tests: 5/5 code-phase suite PASS + reports
Browse files- docs/PROGRESS.md +1 -0
- docs/TEST_REPORT.md +9 -0
docs/PROGRESS.md
CHANGED
|
@@ -27,6 +27,7 @@ Har kaam ke baad yaha entry karo taaki yaad rahe kab kya kiya tha.
|
|
| 27 |
- [x] verify: `viu1_moe.py` 241.7M/166.2M OK, `train.py --smoke` 3 steps OK, push dry-run OK (data/raw excluded)
|
| 28 |
- [x] 2026-09-16: local rename `Mini-ViuAI-50M/` -> `ViuMini-MoE-242M/`, README base_model null fix, HF push OK -> https://huggingface.co/ViuAI/ViuMini-MoE-242M
|
| 29 |
- [x] 2026-09-17: Kaggle smoke FAIL (`data/raw` HF me excluded) -> `train.py` smoke fallback (built-in dummy lines), Kaggle re-test PASS (24 blocks, loss 5.38->5.37), `docs/TEST_REPORT.md` banaya, HF push OK
|
|
|
|
| 30 |
- [ ] TODO: toy router-balance test (Kaggle), FINAL tokenizer (500k lines wait), dense baseline
|
| 31 |
|
| 32 |
## Kaise note likhna hai
|
|
|
|
| 27 |
- [x] verify: `viu1_moe.py` 241.7M/166.2M OK, `train.py --smoke` 3 steps OK, push dry-run OK (data/raw excluded)
|
| 28 |
- [x] 2026-09-16: local rename `Mini-ViuAI-50M/` -> `ViuMini-MoE-242M/`, README base_model null fix, HF push OK -> https://huggingface.co/ViuAI/ViuMini-MoE-242M
|
| 29 |
- [x] 2026-09-17: Kaggle smoke FAIL (`data/raw` HF me excluded) -> `train.py` smoke fallback (built-in dummy lines), Kaggle re-test PASS (24 blocks, loss 5.38->5.37), `docs/TEST_REPORT.md` banaya, HF push OK
|
| 30 |
+
- [x] 2026-09-17: full code-phase suite 5/5 PASS (params 241.7M/166.2M, sliding unit, toy router balanced max 0.203 + loss 6.18->5.95, tok fertility 1.38 unk 0, train smoke), TEST_REPORT update, HF push
|
| 31 |
- [ ] TODO: toy router-balance test (Kaggle), FINAL tokenizer (500k lines wait), dense baseline
|
| 32 |
|
| 33 |
## Kaise note likhna hai
|
docs/TEST_REPORT.md
CHANGED
|
@@ -35,6 +35,15 @@ Har test ke baad yaha entry hogi: date, env, command, result, verdict.
|
|
| 35 |
- Verdict: PASS (screenshot verified)
|
| 36 |
- HF push: `fix: smoke fallback for Kaggle clone without data/raw` done
|
| 37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
## Template (next test ke liye)
|
| 39 |
```
|
| 40 |
## YYYY-MM-DD β <test naam> (<env>)
|
|
|
|
| 35 |
- Verdict: PASS (screenshot verified)
|
| 36 |
- HF push: `fix: smoke fallback for Kaggle clone without data/raw` done
|
| 37 |
|
| 38 |
+
## 2026-09-17 β Full code-phase suite (local, user-approved)
|
| 39 |
+
- Env: Windows, CPU, torch 2.13, repo `ViuMini-MoE-242M` (post-rename)
|
| 40 |
+
- T1 params: `viu1_moe.py --config viu1_moe_config.yaml` β total 241,668,224 (241.7M) / active 166,158,464 (166.2M), MoE=9 dense=19 sliding=19 full=9, forward loss ~12.11 β PASS
|
| 41 |
+
- T2 sliding: unit check T=8 window=3 β row7=[F,F,F,F,F,T,T,T], har row me past-count=min(i+1,3) β PASS (win+causal combined correct)
|
| 42 |
+
- T3 toy router (2L, dim64, 8 experts): usage=[0.156,0.062,0.180,0.121,0.109,0.070,0.203,0.098], max-share 0.203 (<0.6 collapse-line), aux 0.018, 5-step loss 6.18β5.95 smooth β PASS, no collapse
|
| 43 |
+
- T4 tokenizer coverage (TEST tok 714 vocab): fertility 1.38, unk=0 β PASS (FINAL 48k pending, 500k lines wait)
|
| 44 |
+
- T5 train smoke: blocks=13 train=10 val=3, 3 steps ~6.26-6.28, ckpt/eval OK β PASS
|
| 45 |
+
- Verdict: 5/5 PASS β code phase green, bacha: Kaggle toy repeat + FINAL tokenizer + dense baseline (GPU)
|
| 46 |
+
|
| 47 |
## Template (next test ke liye)
|
| 48 |
```
|
| 49 |
## YYYY-MM-DD β <test naam> (<env>)
|