 AlpinDale
|
c05a45f22f
chore: minor updates to throughput benchmark and llm class
|
7 hónapja |
|
AlpinDale
|
458c8b5e33
chore: estimated input speed for tqdm
|
7 hónapja |
|
AlpinDale
|
e989ddf2a1
chore: allow using multiple lora modules in llm.generate()
|
7 hónapja |
|
AlpinDale
|
d00a7517e6
fix: tokenizer delay with using LLM class
|
7 hónapja |
|
AlpinDale
|
8d77c69cbd
feat: support image processor and add llava example
|
7 hónapja |
|
AlpinDale
|
8c61fb9c19
fix: prevent LLM.encode() to be used with causal models
|
7 hónapja |
|
AlpinDale
|
90ceab32ff
refactor: consolidate prompt args to LLM engines
|
7 hónapja |
|
AlpinDale
|
342346afda
improve hashing function
|
8 hónapja |
|
AlpinDale
|
be8154a8a0
feat: proper embeddings API with e5-mistral-7b support
|
8 hónapja |
|
AlpinDale
|
7e6b309f89
tok/s in llm class
|
8 hónapja |
|
AlpinDale
|
35ae01d7ba
refactor: attention metadata term
|
8 hónapja |
|
AlpinDale
|
46159b107a
formatting: pt1
|
8 hónapja |
|
AlpinDale
|
bf2dd2bee9
feat: allow multiple sampling params in LLM class
|
8 hónapja |
 AlpinDale
|
9d81716bfd
[v0.5.3] Release Candidate (#388)
|
10 hónapja |
|
AlpinDale
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
11 hónapja |
|
AlpinDale
|
e42a78381a
feat: switch from pylint to ruff (#322)
|
1 éve |
|
AlpinDale
|
ac82b67f75
feat: naive context shift and various QoL changes (#289)
|
1 éve |
|
AlpinDale
|
c3a221eb02
feat: GGUF, QuIP#, and Marlin support (#228)
|
1 éve |
|
AlpinDale
|
641bb0f6e9
feat: add custom allreduce kernels (#224)
|
1 éve |
|
AlpinDale
|
c0aac15421
feat: S-LoRA support (#222)
|
1 éve |
|
AlpinDale
|
8fa608aeb7
feat: replace Ray with NCCL for control plane comms (#221)
|
1 éve |
|
AlpinDale
|
f013d714c0
chore: merge dev branch into main (#177)
|
1 éve |
|
AlpinDale
|
2755a48d51
merge dev branch into main (#153)
|
1 éve |
|
AlpinDale
|
8834ecf9de
chore: clean up refactor endpoints (#98)
|
1 éve |
|
AlpinDale
|
c70abc7522
fix the LLM class for quantization
|
1 éve |
|
AlpinDale
|
6b9561ef07
adapt TGI incremental detokenization
|
1 éve |
|
AlpinDale
|
388d7545dd
fix: circular import
|
1 éve |
|
AlpinDale
|
c761d38c69
fix: sort outputs and avoid unwanted list copy
|
1 éve |
|
AlpinDale
|
56077f0f29
upstream: trust remote code
|
1 éve |
|
AlpinDale
|
724852dc31
chore: refactoring cont.
|
1 éve |
