add Ulysses Attention, Ring Attention, Unified Attention, and Ulysses Anything Attention
This commit is contained in:
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Base" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Base" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Base" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Distilled" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Distilled" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Distilled" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Mid" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Mid" \
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
# Example: Running inference with 2-GPU parallelism
|
||||
# CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node 2 infer_helios.py \
|
||||
# --enable_parallelism \
|
||||
# --cp_backend "ulysses" \ # ["ring", "ulysses", "unified", "ulysses_anything"]
|
||||
|
||||
CUDA_VISIBLE_DEVICES=0 python infer_helios.py \
|
||||
--base_model_path "BestWishYsh/Helios-Mid" \
|
||||
|
||||
Reference in New Issue
Block a user