 AlpinDale
|
ed6717d0c0
feat: initial support for control vectors
|
4 months ago |
 AlpinDale
|
a8d10fcfee
chore: add contribution guidelines + Code of Conduct (#507)
|
6 months ago |
|
AlpinDale
|
30f9e1f7c9
remove docker
|
4 months ago |
|
AlpinDale
|
206261a3fb
chore: add contribution guidelines + Code of Conduct (#507)
|
6 months ago |
|
AlpinDale
|
9371a33e90
docs: add installation guides
|
4 months ago |
 Ahmed
|
da67a9c8d1
docs: add doc site with example content
|
4 months ago |
|
AlpinDale
|
f91991f584
fix: f-string fixes
|
4 months ago |
|
AlpinDale
|
acbdc50a71
fix: `vocab_size` field access in llava
|
4 months ago |
|
AlpinDale
|
a3e26391e4
chore: add a wrapper for torch.inference_mode decorator
|
4 months ago |
|
AlpinDale
|
79e56506d7
clean up requirements
|
4 months ago |
|
AlpinDale
|
a0d031efcc
feat: initial text-to-text support for Chameleon model
|
4 months ago |
|
AlpinDale
|
9038dea2df
fix: short commit hash import error
|
4 months ago |
|
AlpinDale
|
ba371fbbbd
feat: AWQ marlin kernels (#603)
|
4 months ago |
|
AlpinDale
|
08373fd1ee
fix: asymmetric TP changes breaking the gptq and awq quants (#602)
|
4 months ago |
|
AlpinDale
|
ad68d149d8
chore: refactor and decouple phi3v image embedding
|
4 months ago |
|
AlpinDale
|
a4cbcfe59f
feat: disable logprob serialization to CPU for spec decode
|
4 months ago |
|
AlpinDale
|
f6d2f3c4ba
move `aphrodite.endpoints.openai.chat_utils` -> `aphrodite.endpoints.chat_utils`
|
4 months ago |
|
AlpinDale
|
055963b252
fix: channel-wise fp8 marlin
|
4 months ago |
|
AlpinDale
|
5d98b7ead1
fix: input_scale for w8a8 is optional
|
4 months ago |
|
AlpinDale
|
408ca43d2e
feat: support fbgemm_fp8 quant on ampere
|
4 months ago |
|
AlpinDale
|
fa15bad2ea
chore: minor AMD fixes
|
4 months ago |
|
AlpinDale
|
9be43994fe
feat: fbgemm quantization support (#601)
|
4 months ago |
|
AlpinDale
|
cac8163f77
fix: request abort crashing pipeline parallel
|
4 months ago |
|
AlpinDale
|
45a004874c
chore: allow specifying custom Executor
|
4 months ago |
|
AlpinDale
|
c8f5424d72
add scale_ub inputs to fp8 dynamic per-token quant
|
4 months ago |
|
AlpinDale
|
5f84f0651c
docker: install libibverbs by default
|
4 months ago |
|
AlpinDale
|
d3c474d219
chore: enable dynamic per-token `fp8`
|
4 months ago |
|
AlpinDale
|
c92e46af58
fix shared memory bug w/ multi-node
|
4 months ago |
|
AlpinDale
|
2ebb37d1ee
update time since last collection for AsyncMetricsCollector
|
4 months ago |
|
AlpinDale
|
6ac658b0d6
some small performance improvements
|
4 months ago |
