AlpinDale
|
04b53d2db5
chore: add initializer files
|
1 year ago |
AlpinDale
|
d4950a497e
chore: add function calls
|
1 year ago |
AlpinDale
|
a482c54c04
fix: various fixes from upstream
|
1 year ago |
AlpinDale
|
8f7853c255
fix: typo in args_tools.py
|
1 year ago |
AlpinDale
|
646b514323
feat: add draft for async engine
|
1 year ago |
AlpinDale
|
90596764fb
feat: finish the main engine class
|
1 year ago |
AlpinDale
|
54b6f5becf
feat: add GPU worker class
|
1 year ago |
AlpinDale
|
5e25fe5e64
feat: add model loader
|
1 year ago |
AlpinDale
|
e283de0a88
feat: adapt modeling_llama.py
|
1 year ago |
AlpinDale
|
42682fdaf9
feat: adapt modeling_llama.py
|
1 year ago |
AlpinDale
|
d884c3c7ba
Remove pre-mature commit
|
1 year ago |
AlpinDale
|
c25ee56847
chore: change epsilon variance value
|
1 year ago |
AlpinDale
|
49b8a478ee
feat: add normalization layers
|
1 year ago |
AlpinDale
|
48a75478cb
feat: add generation sampler
|
1 year ago |
AlpinDale
|
6912d232ec
feat: added paged attention
|
1 year ago |
AlpinDale
|
98b81d5921
fix: incorrect act names
|
1 year ago |
AlpinDale
|
3f7a3f7b64
feat: implement swiglu activation
|
1 year ago |
AlpinDale
|
0ec53128b6
feat: add layernorm kernels
|
1 year ago |
AlpinDale
|
67a17a1e93
feat: add rotary embeddings
|
1 year ago |
AlpinDale
|
28866137ea
feat: add swiglu activation
|
1 year ago |
AlpinDale
|
d40a8d6bb0
chore: bind single_query_cached_kv_attention to python
|
1 year ago |
AlpinDale
|
05d0a7e763
feat: adapt the attention kernels
|
1 year ago |
AlpinDale
|
6aa1a9ee79
feat: add bf16 datatype headers
|
1 year ago |
AlpinDale
|
ea88cc1864
fix: typo in fp16 headers
|
1 year ago |
AlpinDale
|
4be4515a78
chore: copyright notice
|
1 year ago |
AlpinDale
|
4061b1721f
chore: add NVIDIA's license to README
|
1 year ago |
AlpinDale
|
82e4cf5161
feat: add fp16 kernels
|
1 year ago |
AlpinDale
|
3c3944153c
feat: add generic attention and FP32 dtype kernels
|
1 year ago |
AlpinDale
|
b48fe85378
chore: utilities for modeling
|
1 year ago |
AlpinDale
|
68027b2c0c
fix: forgot to add import for ray tools
|
1 year ago |