,  7.33it/s]
+
 50%|█████     | 25/50 [00:03<00:03,  7.23it/s]
+
 52%|█████▏    | 26/50 [00:03<00:03,  7.25it/s]
+
 54%|█████▍    | 27/50 [00:03<00:03,  7.26it/s]
+
 56%|█████▌    | 28/50 [00:03<00:03,  7.27it/s]
+
 58%|█████▊    | 29/50 [00:03<00:02,  7.28it/s]
+
 60%|██████    | 30/50 [00:04<00:02,  7.28it/s]
+
 62%|██████▏   | 31/50 [00:04<00:02,  7.30it/s]
+
 64%|██████▍   | 32/50 [00:04<00:02,  7.31it/s]
+
 66%|██████▌   | 33/50 [00:04<00:02,  7.34it/s]
+
 68%|██████▊   | 34/50 [00:04<00:02,  7.39it/s]
+
 70%|███████   | 35/50 [00:04<00:02,  7.28it/s]
+
 72%|███████▏  | 36/50 [00:04<00:01,  7.27it/s]
+
 74%|███████▍  | 37/50 [00:04<00:01,  7.31it/s]
+
 76%|███████▌  | 38/50 [00:05<00:01,  7.29it/s]
+
 78%|███████▊  | 39/50 [00:05<00:01,  6.91it/s]
+
 80%|████████  | 40/50 [00:05<00:01,  7.04it/s]
+
 82%|████████▏ | 41/50 [00:05<00:01,  7.11it/s]
+
 84%|████████▍ | 42/50 [00:05<00:01,  7.15it/s]
+
 86%|████████▌ | 43/50 [00:05<00:00,  7.20it/s]
+
 88%|████████▊ | 44/50 [00:05<00:00,  7.23it/s]
+
 90%|█████████ | 45/50 [00:06<00:00,  7.15it/s]
+
 92%|█████████▏| 46/50 [00:06<00:00,  7.23it/s]
+
 94%|█████████▍| 47/50 [00:06<00:00,  7.28it/s]
+
 96%|█████████▌| 48/50 [00:06<00:00,  7.36it/s]
+
 98%|█████████▊| 49/50 [00:06<00:00,  7.39it/s]
+
100%|██████████| 50/50 [00:06<00:00,  7.41it/s]
                                                 
+
                                               

 58%|█████▊    | 100/171 [08:57<06:11,  5.23s/it]
+
100%|██████████| 50/50 [00:06<00:00,  7.41it/s]
+
                                               
 59%|█████▉    | 101/171 [09:02<08:18,  7.12s/it]
 60%|█████▉    | 102/171 [09:07<07:23,  6.43s/it]
 60%|██████    | 103/171 [09:11<06:40,  5.88s/it]
 61%|██████    | 104/171 [09:16<06:12,  5.56s/it]
 61%|██████▏   | 105/171 [09:21<05:46,  5.25s/it]
 62%|██████▏   | 106/171 [09:26<05:31,  5.11s/it]
 63%|██████▎   | 107/171 [09:30<05:16,  4.94s/it]
 63%|██████▎   | 108/171 [09:35<05:04,  4.83s/it]
 64%|██████▎   | 109/171 [09:39<04:58,  4.81s/it]
 64%|██████▍   | 110/171 [09:44<04:52,  4.80s/it]
                                                 

 64%|██████▍   | 110/171 [09:44<04:52,  4.80s/it]
 65%|██████▍   | 111/171 [09:50<05:00,  5.02s/it]
 65%|██████▌   | 112/171 [09:55<04:57,  5.04s/it]
 66%|██████▌   | 113/171 [10:00<05:00,  5.19s/it]
 67%|██████▋   | 114/171 [10:02<03:49,  4.02s/it]
 67%|██████▋   | 115/171 [10:07<04:09,  4.46s/it]
 68%|██████▊   | 116/171 [10:12<04:15,  4.65s/it]
 68%|██████▊   | 117/171 [10:17<04:17,  4.77s/it]
 69%|██████▉   | 118/171 [10:23<04:25,  5.01s/it]
 70%|██████▉   | 119/171 [10:28<04:20,  5.01s/it]
 70%|███████   | 120/171 [10:33<04:21,  5.14s/it]
                                                 

 70%|███████   | 120/171 [10:33<04:21,  5.14s/it]
 71%|███████   | 121/171 [10:38<04:15,  5.10s/it]
 71%|███████▏  | 122/171 [10:44<04:15,  5.21s/it]
 72%|███████▏  | 123/171 [10:49<04:06,  5.13s/it]
 73%|███████▎  | 124/171 [10:54<04:04,  5.19s/it]
 73%|███████▎  | 125/171 [10:59<03:55,  5.12s/it]
 74%|███████▎  | 126/171 [11:04<03:54,  5.22s/it]
 74%|███████▍  | 127/171 [11:09<03:46,  5.16s/it]
 75%|███████▍  | 128/171 [11:15<03:46,  5.26s/it]
 75%|███████▌  | 129/171 [11:20<03:37,  5.18s/it]
 76%|███████▌  | 130/171 [11:25<03:27,  5.07s/it]
                                                 

 76%|███████▌  | 130/171 [11:25<03:27,  5.07s/it]
 77%|███████▋  | 131/171 [11:29<03:16,  4.92s/it]
 77%|███████▋  | 132/171 [11:34<03:10,  4.88s/it]
 78%|███████▊  | 133/171 [11:39<03:00,  4.75s/it]
 78%|███████▊  | 134/171 [11:43<02:54,  4.70s/it]
 79%|███████▉  | 135/171 [11:48<02:47,  4.64s/it]
 80%|███████▉  | 136/171 [11:52<02:42,  4.64s/it]
 80%|████████  | 137/171 [11:57<02:39,  4.68s/it]
 81%|████████  | 138/171 [12:03<02:43,  4.94s/it]
 81%|████████▏ | 139/171 [12:08<02:38,  4.94s/it]
 82%|████████▏ | 140/171 [12:13<02:39,  5.14s/it]
                                                 

 82%|████████▏ | 140/171 [12:13<02:39,  5.14s/it]
 82%|████████▏ | 141/171 [12:18<02:31,  5.04s/it]
 83%|████████▎ | 142/171 [12:23<02:23,  4.96s/it]
 84%|████████▎ | 143/171 [12:28<02:24,  5.15s/it]
 84%|████████▍ | 144/171 [12:33<02:18,  5.12s/it]
 85%|████████▍ | 145/171 [12:39<02:14,  5.16s/it]
 85%|████████▌ | 146/171 [12:44<02:07,  5.11s/it]
 86%|████████▌ | 147/171 [12:49<02:07,  5.29s/it]
 87%|████████▋ | 148/171 [12:54<01:58,  5.14s/it]
 87%|████████▋ | 149/171 [12:59<01:53,  5.15s/it]
 88%|████████▊ | 150/171 [13:04<01:45,  5.03s/it]
                                                 

 88%|████████▊ | 150/171 [13:04<01:45,  5.03s/it]{'eval_loss': 1.723291039466858, 'eval_runtime': 6.9149, 'eval_samples_per_second': 7.231, 'eval_steps_per_second': 7.231, 'eval_entropy': 1.6743564820289611, 'eval_num_tokens': 187364.0, 'eval_mean_token_accuracy': 0.6074856460094452, 'epoch': 1.76}
+{'loss': 1.685, 'grad_norm': 0.53515625, 'learning_rate': 7.251461988304094e-05, 'entropy': 1.6025203160941601, 'num_tokens': 204326.0, 'mean_token_accuracy': 0.6337072908878326, 'epoch': 1.94}
+{'loss': 1.6754, 'grad_norm': 0.546875, 'learning_rate': 6.0818713450292395e-05, 'entropy': 1.6512395030743368, 'num_tokens': 222326.0, 'mean_token_accuracy': 0.629128275690852, 'epoch': 2.11}
+{'loss': 1.5641, 'grad_norm': 0.55078125, 'learning_rate': 4.912280701754386e-05, 'entropy': 1.5786307081580162, 'num_tokens': 240777.0, 'mean_token_accuracy': 0.6360522732138634, 'epoch': 2.28}
+{'loss': 1.5358, 'grad_norm': 0.58984375, 'learning_rate': 3.742690058479532e-05, 'entropy': 1.4788970686495304, 'num_tokens': 260167.0, 'mean_token_accuracy': 0.6559325378388167, 'epoch': 2.46}
+{'loss': 1.5389, 'grad_norm': 0.6484375, 'learning_rate': 2.5730994152046783e-05, 'entropy': 1.4776077300310135, 'num_tokens': 277801.0, 'mean_token_accuracy': 0.6520779274404049, 'epoch': 2.64}
+
+
  0%|          | 0/50 [00:00<?, ?it/s]
+
  4%|▍         | 2/50 [00:00<00:03, 13.85it/s]
+
  8%|▊         | 4/50 [00:00<00:05,  8.77it/s]
+
 10%|█         | 5/50 [00:00<00:05,  8.18it/s]
+
 12%|█▏        | 6/50 [00:00<00:05,  7.64it/s]
+
 14%|█▍        | 7/50 [00:00<00:05,  7.45it/s]
+
 16%|█▌        | 8/50 [00:01<00:05,  7.28it/s]
+
 18%|█▊        | 9/50 [00:01<00:05,  6.86it/s]
+
 20%|██        | 10/50 [00:01<00:05,  6.79it/s]
+
 22%|██▏       | 11/50 [00:01<00:05,  6.81it/s]
+
 24%|██▍       | 12/50 [00:01<00:05,  6.86it/s]
+
 26%|██▌       | 13/50 [00:01<00:05,  6.85it/s]
+
 28%|██▊       | 14/50 [00:01<00:05,  6.85it/s]
+
 30%|███       | 15/50 [00:02<00:05,  6.82it/s]
+
 32%|███▏      | 16/50 [00:02<00:04,  6.82it/s]
+
 34%|███▍      | 17/50 [00:02<00:04,  6.83it/s]
+
 36%|███▌      | 18/50 [00:02<00:04,  6.84it/s]
+
 38%|███▊      | 19/50 [00:02<00:04,  6.78it/s]
+
 40%|████      | 20/50 [00:02<00:04,  6.64it/s]
+
 42%|████▏     | 21/50 [00:02<00:04,  6.67it/s]
+
 44%|████▍     | 22/50 [00:03<00:04,  6.72it/s]
+
 46%|████▌     | 23/50 [00:03<00:04,  6.73it/s]
+
 48%|████▊     | 24/50 [00:03<00:03,  6.71it/s]
+
 50%|█████     | 25/50 [00:03<00:03,  6.71it/s]
+
 52%|█████▏    | 26/50 [00:03<00:03,  6.81it/s]
+
 54%|█████▍    | 27/50 [00:03<00:03,  6.88it/s]
+
 56%|█████▌    | 28/50 [00:03<00:03,  6.93it/s]
+
 58%|█████▊    | 29/50 [00:04<00:03,  6.96it/s]
+
 60%|██████    | 30/50 [00:04<00:02,  6.96it/s]
+
 62%|██████▏   | 31/50 [00:04<00:02,  7.01it/s]
+
 64%|██████▍   | 32/50 [00:04<00:02,  7.01it/s]
+
 66%|██████▌   | 33/50 [00:04<00:02,  7.04it/s]
+
 68%|██████▊   | 34/50 [00:04<00:02,  7.06it/s]
+
 70%|███████   | 35/50 [00:04<00:02,  7.01it/s]
+
 72%|███████▏  | 36/50 [00:05<00:01,  7.01it/s]
+
 74%|███████▍  | 37/50 [00:05<00:01,  7.03it/s]
+
 76%|███████▌  | 38/50 [00:05<00:01,  7.01it/s]
+
 78%|███████▊  | 39/50 [00:05<00:01,  6.73it/s]
+
 80%|████████  | 40/50 [00:05<00:01,  6.82it/s]
+
 82%|████████▏ | 41/50 [00:05<00:01,  6.87it/s]
+
 84%|████████▍ | 42/50 [00:05<00:01,  6.88it/s]
+
 86%|████████▌ | 43/50 [00:06<00:01,  6.92it/s]
+
 88%|████████▊ | 44/50 [00:06<00:00,  6.90it/s]
+
 90%|█████████ | 45/50 [00:06<00:00,  6.83it/s]
+
 92%|█████████▏| 46/50 [00:06<00:00,  6.86it/s]
+
 94%|█████████▍| 47/50 [00:06<00:00,  6.89it/s]
+
 96%|█████████▌| 48/50 [00:06<00:00,  6.94it/s]
+
 98%|█████████▊| 49/50 [00:06<00:00,  6.97it/s]
+
100%|██████████| 50/50 [00:07<00:00,  6.99it/s]
                                                 
+
                                               

 88%|████████▊ | 150/171 [13:11<01:45,  5.03s/it]
+
100%|██████████| 50/50 [00:07<00:00,  6.99it/s]
+
                                               
 88%|████████▊ | 151/171 [13:17<02:26,  7.34s/it]
 89%|████████▉ | 152/171 [13:22<02:06,  6.65s/it]
 89%|████████▉ | 153/171 [13:27<01:52,  6.24s/it]
 90%|█████████ | 154/171 [13:32<01:39,  5.88s/it]
 91%|█████████ | 155/171 [13:37<01:30,  5.68s/it]
 91%|█████████ | 156/171 [13:42<01:22,  5.48s/it]
 92%|█████████▏| 157/171 [13:48<01:15,  5.40s/it]
 92%|█████████▏| 158/171 [13:53<01:09,  5.31s/it]
 93%|█████████▎| 159/171 [13:58<01:03,  5.32s/it]
 94%|█████████▎| 160/171 [14:03<00:58,  5.34s/it]
                                                 

 94%|█████████▎| 160/171 [14:03<00:58,  5.34s/it]
 94%|█████████▍| 161/171 [14:09<00:53,  5.39s/it]
 95%|█████████▍| 162/171 [14:14<00:47,  5.27s/it]
 95%|█████████▌| 163/171 [14:19<00:42,  5.30s/it]
 96%|█████████▌| 164/171 [14:24<00:36,  5.19s/it]
 96%|█████████▋| 165/171 [14:30<00:31,  5.25s/it]
 97%|█████████▋| 166/171 [14:35<00:25,  5.16s/it]
 98%|█████████▊| 167/171 [14:40<00:20,  5.22s/it]
 98%|█████████▊| 168/171 [14:45<00:15,  5.14s/it]
 99%|█████████▉| 169/171 [14:50<00:10,  5.21s/it]
 99%|█████████▉| 170/171 [14:55<00:05,  5.13s/it]
                                                 

 99%|█████████▉| 170/171 [14:55<00:05,  5.13s/it]
100%|██████████| 171/171 [14:56<00:00,  3.96s/it]
                                                 

100%|██████████| 171/171 [14:56<00:00,  3.96s/it]
100%|██████████| 171/171 [14:56<00:00,  5.25s/it]
+{'eval_loss': 1.740199327468872, 'eval_runtime': 7.2961, 'eval_samples_per_second': 6.853, 'eval_steps_per_second': 6.853, 'eval_entropy': 1.6042121207714082, 'eval_num_tokens': 277801.0, 'eval_mean_token_accuracy': 0.604690465927124, 'epoch': 2.64}
+{'loss': 1.5529, 'grad_norm': 0.671875, 'learning_rate': 1.4035087719298246e-05, 'entropy': 1.504924265295267, 'num_tokens': 296656.0, 'mean_token_accuracy': 0.6426991727203131, 'epoch': 2.82}
+{'loss': 1.5501, 'grad_norm': 0.68359375, 'learning_rate': 2.3391812865497075e-06, 'entropy': 1.5225243382155895, 'num_tokens': 315417.0, 'mean_token_accuracy': 0.6515443682670593, 'epoch': 3.0}
+{'train_runtime': 896.9171, 'train_samples_per_second': 1.505, 'train_steps_per_second': 0.191, 'train_loss': 1.714820304809258, 'entropy': 1.5296188592910767, 'num_tokens': 315915.0, 'mean_token_accuracy': 0.6488392353057861, 'epoch': 3.0}
+
+Peak GPU memory during QLoRA training:
+  cuda:0 (NVIDIA A30): 5.12 GiB allocated, 5.43 GiB reserved
+Saved QLoRA adapter to: chapter05/runs/it_qlora_rerun_2026-09-18
diff --git a/code/chapter07/eval/test_split/format_adherence_test.log b/code/chapter07/eval/test_split/format_adherence_test.log
new file mode 100644
index 0000000..ae473e8
--- /dev/null
+++ b/code/chapter07/eval/test_split/format_adherence_test.log
@@ -0,0 +1,35 @@
+Failed to load /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_mxfp8.cpython-310-x86_64-linux-gnu.so: Could not load this library: /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_mxfp8.cpython-310-x86_64-linux-gnu.so
+Failed to load /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_cutlass_90a.abi3.so: Could not load this library: /home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torchao/_C_cutlass_90a.abi3.so
+/home/amit/FTBook-pvt/code/.venv/lib/python3.12/site-packages/torch/cuda/__init__.py:1061: UserWarning: Can't initialize NVML
+  raw_cnt = _raw_device_count_nvml()
+Loaded 50 prompts from data/it_support/test.jsonl
+generating base...
+
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:00<00:01,  1.66it/s]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:01<00:00,  1.61it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:01<00:00,  2.39it/s]
+The following generation flags are not valid and may be ignored: ['temperature', 'top_p', 'top_k']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The attention mask is not set and cannot be inferred from input because pad token is same as eos token. As a consequence, you may observe unexpected behavior. Please pass your input's `attention_mask` to obtain reliable results.
+  base format 0% (0/50)
+generating teacher...
+
Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards:  50%|█████     | 1/2 [00:00<00:00,  1.13it/s]
Loading checkpoint shards: 100%|██████████| 2/2 [00:01<00:00,  1.54it/s]
Loading checkpoint shards: 100%|██████████| 2/2 [00:01<00:00,  1.46it/s]
+  teacher format 100% (50/50)
+generating student...
+  chapter07/runs/student_run1 is a LoRA adapter -> loading base + PeftModel
+
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:00<00:01,  1.64it/s]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:01<00:00,  1.62it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:01<00:00,  2.40it/s]
+  student format 100% (50/50)
+
+Saved chapter07/eval/test_split/format_report_test.json
+{
+  "num_examples": 50,
+  "valid_set": "data/it_support/test.jsonl",
+  "system_prompt": "You are an IT support assistant. Provide clear, step-by-step answers.",
+  "format_regex": "**Summary:** ... **Steps:** ... 1.",
+  "base_model": "Qwen/Qwen3-4B-Instruct-2507",
+  "teacher_dir": "chapter06/runs/sft_run1",
+  "student_dir": "chapter07/runs/student_run1",
+  "metric": "house_format_adherence (fraction matching **Summary:/Steps:** regex)",
+  "base_format_adherence": 0.0,
+  "base_format_count": "0/50",
+  "teacher_format_adherence": 1.0,
+  "teacher_format_count": "50/50",
+  "student_format_adherence": 1.0,
+  "student_format_count": "50/50"
+}