ROCm/FastFlowLM

View on GitHub

Run LLMs on AMD Ryzen™ AI NPUs in minutes; purpose-built and deeply optimized for the AMD NPUs.

Stars: 1.9k
Forks: 152
Open beginner issues: 0
Indexed issues: 152
Avg merge: 4h 14m
Merged PRs (30d): 11

Dominant language: C++
License: MIT
Last GitHub push: Sep 18, 2026
Latest indexed: Sep 20, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

152 open issues indexed

[Bug, Issue]: ASR model (whisper-large-v3-turbo) not explicitly named in v1/models, so doesn't work with stricter openAI compatible connections Open

Difficulty 3/5 1-2 days Newbie friendliness 65/100

ROCm/FastFlowLM#746 · 1 comment · Opened Sep 19, 2026
[Feature]: Add support for Cactus Compute Needle model Open

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

ROCm/FastFlowLM#745 · Opened Sep 18, 2026
No valid checkpoint to restore + Max length reached on multi-turn tool calls (Qwen3.6-MoE, NPU) Open

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

ROCm/FastFlowLM#744 · 1 comment · Opened Sep 18, 2026
Truncated tool call is emitted as a complete one (unterminated heredoc reaches the client) Open

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

ROCm/FastFlowLM#741 · 1 comment · Opened Sep 18, 2026
Non-streaming responses leak <think> tags into content for qwen3.5, qwen3.6-moe, qwen3vl and qwen3.5-omni Open

Difficulty 3/5 1-2 days Newbie friendliness 70/100

ROCm/FastFlowLM#739 · 1 comment · Opened Sep 18, 2026
[Feature]: KV-cache reuse for shared-prefix one-shot requests (prefix caching / partial truncate) Open

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

ROCm/FastFlowLM#737 · Opened Sep 17, 2026
Prompt cache: identical resend never hits, partial match discards everything, and nothing is sent during prefill Open

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

ROCm/FastFlowLM#733 · 2 comments · Opened Sep 17, 2026
v1.0.4 regression: embed-gemma:300m embeddings via --embed 1 are wrong (uncorrelated with v0.9.39 on identical weights) Open

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

ROCm/FastFlowLM#732 · Opened Sep 17, 2026
[Feature/Model Request]: Support for NPU-accelerated Text-to-Speech (TTS) models Open

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

ROCm/FastFlowLM#731 · Opened Sep 17, 2026
[Issue]: `usage.prompt_tokens` on the OpenAI-compatible endpoints reports only the tokens prefilled on the current turn Open

Difficulty 2/5 1-3 hours Newbie friendliness 75/100

ROCm/FastFlowLM#730 · Opened Sep 17, 2026
[Documentation]: How to generate SRT/VTT transcripts Open

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

ROCm/FastFlowLM#726 · Opened Sep 13, 2026
[Issue]: [Bug][Linux] flm serve crash: C++ assertion failure '!this->empty()' in std::vector<logits_t>::back() Open

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

ROCm/FastFlowLM#725 · Opened Sep 12, 2026
[Issue]: Gemma4 tool-call parsing produces wrong name and arguments (v1.0.5) Open

Difficulty 3/5 1-2 days Newbie friendliness 66/100

ROCm/FastFlowLM#722 · 3 comments · Opened Sep 10, 2026
Flm with gemma 4 has very poor ability to call tools Open

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

ROCm/FastFlowLM#721 · 2 comments · 1 reaction · Opened Sep 10, 2026
Inconsistent memory consumption? Open

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

ROCm/FastFlowLM#720 · Opened Sep 10, 2026
[Issue]: JSON literal data types instead of strings Open

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

ROCm/FastFlowLM#719 · Opened Sep 10, 2026
The directory for models is not configurable in the new MSI installer Open

Difficulty 3/5 1-2 days Newbie friendliness 55/100

ROCm/FastFlowLM#718 · Opened Sep 10, 2026
Server silently serves the resident model for unresolvable model tags, echoing the requested name Open

Difficulty 4/5 3-5 days Newbie friendliness 64/100

ROCm/FastFlowLM#716 · 1 comment · Opened Sep 9, 2026
[Feature]: Support MiniCPM5-2B top-ranked open 2B model Open

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

ROCm/FastFlowLM#715 · Opened Sep 8, 2026
Qwen3-8B on XDNA2 (Strix Halo): only ~10 tok/s decode + slow TTFT vs GPU - is this expected? Open

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

ROCm/FastFlowLM#713 · Opened Sep 8, 2026
[Model Support]: Unofficial MiniCPM5-2B Q4NX pack for XDNA 2 Open

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

ROCm/FastFlowLM#712 · 11 comments · Opened Sep 8, 2026
[Feature]: Publish signed binaries Open

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

ROCm/FastFlowLM#711 · Opened Sep 7, 2026
Qwen3.6-35B-A3B Linux v1.0.4: ~2.7s fixed per-request TTFT = ~4,900 serialized per-expert NPU execs — validated coarsening fix (concatenated-expert GEMM) Open

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

ROCm/FastFlowLM#710 · 2 reactions · Opened Sep 6, 2026
[Linux] Weight bos are MAP_LOCKED — models >~18GB cannot load on 22.6GB RAM machines and some request Open

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

ROCm/FastFlowLM#709 · Opened Sep 5, 2026
[Issue]: v1.0.4 causes NPU hardware crash (0xc00002b6) and Kernel Deadlock on AMD Ryzen AI (v1.0.3 works fine) Open

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

ROCm/FastFlowLM#708 · 1 comment · Opened Sep 5, 2026
[Feature]: Trivial/Low-Priority/Unimportant: F.L.M. Mel & Kim Open

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

ROCm/FastFlowLM#707 · 1 reaction · Opened Sep 4, 2026
[Issue]: Qwen3.5 / Qwen3.6 family fails to load on Strix Halo (Ryzen AI MAX+ 395, NPU2) — CREATE_HWCTX err=-22, while all other models work Open

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

ROCm/FastFlowLM#705 · 3 comments · Opened Sep 3, 2026
Support request: Bielik-11B-v3.0-Instruct-heretic-MPOA on Ryzen AI NPU Open

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

ROCm/FastFlowLM#704 · Opened Sep 3, 2026
[Bug] Whisper transcription truncated after ~16 tokens (any clip with >~4s of continuous speech) Open

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

ROCm/FastFlowLM#698 · 1 comment · Opened Sep 1, 2026
[Issue]: Todos status argument missing in Tool calling Open

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

ROCm/FastFlowLM#697 · 2 comments · Opened Sep 1, 2026
[Feature]: Support for iFlytek's Newly Open-Sourced Edge-Side Models (Spark X2.5-4B & 1.7B) Open

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

ROCm/FastFlowLM#696 · 1 comment · 1 reaction · Opened Sep 1, 2026
[Issue]: Whisper periodic forced timestamp sampling truncates Japanese sentence endings Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

ROCm/FastFlowLM#694 · 1 comment · Opened Aug 31, 2026
[Feature]: Qwen3.8-Flash-Next Open

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

ROCm/FastFlowLM#692 · 3 comments · 8 reactions · Opened Aug 28, 2026
[Feature]: Model Request: JetBrains Mellum2‑12B‑A2.5B‑Thinking & Mellum2‑12B‑A2.5B‑Instruct (MoE) Open

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

ROCm/FastFlowLM#691 · 3 reactions · Opened Aug 28, 2026
[ Request/Plea ] Let the Open Source Community Support You Open

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

ROCm/FastFlowLM#690 · 38 comments · 7 reactions · Opened Aug 27, 2026
v1.0.2 crashes (0xC0000005) loading qwen3.6-moe:35b-a3b weights downloaded with v0.9.45 — `flm pull` doesn't detect outdated weights Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

ROCm/FastFlowLM#685 · 1 comment · Opened Aug 25, 2026
gpt-oss:20b never emits tool_calls even with valid OpenAI-style tools schema + tool_choice:auto Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

ROCm/FastFlowLM#684 · 1 reaction · Opened Aug 24, 2026
[Model request]: Meta Muse Glimmer 30B Open

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

ROCm/FastFlowLM#683 · 1 comment · 1 reaction · Opened Aug 24, 2026
Server stops accepting connections after 10 client aborts — connection slot leak ("Connection limit reached (10)") Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

ROCm/FastFlowLM#680 · 1 comment · 1 reaction · Opened Aug 20, 2026
FLM 1.0.2 on Linux: Qwen3.5 4B and 9B hang during prefill, while Llama 3.2 and Qwen3.6 work Open

Difficulty 4/5 3-5 days Newbie friendliness 42/100

ROCm/FastFlowLM#678 · 2 comments · 1 reaction · Opened Aug 19, 2026
Qwen3.6-35B-A3B emits garbage tokens and gpt-oss-20b crashes on Gorgon Point (Ryzen AI 9 HX 470); HRX source build crashes in qwen3_npu.dll Open

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

ROCm/FastFlowLM#677 · 7 comments · 3 reactions · Opened Aug 19, 2026
[Feature]: New Python wrapper Open

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

ROCm/FastFlowLM#676 · 1 reaction · Opened Aug 19, 2026
[Feature]: Qwen3.8-27B please Open

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

ROCm/FastFlowLM#673 · 4 comments · 12 reactions · Opened Aug 18, 2026
/api/show returns stubs — populate model_info context_length + details so Ollama-compatible clients can auto-configure Open

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

ROCm/FastFlowLM#671 · 3 comments · 1 reaction · Opened Aug 15, 2026
flm-real segfaults on NPU firmware timeout when GPU is concurrently active via ROCm (Strix Halo) — companion to amd/xdna-driver#1605 Open

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

ROCm/FastFlowLM#670 · 1 comment · Opened Aug 14, 2026
[ GUIDE **WILL NEVER WORK ABOVE 1.0.1** ] - Converting and adding finetunes (unofficial but functional) Open

Difficulty 3/5 1-2 days Newbie friendliness 38/100

ROCm/FastFlowLM#669 · 1 comment · 1 reaction · Opened Aug 14, 2026
Apply a "chat" label to LLMs Open

Difficulty 3/5 1-2 days Newbie friendliness 52/100

ROCm/FastFlowLM#668 · Opened Aug 13, 2026
embed-gemma:300m returns a collapsed embedding space: deterministic, but unusable for retrieval (same model via llama.cpp is correct) Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

ROCm/FastFlowLM#661 · 13 comments · Opened Aug 10, 2026
IOCTL regression using ASR and embedding models simultaneously (Krackan Point) Open

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

ROCm/FastFlowLM#655 · 7 comments · Opened Aug 6, 2026
Nanbeige4.2-3B please Open

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

ROCm/FastFlowLM#653 · 1 comment · Opened Aug 6, 2026

ROCm/FastFlowLM

Get new issues in your inbox