unsloth_zoo>=2025.9.10
torch>=2.4.0
xformers>=0.0.27.post2
bitsandbytes
packaging
tyro
transformers!=4.47.0,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,<=4.56.2,>=4.51.3
datasets<4.0.0,>=3.4.1
sentencepiece>=0.2.0
tqdm
psutil
wheel>=0.42.0
numpy
accelerate>=0.34.1
trl!=0.15.0,!=0.19.0,!=0.9.0,!=0.9.1,!=0.9.2,!=0.9.3,>=0.7.9
peft!=0.11.0,>=0.7.1
protobuf
huggingface_hub>=0.34.0
hf_transfer
diffusers
torchvision

[:platform_system == "Linux"]
triton>=3.0.0

[:platform_system == "Windows"]
triton-windows

[base]
unsloth[huggingface]

[colab]
unsloth[cu121]

[colab-ampere]
unsloth[colab-ampere-torch220]
unsloth[flashattention]

[colab-ampere-torch211]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch211]
unsloth[flashattention]

[colab-ampere-torch220]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch220]
unsloth[flashattention]

[colab-new]
unsloth_zoo>=2025.9.10
packaging
tyro
transformers!=4.47.0,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,<=4.56.2,>=4.51.3
datasets<4.0.0,>=3.4.1
sentencepiece>=0.2.0
tqdm
psutil
wheel>=0.42.0
numpy
protobuf
huggingface_hub>=0.34.0
hf_transfer
bitsandbytes>=0.43.3
unsloth[triton]

[colab-no-deps]
accelerate>=0.34.1
trl!=0.15.0,!=0.19.0,!=0.9.0,!=0.9.1,!=0.9.2,!=0.9.3,>=0.7.9
peft>=0.7.1
xformers<0.0.27
bitsandbytes>=0.43.3
protobuf

[colab-torch211]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch211]

[colab-torch220]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch220]

[conda]
unsloth[huggingface]

[cu118]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118only]

[cu118-ampere]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118only]
unsloth[flashattention]

[cu118-ampere-torch211]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch211]
unsloth[flashattention]

[cu118-ampere-torch220]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch220]
unsloth[flashattention]

[cu118-ampere-torch230]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch230]
unsloth[flashattention]

[cu118-ampere-torch240]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch240]
unsloth[flashattention]

[cu118-ampere-torch250]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch250]
unsloth[flashattention]

[cu118-ampere-torch251]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch251]
unsloth[flashattention]

[cu118-ampere-torch260]
unsloth[huggingface]
bitsandbytes>=0.45.1
unsloth[cu118onlytorch260]
unsloth[flashattention]

[cu118-ampere-torch270]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu118onlytorch270]
unsloth[flashattention]

[cu118-ampere-torch271]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu118onlytorch271]
unsloth[flashattention]

[cu118-ampere-torch280]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu118onlytorch280]
unsloth[flashattention]

[cu118-torch211]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch211]

[cu118-torch212]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch212]

[cu118-torch220]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch220]

[cu118-torch230]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch230]

[cu118-torch240]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch240]

[cu118-torch250]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch250]

[cu118-torch251]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu118onlytorch251]

[cu118-torch260]
unsloth[huggingface]
bitsandbytes>=0.45.1
unsloth[cu118onlytorch260]

[cu118-torch270]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu118onlytorch270]

[cu118-torch271]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu118onlytorch271]

[cu118-torch280]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu118onlytorch280]

[cu118only]
xformers==0.0.22.post7

[cu118onlytorch211]
xformers==0.0.23

[cu118onlytorch212]
xformers==0.0.23.post1

[cu118onlytorch220]
xformers==0.0.24

[cu118onlytorch230]
xformers==0.0.27

[cu118onlytorch240]
xformers==0.0.27.post2

[cu118onlytorch250]
xformers==0.0.28.post2

[cu118onlytorch251]
xformers==0.0.29.post1

[cu118onlytorch260]
xformers==0.0.29.post3

[cu118onlytorch270]
xformers==0.0.30

[cu118onlytorch271]
xformers==0.0.31.post1

[cu118onlytorch280]
xformers==0.0.32.post2

[cu121]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121only]

[cu121-ampere]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121only]
unsloth[flashattention]

[cu121-ampere-torch211]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch211]
unsloth[flashattention]

[cu121-ampere-torch220]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch220]
unsloth[flashattention]

[cu121-ampere-torch230]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch230]
unsloth[flashattention]

[cu121-ampere-torch240]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch240]
unsloth[flashattention]

[cu121-ampere-torch250]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch250]
unsloth[flashattention]

[cu121-ampere-torch251]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch251]
unsloth[flashattention]

[cu121-torch211]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch211]

[cu121-torch212]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch212]

[cu121-torch220]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch220]

[cu121-torch230]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch230]

[cu121-torch240]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch240]

[cu121-torch250]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch250]

[cu121-torch251]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu121onlytorch251]

[cu121only]
xformers==0.0.22.post7

[cu121onlytorch211]
xformers==0.0.23

[cu121onlytorch212]
xformers==0.0.23.post1

[cu121onlytorch220]
xformers==0.0.24

[cu121onlytorch230]
xformers==0.0.27

[cu121onlytorch240]
xformers==0.0.27.post2

[cu121onlytorch250]
xformers==0.0.28.post2

[cu121onlytorch251]
xformers==0.0.29.post1

[cu124-ampere-torch240]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu124onlytorch240]
unsloth[flashattention]

[cu124-ampere-torch250]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu124onlytorch250]
unsloth[flashattention]

[cu124-ampere-torch251]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu124onlytorch251]
unsloth[flashattention]

[cu124-ampere-torch260]
unsloth[huggingface]
bitsandbytes>=0.45.1
unsloth[cu124onlytorch260]
unsloth[flashattention]

[cu124-torch240]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu124onlytorch240]

[cu124-torch250]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu124onlytorch250]

[cu124-torch251]
unsloth[huggingface]
bitsandbytes>=0.43.3
unsloth[cu124onlytorch251]

[cu124-torch260]
unsloth[huggingface]
bitsandbytes>=0.45.1
unsloth[cu124onlytorch260]

[cu124onlytorch250]
xformers==0.0.28.post2

[cu124onlytorch251]
xformers==0.0.29.post1

[cu124onlytorch260]
xformers==0.0.29.post3

[cu126-ampere-torch260]
unsloth[huggingface]
bitsandbytes>=0.45.1
unsloth[cu126onlytorch260]
unsloth[flashattention]

[cu126-ampere-torch270]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu126onlytorch270]
unsloth[flashattention]

[cu126-ampere-torch271]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu126onlytorch271]
unsloth[flashattention]

[cu126-ampere-torch280]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu126onlytorch280]
unsloth[flashattention]

[cu126-torch260]
unsloth[huggingface]
bitsandbytes>=0.45.1
unsloth[cu126onlytorch260]

[cu126-torch270]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu126onlytorch270]

[cu126-torch271]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu126onlytorch271]

[cu126-torch280]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu126onlytorch280]

[cu126onlytorch260]
xformers==0.0.29.post3

[cu126onlytorch270]
xformers==0.0.30

[cu126onlytorch271]
xformers==0.0.31.post1

[cu126onlytorch280]
xformers==0.0.32.post2

[cu128-ampere-torch270]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu128onlytorch270]
unsloth[flashattention]

[cu128-ampere-torch271]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu128onlytorch271]
unsloth[flashattention]

[cu128-ampere-torch280]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu128onlytorch280]
unsloth[flashattention]

[cu128-torch270]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu128onlytorch270]

[cu128-torch271]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu128onlytorch271]

[cu128-torch280]
unsloth[huggingface]
bitsandbytes>=0.45.5
unsloth[cu128onlytorch280]

[cu128onlytorch270]
xformers==0.0.30

[cu128onlytorch271]
xformers==0.0.31.post1

[cu128onlytorch280]
xformers==0.0.32.post2

[flashattention]

[flashattention:platform_system == "Linux"]
packaging
ninja
flash-attn>=2.6.3

[huggingface]
unsloth_zoo>=2025.9.10
packaging
tyro
transformers!=4.47.0,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,<=4.56.2,>=4.51.3
datasets<4.0.0,>=3.4.1
sentencepiece>=0.2.0
tqdm
psutil
wheel>=0.42.0
numpy
accelerate>=0.34.1
trl!=0.15.0,!=0.19.0,!=0.9.0,!=0.9.1,!=0.9.2,!=0.9.3,>=0.7.9
peft!=0.11.0,>=0.7.1
protobuf
huggingface_hub>=0.34.0
hf_transfer
diffusers
torchvision

[kaggle]
unsloth[huggingface]

[kaggle-new]
unsloth[huggingface]
bitsandbytes>=0.43.3
