amd/Qwen2.5-3B-Instruct-onnx-ryzenai-npu warn
Weights only ship in a format that can run code when loaded; its license differs from its base model's; the chat template differs from its base model, which changes behavior. Plus 2 minor notes.
claims base: Qwen/Qwen2.5-3B-Instruct · chat template: present · view on Hugging Face ↗
Scan coverageStatic battery2026-08-26Weights battery2026-08-26Behavioral batterynot rundetails
| Battery | Looks at | Status |
|---|---|---|
| Static battery | Metadata & packaging | complete 2026-08-26 |
| Weights battery | Weights forensics — no GPU, no download | complete 2026-08-26no token vocabulary — pickle audit only · pickle audit clean |
| Behavioral battery | Live-inference differentials | not run |
Ingot runs three batteries against a model. What each one checks →
Findings
Scanned 2026-08-26 · published from a community scan.
medium Pickle-serialized weights, no safetensors
Weights ship only as pickle-based files (prefill.bin, prefill.pb.bin). Loading pickle executes arbitrary code from the file — prefer a safetensors release or load in a sandbox.
How to fix
Convert the weights to safetensors before loading them anywhere that matters.
- Do not load the pickle files in-process — pickle deserialization executes arbitrary code from the file.
- Convert locally in a sandbox: `pip install safetensors` and use `safetensors.torch.save_file` on a state dict loaded with `torch.load(..., weights_only=True)` (refuses most code-execution payloads), or use Hugging Face's `convert.py` space/script.
- Pin the exact revision hash you converted from, and load only your converted safetensors artifact from then on.
low Padding token is the EOS token
The pad token and the (only) EOS token are the same. Fine-tuning frameworks mask pad positions out of the loss, so training on this checkpoint teaches the model to never emit EOS — the Phi-4 / Qwen 2.5 / DeepSeek R1 infinite-generation bug. Safe to serve, hazardous to fine-tune; repoint pad_token at a dedicated token first.
How to fixingot patch
Align the stop-token declarations — a pure metadata fix to `generation_config.json` (and `config.json`).
- Identify the token the chat template actually ends assistant turns with (e.g. `<|eot_id|>`, `<end_of_turn>`, `<|im_end|>`) and make sure its id is in `generation_config.json`'s `eos_token_id` list.
- Keep `config.json`'s `eos_token_id` consistent with (or a subset of) `generation_config.json`'s — runtimes differ in which file they read.
- For the pad-equals-EOS hazard: repoint `pad_token` at a dedicated padding token before fine-tuning; serving is unaffected.
- Until the repo is fixed, pass explicit stop tokens to your serving stack (e.g. vLLM `stop_token_ids`, llama.cpp `--override-kv tokenizer.ggml.eos_token_id`).
medium License differs from claimed parent (mit vs other)
This model declares mit while its claimed base Qwen/Qwen2.5-3B-Instruct declares other. Verify the re-license is permitted before commercial use.
How to fix
Verify the re-license is actually permitted before relying on it.
- Read the parent's license for derivative-work and re-licensing terms — many open-weight licenses (e.g. Llama-family) do not permit arbitrary re-licensing.
- If the re-license is not permitted, the parent's terms govern your use regardless of what this repo declares.
medium Chat template differs from claimed parent
The chat template does not match Qwen/Qwen2.5-3B-Instruct's. Template drift silently changes model behavior even when weights are identical — 37% of drifted derivatives in our census left it undisclosed. Diff the templates before deploying.
How to fixingot patch
Restore the parent's chat template in `tokenizer_config.json` — a pure metadata fix.
- Run `ingot patch <owner/model>` — the patch manifest carries the parent's template and applies it to a local copy's `tokenizer_config.json`.
- Or fix by hand: copy the `chat_template` value from the parent repo's `tokenizer_config.json` into this model's, and pin your serving stack to that file.
- If the drift was intentional (the author retrained on a new template), confirm that in the model card before "fixing" it — restoring the parent template on retrained weights changes behavior too.
info Pickle static analysis clean
Opcode-level parse of prefill.bin, prefill.pb.bin (no code executed) found only standard serialization globals (0 distinct, all torch/collections/numpy). Pickle remains an executable format — this verifies the current bytes, not future uploads; prefer a safetensors release.
low Pickle checkpoint only partially analyzable
Static analysis could not fully parse: prefill.bin: legacy parse stopped after 0 pickle(s): unsupported pickle opcode 0xa6 at 0, prefill.pb.bin: legacy parse stopped after 0 pickle(s): unsupported pickle opcode 0xa at 0. Unparsed content is unverified.
Fix it
Some findings are metadata-level and patchable — apply the fixes to your local copy (your weights never leave your machine):
npx @ingotai/scan patch amd/Qwen2.5-3B-Instruct-onnx-ryzenai-npu
Remediation guidance addresses the documented findings only. It is evidence-driven repair, not a safety certification of the model.
Fingerprint
The durable profile of this model: measured weights-and-metadata facts, rebuilt on every scan and battery run. Updated 2026-08-26.
| license | mit |
| serialization | no safetensors pickle |
| chat template | present (chat_template.jinja) · sha256:8aa40ce145adb73c |
| claimed lineage | Qwen/Qwen2.5-3B-Instruct |
| lineage verified | unverified — weights battery pending |
Full measured fingerprint
| pipeline | text-generation |
| repo files | 21 — pickle: prefill.bin, prefill.pb.bin |
| revision | 94f88326aa8c |
| HF snapshot | 156 downloads · 2 likes · updated 2025-10-23 · captured 2026-08-25 |
| weights battery | token-embedding scan n/a — no token-embedding tensor found in the pickle checkpoint(s) — pickle checkpoint(s) statically analyzed anyway |
| pickle audit | prefill.bin, prefill.pb.bin — 0 standard global(s) · legacy (pre-1.6) format, head-scan only |
Battery runs (1)the run trace behind the findings — what each job measured
| battery | status | queued | duration | attempts |
|---|---|---|---|---|
| weights | complete | 2026-08-25 21:51 | 3s | 1 |
weights run 2026-08-25 — measurements
| probes run | pickle-static-analysis |
Verdict badge
Ship the verdict in your README — it always shows the latest published analysis:
[](https://ingot.tools/models/amd/Qwen2.5-3B-Instruct-onnx-ryzenai-npu)