 AlpinDale
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
4 mesiacov pred |
|
AlpinDale
|
9d81716bfd
[v0.5.3] Release Candidate (#388)
|
8 mesiacov pred |
|
AlpinDale
|
78d66f16d1
Chunked Prefill Part 1 (#384)
|
9 mesiacov pred |
|
AlpinDale
|
feb5840f2a
feat: async tokenization (#374)
|
9 mesiacov pred |
|
AlpinDale
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
9 mesiacov pred |
|
AlpinDale
|
c41462cfcd
feat: exllamav2 quantization (#305)
|
10 mesiacov pred |
|
AlpinDale
|
9810daa699
feat: INT8 KV Cache (#298)
|
10 mesiacov pred |
|
AlpinDale
|
e0c35bb353
feat: bitsandbytes and `--load-in{4,8}bit` support (#294)
|
10 mesiacov pred |
|
AlpinDale
|
705821a7fe
feat: AQLM quantization support (#293)
|
10 mesiacov pred |
|
AlpinDale
|
ac82b67f75
feat: naive context shift and various QoL changes (#289)
|
10 mesiacov pred |
|
AlpinDale
|
72229a94da
feat: better marlin kernels (#285)
|
10 mesiacov pred |
|
AlpinDale
|
657aec0cbd
refactor: OpenAI endpoint (#261)
|
10 mesiacov pred |
|
AlpinDale
|
4d04ade9ef
feat: fine-grained seeds (#279)
|
11 mesiacov pred |
|
AlpinDale
|
ea0f57b233
feat: allow further support for non-cuda devices (#247)
|
11 mesiacov pred |
|
AlpinDale
|
c3a221eb02
feat: GGUF, QuIP#, and Marlin support (#228)
|
1 rok pred |
|
AlpinDale
|
31c95011a6
feat: FP8 E5M2 KV Cache (#226)
|
1 rok pred |
|
AlpinDale
|
641bb0f6e9
feat: add custom allreduce kernels (#224)
|
1 rok pred |
|
AlpinDale
|
c0aac15421
feat: S-LoRA support (#222)
|
1 rok pred |
|
AlpinDale
|
8fa608aeb7
feat: replace Ray with NCCL for control plane comms (#221)
|
1 rok pred |
|
AlpinDale
|
15a0454172
feat: FP8 KV Cache (#185)
|
1 rok pred |
|
AlpinDale
|
7d91e9e0f2
feat: CUDA graphs (#172)
|
1 rok pred |
|
AlpinDale
|
653da510d1
chore: rewrite InputMetadata (#143)
|
1 rok pred |
|
AlpinDale
|
1334a833a4
feat: AMD ROCm support (#95)
|
1 rok pred |
|
AlpinDale
|
2b1ba581f9
feat: re-implement GPTQ (#141)
|
1 rok pred |
|
AlpinDale
|
8223f85c1b
feat: SqueezeLLM support (#140)
|
1 rok pred |
|
AlpinDale
|
1aab8a7d6f
feat: speedup compilation times by 3x (#130)
|
1 rok pred |
|
AlpinDale
|
237d2ec28d
fix: CPU OOM for large models (#128)
|
1 rok pred |
|
AlpinDale
|
8834ecf9de
chore: clean up refactor endpoints (#98)
|
1 rok pred |
|
AlpinDale
|
efc6f7fbec
chore: reformats (#90)
|
1 rok pred |
|
AlpinDale
|
3d72f05c7b
feat: flattened 1D tensor -> 2D tensor (#85)
|
1 rok pred |
