turboderp-org ExLlamaV2 GPTQ Index q_matrix.cu make_sequential out-of-bounds write
- Published
- Last modified
- Where the record stands: published, rejected by its publisher, or superseded by another id.What do these mean? →
- published
- Whether CISA lists it in its Known Exploited Vulnerabilities catalog: proof that attackers are using it.What do these mean? →
- Not listed in CISA KEV
- Whether working attack code is public (Exploit-DB, Metasploit, Nuclei or GitHub). Public code makes attacks easier.What do these mean? →
- None on Exploit-DB, Metasploit or GitHub
Summary
ExLlamaV2, a Python library for running large language models on NVIDIA GPUs, is affected by this flaw in its GPTQ weight-handling code. A local attacker who can supply crafted input data may be able to write memory outside its intended bounds, which could crash the software or potentially allow code execution.
Summarized by anthropic/claude-haiku-5.5. Verify details against the original sources.
Read the original description
A flaw has been found in turboderp-org ExLlamaV2 up to 0.3.2. The impacted element is the function QMatrix::make_sequential of the file exllamav2/exllamav2_ext/cuda/q_matrix.cu of the component GPTQ Index Handler. Executing a manipulation of the argument g_idx can lead to out-of-bounds write. The attack needs to be launched locally. The project was informed of the problem early through an issue report but has not responded yet.
Timeline
- CVE published
- First seen by cve.now
Affected versions
- turboderp-org ExLlamaV2
0.3.0, 0.3.1, 0.3.2