All issues

Search results

ROCm/HIP Windows wheel: Translator destruction hangs indefinitely, even in CPU-only mode with no AMD GPU present

Status: Open.

#2101 In OpenNMT/CTranslate2;

· Nachoco000 opened on Sep 12, 2026

Significant Korean accuracy degradation with Whisper inputs shorter than 30 seconds

Status: Open.

#2100 In OpenNMT/CTranslate2;

· feel012-web opened on Sep 11, 2026

Chutki

Status: Open.

#2093 In OpenNMT/CTranslate2;

· kumarkunal18128-debug opened on Aug 30, 2026

HIP: hipMallocAsync silently corrupts buffers on gfx1030 (RDNA2) although the device reports memory-pool support; default allocator should be cub_caching on HIP

Status: Open.

#2090 In OpenNMT/CTranslate2;

· Krzysztofdaw opened on Aug 29, 2026

Host memory grows ~6.8 KB per inference on aarch64 + CUDA (not on x86_64, not on CPU)

Status: Open.

#2088 In OpenNMT/CTranslate2;

· paruparu opened on Aug 25, 2026

Windows 11 process-exit hang after ROCm Whisper inference with AMD GPU

Status: Open.

#2085 In OpenNMT/CTranslate2;

· Sethx86 opened on Aug 17, 2026

Source builds need workarounds for sdist, CUDA arches, non-x86 CPU dispatch, and setup.py hardcoding

Status: Open.

#2082 In OpenNMT/CTranslate2;

· RemiArata opened on Aug 11, 2026

Failure to compile on 32-bit x86 GNU/Linux

Status: Open.

#2081 In OpenNMT/CTranslate2;

· robertkirkman opened on Jul 31, 2026

import ctranslate2 eagerly imports torch (via converters + specs), even for inference-only use

Status: Open.

#2079 In OpenNMT/CTranslate2;

· vannhph53095 opened on Jul 29, 2026

Allow no-timestamp Whisper decoding beyond 224 tokens

Status: Open.

#2074 In OpenNMT/CTranslate2;

· Emre-Akgul opened on Jul 11, 2026

Error on using converted model

Status: Open.

#2072 In OpenNMT/CTranslate2;

· WhoMI7 opened on Jul 5, 2026

4.8.0 regression: WhisperModel.align() fails with parallel_for failed: cudaErrorInvalidDevice on Tesla T4 (works on 4.7.0)

Status: Open.

#2064 In OpenNMT/CTranslate2;

· christopher5106 opened on Jun 11, 2026