Views
No views yet
hotdogs/Qwen3.8-27B-abliterated, trained on the v2 dataset that fixes the template-collapse problem of v1.reasoning_content when served) and then answers.v1 → v2: v1 was trained on a synthetic placeholder dataset (15 unique code bodies, 29–44 char answers like## Review\n\nFound N issue(s) in L lines.). The model faithfully reproduced the template — it answered "No bugs found. Code is clean." and missed real bugs. v2 was retrained on 21,009 real code+bug+answer rows across 5 languages with 550–880 char detailed answers. The model now actually finds the bugs.
fetch not checking res.ok, async races, etc.reasoning_content, user sees only the answermtp.* / blk.64.nextn.*) kept for speculative decoding| Step | Detail |
|---|---|
| Base | hotdogs/Qwen3.8-27B-abliterated (abliterated, ~27B) |
| Method | Unsloth LoRA, r=32, 233M trainable params (0.85%) |
| Dataset | hotdogs/code-analysis-sft-qwen38-v2 — 21,009 train / 1,900 valid |
| Languages | Python, JavaScript, Go, Rust, C |
| Answer style | 550–880 chars, line numbers, severity, fix code block |
| Sequence | max 2048 tokens, bf16, 5× RTX 3090 |
| Early stop | step 400 / 1313 (epoch ~0.30, loss ~0.0003) — stopped before the 15 archetypes were memorized to death |
| Merge | merge_and_unload, MTP 15 tensors recovered, no triple-nest |
| Case | Result |
|---|---|
| Off-by-one (in-archetype) | 🟢 Found it + fix + docstring note |
| Async race (unseen) | 🟢 "no cache-hit fast path" + concurrency |
| Clean code (hallucination test) | 🟢 "correct, no bugs" + minor float/bool note |
1from transformers import AutoModelForImageTextToText, AutoTokenizer
2import torch
3
4MODEL = "hotdogs/Qwen3.8-27B-abliterated-code-analysis-preview"
5tok = AutoTokenizer.from_pretrained(MODEL, trust_remote_code=True)
6model = AutoModelForImageTextToText.from_pretrained(
7 MODEL, torch_dtype=torch.bfloat16,
8 device_map="auto", trust_remote_code=True, attn_implementation="sdpa")
9model.eval()
10
11def review(code, max_new=600):
12 text = tok.apply_chat_template(
13 [{"role": "user", "content": "Review this code and report any bugs you find.\n\n```python\n" + code + "\n```"}],
14 tokenize=False, add_generation_prompt=True)
15 inputs = tok(text, return_tensors="pt").to(model.device)
16 with torch.no_grad():
17 out = model.generate(input_ids=inputs["input_ids"],
18 attention_mask=inputs["attention_mask"],
19 max_new_tokens=max_new, do_sample=False,
20 repetition_penalty=1.05)
21 new = out[0][inputs["input_ids"].shape[1]:]
22 return tok.decode(new, skip_special_tokens=True)hotdogs/Qwen3.8-27B-abliterated-code-analysis-preview-mtp-GGUF