Cool. Running it on my windows laptop. Installation was smooth, took me a few trials to figure out there are many dolphins in the sea and DevQuasar is not the only one from Venice... The command to run remains "ollama run hf.co/DevQuasar/dphn.Dolphin-Mistral-24B-Venice-Edition-GGUF" (including hf.co, also after installing).
Would be healthy for people to understand the required power / hardware to run all their braindraining prompts, as the difference between my Claude environment and local is huge. My PC is not under spec, but I'm afraid I'm a little too spoiled to give up the output speed of cloud based AI.
Some ideas I get from this (probably already exist):
- Set up ollama + several models in a cloud environment (VPS / idk?) and would that be competitive with my Claude subscription + experience?
- Do something similar, but on a more expensive plan and share it like a Plex environment with friends (or collegues).
Question:
- I have not spent much time on HF yet. How would you advice to explore this platform? It seems like an endless library that will keep on growing and filtering is limited. Just filter on best liked? Or..?
Thanks for the update on windows running, I appreciate it!
- The performance is indeed limited by hardware, there is a reason a datacenter-level GPU costs tens of thousands
- Ollama actually has a local API you could use with something like openclaw
- I actually thought about this idea of running locally and sharing with friends, not sure if there is a good option for it
To your question: HF is not a good explorer unless you know what you are looking for. You can use the filtering tO find GGUF models wit certain sizes etc. They do have a leaderboard, but you are better off using perplexity combined with benchmarking websites
Cool. Running it on my windows laptop. Installation was smooth, took me a few trials to figure out there are many dolphins in the sea and DevQuasar is not the only one from Venice... The command to run remains "ollama run hf.co/DevQuasar/dphn.Dolphin-Mistral-24B-Venice-Edition-GGUF" (including hf.co, also after installing).
Would be healthy for people to understand the required power / hardware to run all their braindraining prompts, as the difference between my Claude environment and local is huge. My PC is not under spec, but I'm afraid I'm a little too spoiled to give up the output speed of cloud based AI.
Some ideas I get from this (probably already exist):
- Set up ollama + several models in a cloud environment (VPS / idk?) and would that be competitive with my Claude subscription + experience?
- Do something similar, but on a more expensive plan and share it like a Plex environment with friends (or collegues).
Question:
- I have not spent much time on HF yet. How would you advice to explore this platform? It seems like an endless library that will keep on growing and filtering is limited. Just filter on best liked? Or..?
Looking forward to your next one :)
Thanks for the update on windows running, I appreciate it!
- The performance is indeed limited by hardware, there is a reason a datacenter-level GPU costs tens of thousands
- Ollama actually has a local API you could use with something like openclaw
- I actually thought about this idea of running locally and sharing with friends, not sure if there is a good option for it
To your question: HF is not a good explorer unless you know what you are looking for. You can use the filtering tO find GGUF models wit certain sizes etc. They do have a leaderboard, but you are better off using perplexity combined with benchmarking websites