Search results

can this project run deepseek V4 flash on an 8G VRam?

Status: Open.

#299 In lyogavin/airllm;

· aibfarm opened on Jun 22, 2026

Proposal: Layer Prefetching with Double Buffering to Reduce I/O Latency During Inference

Status: Open.

#298 In lyogavin/airllm;

· arthurg-santos opened on Jun 21, 2026

delete_original can delete a shard and then crash

Status: Open.

#297 In lyogavin/airllm;

· tg12 opened on Jun 21, 2026

Model splitter unpickles remote .bin shards by default

Status: Open.

#296 In lyogavin/airllm;

· tg12 opened on Jun 21, 2026

Headline VRAM and performance claims lack reproducible verification

Status: Open.

#295 In lyogavin/airllm;

· tg12 opened on Jun 21, 2026

setup.py runs a nested pip upgrade during install

Status: Open.

#294 In lyogavin/airllm;

· tg12 opened on Jun 21, 2026

AutoModel enables Hugging Face remote code execution by default

Status: Open.

#293 In lyogavin/airllm;

· tg12 opened on Jun 21, 2026

Error: ValueError: Module does not have parameter named "shared_expert".

Status: Open.

#290 In lyogavin/airllm;

· umangpandey opened on Jun 3, 2026

If you can't install and still wanna use it on Windows 10+

Status: Open.

#289 In lyogavin/airllm;

· maxartemev163-ai opened on May 30, 2026

web

Status: Open.

#287 In lyogavin/airllm;

· alaaalhaydari205-oss opened on May 10, 2026

Any plan to support for qwen3.6-35b-a3b?

Status: Open.

#285 In lyogavin/airllm;

· Sblash opened on May 6, 2026

I found the perfect use case for airllm

Status: Open.

#284 In lyogavin/airllm;

· thiswillbeyourgithub opened on May 6, 2026