 AlpinDale
|
8db2fa8e2e
why was that not committed?
|
8 months ago |
|
AlpinDale
|
54678c91f3
fix outlines requirements
|
8 months ago |
|
AlpinDale
|
e758c2aef4
fix micromamba url
|
8 months ago |
|
AlpinDale
|
50c2434267
move megatron to a top-level directory
|
8 months ago |
|
AlpinDale
|
92b649e81b
fix case where hf_config==None
|
8 months ago |
|
AlpinDale
|
63c2508ab4
no key sorting for outlines
|
8 months ago |
|
AlpinDale
|
2fffd10bd3
head_size 256 for gemma in triton FA
|
8 months ago |
|
AlpinDale
|
029ad054f0
skip rows in logits added for the prompt tokens
|
8 months ago |
|
AlpinDale
|
8d737c8a9a
fix types in merge_dict
|
8 months ago |
 AlpinDale
|
4d33ce60da
feat: Triton flash attention backend for ROCm (#407)
|
8 months ago |
|
AlpinDale
|
9e9057a614
separate init_distributed_environment from worker
|
9 months ago |
|
AlpinDale
|
40f63268ee
disable new layernorm kernels for CUDA < 12.0
|
9 months ago |
|
AlpinDale
|
8e87b290c2
enable attention bias support in llama
|
9 months ago |
|
AlpinDale
|
893c791152
fix TP for llava
|
9 months ago |
|
AlpinDale
|
f52aa64fe6
use the get_len() method instead of manual len calculation
|
9 months ago |
|
AlpinDale
|
c2aaaefd57
allow out-of-tree model registry
|
9 months ago |
|
AlpinDale
|
082d4e6972
feat: add chunked prefill scheduler (#406)
|
9 months ago |
 sgsdxzy
|
638547ec98
fix: Improve cohere model. (#404)
|
9 months ago |
|
AlpinDale
|
b4fcaf7aa3
add sampling param for left-truncating prompt tokens
|
9 months ago |
|
AlpinDale
|
0b1aad2924
split requirements
|
9 months ago |
|
AlpinDale
|
355a21f1ba
make nccl wrapper more robust
|
9 months ago |
|
AlpinDale
|
2f4e7aba13
update torch to 2.2.1
|
9 months ago |
|
AlpinDale
|
7528e0ce3e
make detokenization optional
|
9 months ago |
|
AlpinDale
|
fb23720c72
fix CPU build
|
9 months ago |
|
AlpinDale
|
23a1114e4f
enable hf_transfer if installed
|
9 months ago |
|
AlpinDale
|
282d7b7f9c
fix multi-gpu ray tokenizer for trust_remote_code
|
9 months ago |
|
AlpinDale
|
071269e406
feat: FP8 E4M3 KV Cache (#405)
|
9 months ago |
|
AlpinDale
|
6f00203041
refactor scheduler for chunked prefill, remove reorder policy for now
|
9 months ago |
|
AlpinDale
|
14f39af8b5
add dict merging util
|
9 months ago |
|
AlpinDale
|
6f1d13d30a
better recognize cpu build
|
9 months ago |
