.. |
containers
|
75db3d7da7
Fix SD workflow to work with latest diffusers version (#4918)
|
9 月之前 |
__init__.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 年之前 |
auto_tp.py
|
1787673edc
fix num_kv_heads sharding in uneven autoTP for Falcon-40b (#4712)
|
9 月之前 |
auto_tp_model_utils.py
|
c20f6fa4e0
support baichuan model: (#4721)
|
10 月之前 |
fusedqkv_utils.py
|
85132adc31
enable starcode((kv_head=1)) autotp (#4896)
|
9 月之前 |
inject.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 年之前 |
layers.py
|
f15cccfa0c
[AutoTP] Make AutoTP work when num_heads not divisible by number of workers (#4011)
|
1 年之前 |
load_checkpoint.py
|
d72edb3b0d
fix lm head overriden issue, move it from checkpoint in-loop loading to out loop (#4206)
|
1 年之前 |
module_quantize.py
|
430510bfce
Checks for user injection policy (#3052)
|
1 年之前 |
policy.py
|
0a61d5d664
Hybrid Engine Refactor and Llama Inference Support (#3425)
|
1 年之前 |
replace_module.py
|
85132adc31
enable starcode((kv_head=1)) autotp (#4896)
|
9 月之前 |
replace_policy.py
|
468882fb68
Add the policy to run llama model from the official repo (#4313)
|
1 年之前 |
tp_shard.py
|
85132adc31
enable starcode((kv_head=1)) autotp (#4896)
|
9 月之前 |
utils.py
|
468882fb68
Add the policy to run llama model from the official repo (#4313)
|
1 年之前 |