medicis123

6 ms·

1.
by medicis123 2mo ago
We did something similar - Streaming experts. Maintaining an expert cache, optimizing it to simulate running a multi-model agentic workflow on a 2-DGC Spark Cluster. The models we ran were: DeepSeek V4 Flash, Gemma 4 26B A4B, and Nemotron 3
2.
New Inference Server for DGX Spark: large model C4:55-90 tok/s no spec decode
4 points by medicis123 2mo ago|0 comments
3.
Show HN: Stop GPU pods placement getting bottlenecked by reserved VRAM
2 points by medicis123 7mo ago|0 comments
4.
A New Approach to GPU Sharing: Deterministic, SLA-Based GPU Kernel Scheduling
1 points by medicis123 10mo ago|0 comments
5.
Show HN: Disaggregating GPU compute from CPU in ML job execution to scale GPUs(woolyai.com)
1 points by medicis123 10mo ago|0 comments
6.
Show HN: Run PyTorch on CPU boxes, offload kernels to remote GPUs
1 points by medicis123 11mo ago|0 comments
7.
Running Nvidia CUDA PyTorch container project/pipelines on AMD with no changes
1 points by medicis123 1y ago|0 comments
8.
GPU-accelerated code on CPU-only environments -Remote GPU Kernel Execution(youtube.com)
1 points by medicis123 1y ago|1 comments
9.
by medicis123 1y ago
We built this feature in our GPU Hypervisor, which cleanly separates your user-space ML environment from the GPU runtime, so you can code locally, run remotely, and execute kernels remotely on GPU hosts. Would love to hear how this impacts
10.
Sharing base model in GPU VRAM across multiple inference stack process [video](youtube.com)
7 points by medicis123 1y ago|1 comments
11.
by medicis123 1y ago
We have just published a short demo of the WoolyAI GPU Hypervisor, showcasing VRAM memory sharing/deduplication. Load a single base model once, then run multiple isolated LoRA stacks or VLLM stacks on the same GPU. Why this matters Hig
12.
by medicis123 2y ago
Looks interesting. Will give it a try. I have been following AI startups developing various agents(Agentic phenomenon). Are you using OPenAi and any other LLM service APIs or you fine-tuned an open source model?
13.
Sharing actual GPU core and VRAM utilization metrics for query on 10 LLM models(woolyai.com)
1 points by medicis123 2y ago|1 comments
14.
by medicis123 2y ago
We ran it on WoolyAI Acceleration Service https://docs.woolyai.com/getting-started/running-your-first-... There are some interesting insights just looking at these numbers. Environment Details Wooly Client: Linux non-G
15.
Show HN: WoolyAI-CUDA Abstraction Layer to Decouple Kernel Shader Exec on GPU(woolyai.com)
4 points by medicis123 2y ago|0 comments
16.
Locally delivered and centrally managed macOS envs for privileged access setup(veertu.com)
1 points by medicis123 7y ago|0 comments
17.
by medicis123 8y ago
Wanted to share that we developed a GitLab CI Runner/executor for Anka Build and have made it public (It was done for one of our user). It basically enables you to run your iOS/macOS GitLab CI pipelines/jobs on Anka build mac
18.
Shopify scaling iOS CI with Anka(engineering.shopify.com)
1 points by medicis123 8y ago|0 comments