OpenNMT/CTranslate2

View on GitHub

Fast inference engine for Transformer models

Stars: 4.7k
Forks: 536
Open beginner issues: 0
Indexed issues: 230
Avg merge: 12h 12m
Merged PRs (30d): 4

Dominant language: C++
License: MIT
Last GitHub push: Aug 31, 2026
Latest indexed: Sep 19, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: help wanted

Beginner friendly Everything

Open All

230 open issues indexed

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

Difficulty 5/5 Over a week Newbie friendliness 35/100

OpenNMT/CTranslate2#2101 · Opened Sep 12, 2026
Significant Korean accuracy degradation with Whisper inputs shorter than 30 seconds Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

OpenNMT/CTranslate2#2100 · Opened Sep 11, 2026
Chutki Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

OpenNMT/CTranslate2#2093 · 1 comment · Opened 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 Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

OpenNMT/CTranslate2#2090 · 1 comment · Opened Aug 29, 2026
Host memory grows ~6.8 KB per inference on aarch64 + CUDA (not on x86_64, not on CPU) Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#2088 · 2 comments · Opened Aug 25, 2026
Windows 11 process-exit hang after ROCm Whisper inference with AMD GPU Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

OpenNMT/CTranslate2#2085 · Opened Aug 17, 2026
Source builds need workarounds for sdist, CUDA arches, non-x86 CPU dispatch, and setup.py hardcoding Open

Difficulty 5/5 Over a week Newbie friendliness 38/100

OpenNMT/CTranslate2#2082 · Opened Aug 11, 2026
Failure to compile on 32-bit x86 GNU/Linux Open

Difficulty 3/5 1-2 days Newbie friendliness 58/100

OpenNMT/CTranslate2#2081 · 4 comments · Opened Jul 31, 2026
import ctranslate2 eagerly imports torch (via converters + specs), even for inference-only use Open

Difficulty 3/5 1-2 days Newbie friendliness 74/100

OpenNMT/CTranslate2#2079 · 1 comment · Opened Jul 29, 2026
Allow no-timestamp Whisper decoding beyond 224 tokens Open

Difficulty 2/5 1-3 hours Newbie friendliness 72/100

OpenNMT/CTranslate2#2074 · Opened Jul 11, 2026
Error on using converted model Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#2072 · 5 comments · Opened Jul 5, 2026
4.8.0 regression: WhisperModel.align() fails with parallel_for failed: cudaErrorInvalidDevice on Tesla T4 (works on 4.7.0) Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#2064 · 1 comment · Opened Jun 11, 2026
macOS Python 3.14 CPU backend: intra_threads=0 aborts on 4.7.2 and oversubscribes on 4.8.0 Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

OpenNMT/CTranslate2#2063 · 1 comment · Opened Jun 7, 2026
Expose per-step decoder cross-attention scores during generate() (for streaming / AlignAtt policies) Open

Difficulty 5/5 Over a week Newbie friendliness 45/100

OpenNMT/CTranslate2#2059 · Opened Jun 3, 2026
Add runtime LoRA adapter swapping Open

Difficulty 5/5 Over a week Newbie friendliness 38/100

OpenNMT/CTranslate2#2045 · Opened May 14, 2026
del model deadlocks on ROCm 7.2.1 + gfx1100 (Windows) — HIP allocator free path Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

OpenNMT/CTranslate2#2038 · Opened May 8, 2026
bfloat16 GEMM is noticeably slower than float16 Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#2033 · Opened Apr 29, 2026
Feature request: intel GPU support Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

OpenNMT/CTranslate2#2032 · 7 reactions · Opened Apr 14, 2026
Is it possible to convert CTranslate2 weights back to Hugging Face Transformers? Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

OpenNMT/CTranslate2#2031 · Opened Apr 12, 2026
📋 Documentation Enhancement Suggestion Open

Difficulty 2/5 1-3 hours Newbie friendliness 25/100

OpenNMT/CTranslate2#2026 · Opened Mar 13, 2026
ROCm (gfx1201/RX 9070 XT): faster-whisper crash in CTranslate2 4.7.1 Open

Difficulty 4/5 3-5 days Newbie friendliness 38/100

OpenNMT/CTranslate2#2021 · 5 comments · Opened Feb 28, 2026
Cannot get correct translation results (NLLB-200-distilled-600M) Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

OpenNMT/CTranslate2#2017 · 1 comment · Opened Feb 19, 2026
AMD wheels for Windows built against ROCm 7.2, but Windows HIP SDK installer only goes to 7.1.1 Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#2016 · 3 comments · Opened Feb 14, 2026
[Question] TranslateGemma support? Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

OpenNMT/CTranslate2#2015 · 2 comments · Opened Feb 14, 2026
return_alternatives next-pos tokens don't match top-probability tokens of sampling mode Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

OpenNMT/CTranslate2#2014 · 3 comments · Opened Feb 13, 2026
CTranslate2 4.7.0 Bug on NVIDIA GPU Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

OpenNMT/CTranslate2#2009 · 2 comments · Opened Feb 4, 2026
When it comes to super fast inference Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

OpenNMT/CTranslate2#1995 · 1 comment · 3 reactions · Opened Jan 28, 2026
How to return_attention from Generator.generate_tokens ? Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

OpenNMT/CTranslate2#1994 · Opened Jan 27, 2026
RuntimeError: CUDA failed with error CUDA driver version is insufficient for CUDA runtime version Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

OpenNMT/CTranslate2#1988 · 1 comment · Opened Jan 22, 2026
Clang ThreadSanitizer in CI Open

Difficulty 2/5 1-3 hours Newbie friendliness 35/100

OpenNMT/CTranslate2#1987 · Opened Jan 22, 2026
AWQ and Gemma3 compatibility problem Open

Difficulty 5/5 Over a week Newbie friendliness 28/100

OpenNMT/CTranslate2#1956 · 2 comments · Opened Dec 20, 2025
Making it easier for users to get started with CTranslate2 models Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

OpenNMT/CTranslate2#1950 · 27 comments · 3 reactions · Opened Dec 15, 2025
C++ API Documentation Missing / Insufficient for New Users Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

OpenNMT/CTranslate2#1945 · 2 comments · 1 reaction · Opened Dec 3, 2025
Feature request: CUDA 13 support Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#1933 · 9 comments · 43 reactions · Opened Nov 17, 2025
For Llama 3.1-8b-Instruct, logprobs differ significantly from the transformers' model Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#1921 · Opened Oct 20, 2025
bug: Cannot locate libctranslate2.so.4 when repairing Python wheel Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

OpenNMT/CTranslate2#1919 · 1 comment · Opened Oct 17, 2025
The result for T5 is not the same Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

OpenNMT/CTranslate2#1918 · Opened Oct 2, 2025
Hi, I'm unsure how to implement the ops::Div division operator Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

OpenNMT/CTranslate2#1916 · Opened Sep 9, 2025
Could not locate cudnn_cnn_infer64_8.dll. Please make sure it is in your library path! Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

OpenNMT/CTranslate2#1915 · 1 comment · Opened Aug 27, 2025
Not able to convert the last_hidden_state values when using a bf16 converted model weights Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

OpenNMT/CTranslate2#1910 · 4 comments · Opened Jul 29, 2025
The pkg_resources package is slated for removal as early as 2025-11-30 Open

Difficulty 2/5 1-3 hours Newbie friendliness 35/100

OpenNMT/CTranslate2#1909 · 1 comment · 3 reactions · Opened Jul 24, 2025
run fasterwhisper with ctranslate2 on aarch64 Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

OpenNMT/CTranslate2#1908 · Opened Jul 22, 2025
add build to release Open

Difficulty 4/5 3-5 days Newbie friendliness 38/100

OpenNMT/CTranslate2#1900 · Opened Jul 2, 2025
ONNXRuntime-GPU Fails in ComfyUI – Possible cuDNN Conflict with CTranslate2 Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

OpenNMT/CTranslate2#1897 · Opened Jun 26, 2025
Performance comparison with TensorRT-LLM Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

OpenNMT/CTranslate2#1896 · 1 comment · Opened Jun 11, 2025
CPU Dispatch incompatible with higher CPU baselines Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

OpenNMT/CTranslate2#1895 · 1 comment · Opened Jun 7, 2025
pkg_resources needs to be replaced Open

Difficulty 2/5 1-3 hours Newbie friendliness 35/100

OpenNMT/CTranslate2#1894 · 1 comment · 7 reactions · Opened Jun 2, 2025
Constrained beam search like Fairseq Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

OpenNMT/CTranslate2#1893 · Opened Jun 2, 2025
how to build, so that auditwheel can bundle dependencies Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

OpenNMT/CTranslate2#1892 · 1 comment · Opened May 23, 2025
I can not get expected result following the quick start demo Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

OpenNMT/CTranslate2#1890 · 1 comment · 1 reaction · Opened May 8, 2025

OpenNMT/CTranslate2

Get new issues in your inbox