thefahien / elpis-on-7900xtx
Qwen3.8-27B EXL3 + DFlash2 speculative decoding on an AMD RX 7900 XTX (exllamav3-rocm): recipe, patches, benchmarks
Why this job: matched by what the repo talks about: benchmark, speculative, qwen3, patche
What it does
This repo runs the "elpis" model pair on an AMD RX 7900 XTX. The pair is Qwen3.8-27B as EXL3 4.00bpw plus a DFlash2 speculative draft, both by r0b0tlab. It uses phoenixhaxor/exllamav3-rocm and its patched TabbyAPI. The repo contains the recipe, two small patches, the benchmark scripts and the raw numbers.
Get started
No single install command. Read the steps in the README (opens in new tab)
- Maintenance
- Active
- Last commit 11 days ago
- Stars
- 23
- +2 this week
- Licence
- MIT
- Allows business use
- Downloads for
- None listed
- No release files
- Latest version
- None yet
- No releases
- Built with
- Python
- 79% of the code
Is it looked after?
- Contributors
- 1
- Last 90 days
- Issues
- None
- No issues filed
- Pull requests
- 0 merged
- 0 waiting
- Releases
- None
- Never published
- No contributing guide
- No security policy
- No code of conduct
More: languages, topics and project record
Languages
- Python 79%
- Shell 21%
Topics
- amd
- exllamav3
- llm-inference
- qwen
- rocm
- rx-7900-xtx
- speculative-decoding
Project record
- JobEvals and observability
- StatusNew to our list
- Commits6
- Forks0
- Watching0
- CreatedNov 1, 2026
- First seen by usNov 7, 2026
- CheckedNov 12, 2026