Model page

lu-vae/llama-68m-fft warn

Weights only ship in a format that can run code when loaded; EOS ids disjoint between config.json and generation_config.json; its tokenizer differs from its claimed base model.

downloads 513likes 0license apache-2.0arch llamaupdated 2024-05-02

claims base: JackFram/llama-68m · chat template: not found · view on Hugging Face ↗

Scan coverageStatic battery2026-08-25Weights battery2026-08-25Behavioral batterynot rundetails
BatteryLooks atStatus
Static batteryMetadata & packagingcomplete 2026-08-25
Weights batteryWeights forensics — no GPU, no downloadcomplete 2026-08-2532,003-token embedding scanned · 0 undertrained · lineage consistent · pickle audit clean
Behavioral batteryLive-inference differentialsnot run

Ingot runs three batteries against a model. What each one checks →

Findings

Scanned 2026-08-25 · published from a community scan.

medium Pickle-serialized weights, no safetensors

Weights ship only as pickle-based files (pytorch_model.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.

  1. Do not load the pickle files in-process — pickle deserialization executes arbitrary code from the file.
  2. 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.
  3. Pin the exact revision hash you converted from, and load only your converted safetensors artifact from then on.

medium EOS ids disjoint between config.json and generation_config.json

config.json declares eos_token_id [32000] while generation_config.json declares [2] with no overlap. Runtimes read one or the other, so at least one of them stops generation on the wrong token (or never). Align both files on the token the chat template actually ends turns with.

How to fixingot patch

Align the stop-token declarations — a pure metadata fix to `generation_config.json` (and `config.json`).

  1. 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.
  2. Keep `config.json`'s `eos_token_id` consistent with (or a subset of) `generation_config.json`'s — runtimes differ in which file they read.
  3. For the pad-equals-EOS hazard: repoint `pad_token` at a dedicated padding token before fine-tuning; serving is unaffected.
  4. 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 Vocabulary size differs from claimed parent (32003 vs 32000)

A changed vocab means changed tokenization: strings will split differently than on JackFram/llama-68m, which can shift behavior on identifiers, codes, and non-English text.

How to fixweight-level

Not patchable: the vocab size mirrors the embedding matrix in the weights. Verify the change was intentional.

  1. Do not edit `vocab_size` in config.json to "match the parent" — it must equal the embedding table in the shipped weights or the model won't load.
  2. Diff the tokenizers (`tokenizer.json` / added_tokens) against the parent to see what was added or removed, and test your own identifiers, codes, and non-English text through both.
  3. If the drift is unexplained by the model card, treat tokenization-sensitive behavior as unvalidated on this model.

info Embedding-norm glitch scan clean

No undertrained tokens found: every non-special token's embedding norm is above 0.3× the vocabulary median (1.263). The glitch-token data-corruption class has no candidate surface in this model.

info Pickle static analysis clean

Opcode-level parse of pytorch_model.bin (no code executed) found only standard serialization globals (3 distinct, all torch/collections/numpy). Pickle remains an executable format — this verifies the current bytes, not future uploads; prefer a safetensors release.

info Weights consistent with claimed parent JackFram/llama-68m

Mean cosine similarity of 64 sampled token-embedding rows against JackFram/llama-68m is 0.999 — the weights plausibly descend from the declared base (relation: unspecified).

How to fix

Fix or verify the `base_model` declaration so lineage checks can run.

  1. If you own the repo: correct the `base_model` field in the model card metadata to the real, public parent.
  2. If you don't: identify the true parent (config architecture + weight shapes narrow it fast) and re-scan with that lineage in mind.

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 lu-vae/llama-68m-fft

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 2026-08-25.

architecturellama · 2 layers · 768-dim
vocabulary32,003 tokens
licenseapache-2.0
serializationno safetensors pickle
chat templatenone
claimed lineageJackFram/llama-68m
lineage verifiedconsistent vs JackFram/llama-68m — embedding-row cosine 0.999
glitch-token surfaceclean no undertrained tokens
Full measured fingerprint
architecturesLlamaForCausalLM
librarytransformers
pipelinetext-generation
repo files12 — pickle: pytorch_model.bin
revision24303d7edc3e
HF snapshot513 downloads · 0 likes · updated 2024-05-02 · captured 2026-08-25
pickle auditpytorch_model.bin3 standard global(s)
embedding tensormodel.embed_tokens.weight · BF16 · 32,003×768
embedding normsmedian 1.2631 · mean 1.2063
lineage checkconsistent — cosine 0.9992 over 64 sampled rows vs JackFram/llama-68m
Battery runs (1)the run trace behind the findings — what each job measured
batterystatusqueueddurationattempts
weightscomplete2026-08-25 21:4831s1
weights run 2026-08-25 measurements
probes runglitch-norm-scan, zero-template-token-scan, pickle-static-analysis, lineage-norm-correlation
embedding tensormodel.embed_tokens.weight · BF16 · 32,003×768
glitch surface0 undertrained, 0 plain-ASCII
lineage checkconsistent — cosine 0.9992 over 64 rows vs JackFram/llama-68m

Verdict badge

Ship the verdict in your README — it always shows the latest published analysis:

Ingot verdict: warn

[![Ingot scan](https://ingot.tools/api/v1/models/lu-vae/llama-68m-fft/badge.svg)](https://ingot.tools/models/lu-vae/llama-68m-fft)
Gate it in CI