ac35738978
Announce phi3 exllamav2 dev variant
3 days ago
1fdefb70de
Remove deprecated information
1 week ago
2f8b2fe4bb
Update deployment parameters
...
- Remove compose `version`
- Update folder structure with
- `cache`
- `instruction-templates`
- Add examples for ROCM/AMD deployment
- Add examples for custom embedding configurations
1 week ago
afeeb9e8ae
Fix launch arg array handling
1 month ago
6726a71bf6
Bump torch versions
...
Closes #44
2 months ago
994e701bf2
Rephrase ROCM and Arc variants support
2 months ago
6f2d496213
Delete `arc` and `rocm` nightlies
2 months ago
02ebf86b5f
Add ROCM test results to README.md
2 months ago
083d958ced
Create LICENSE
...
AGPL-3.0 license, matching the containerised project.
3 months ago
58392b249b
Update docs for refactored variants
3 months ago
d21346e3f4
Nightly workflows refactor ( #39 )
...
* Create default-nightly.yml
* Nightly variants
* Schedule nightly builds
* Disable arc nightly, More space required on builder
* Disable rocm nightly, More disk space required on builder
3 months ago
f971ca054d
Remove `unstable``tags from rocm and arc
3 months ago
76742f1fbd
Disable nightly builds
3 months ago
2a1dbbf43f
Update docs
3 months ago
0b6b7bc523
Update compose
3 months ago
b38cb4c9f9
Refactor Dockerfile and variants
3 months ago
e6f0ec9837
Bump CUDA version to 12.1
...
Fixes #31
7 months ago
d4b58daffe
Separate nightly builds
7 months ago
89aa0183ab
Improve documentation
...
- Add Quick-Start section
- Expand Usage description
- Signpost Kubernetes issue
- Deprecate `monkey-patch`
7 months ago
0575195a24
Pin torch to CUDA 11.8
7 months ago
123618d6bf
Remove hotfix for ExLlamaV2
...
Issue is resolved upstream
8 months ago
e704efd2fa
Hotfix for ExLlamaV2
...
See:
https://github.com/oobabooga/text-generation-webui/issues/4002
and
ec5164b8a8
8 months ago
13c7bad5cd
Update README.md
8 months ago
8f7d865b5e
README.md spelling ( #24 )
8 months ago
faab710cbc
Add Exllamav2 to base image
8 months ago
e1d999f0b0
Document dated milestone pseudo-version releases
8 months ago
5ff0a9571c
Delete docker-nightly.yml
9 months ago
ffd25bc4e1
Update README.md about versions and nightlies
9 months ago
65897fd7c1
Implement nightly builds of all variants
9 months ago
43392402e0
Deprecate `llama-cublas` variant
...
`default` already includes CUDA GPU offloading for llama
9 months ago
f85bf702ea
Implement versioned builds
...
- Introduce new build arg `VERSION_TAG`
- Update version freshness checking
- Rename docker-compose build example file
9 months ago
79c7cf645e
Fix `CMD` for `llama-cpu` variant
9 months ago
00340c0504
Create `llama-cpu` variant for systems without GPU
...
Fixes #16
9 months ago
6562cf0e16
Update config directories and compose example
...
Add `characters` directory to config
Remove softprompts from compose
10 months ago
b6b1ca1391
Implement per-extension initialisation
...
(Plansee)
10 months ago
4c24344796
Remove ExLlama manual installation (no longer required)
11 months ago
ef85657f3c
Implement Nightly builds as Github workflow ( #10 )
11 months ago
5b6477ddf3
Move nightly builds to dedicated branch
11 months ago
b9d7caffdf
Update docker-nightly.yml
11 months ago
f4638e1bd2
Create docker-nightly.yml
11 months ago
d8fcbd7ca3
Fix example standalone run commands
...
Credit to @sdizen in #7 connection refused issue - thanks for sharing the fix!
11 months ago
84c1bbe883
Integrate `llama-cublas` into base image
...
- Update README.md with deprecation warnings and improved descriptions.
- Fix the build error in the latest `llama-cpp-python` version by removing CMAKE directives.
11 months ago
ff496b6929
Integrate ExLlama in base image
...
Closes #6
11 months ago
ee3f779157
Update README.md
11 months ago
56f068d4e8
Disable extensions folder mapping by default
11 months ago
068bea0948
Minor update README.me
11 months ago
19f9c1b1ac
Implement persistent extensions with optional runtime build
...
- Extensions are now explicitly copied from source to named volume
- Optional rebuild at runtime allows for extensions not present during build
- Closes #3
11 months ago
7caaaa4a7c
Update to upstream changes
...
- AutoGPTQ manual installation removed (it is included in requirements)
- Softprompt config removed
- Build date print-out added to `docker-entrypoint.sh`
- `README.md` updated
11 months ago
524dad64c9
Fix `api` set up hint in README.md
11 months ago
a0137d40b8
Refactor to pull Docker hub images
...
Also fixes `triton` dependency conflict
11 months ago
Atinoda
Benjamin McLean