diff --git a/pyproject.toml b/pyproject.toml index de1583e9e3..73e69dcd4a 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -40,7 +40,7 @@ triton = [ ] windows=[ - "unsloth_zoo>=2025.2.7", + "unsloth_zoo>=2025.3.1", "packaging", "tyro", "transformers>=4.46.1,!=4.47.0", @@ -61,7 +61,7 @@ windows=[ "xformers>=0.0.22.post7 ; platform_system == 'Windows'", ] huggingface = [ - "unsloth_zoo>=2025.2.7", + "unsloth_zoo>=2025.3.1", "packaging", "tyro", "transformers>=4.46.1,!=4.47.0", diff --git a/unsloth/__init__.py b/unsloth/__init__.py index caa06b012d..c8f2926985 100644 --- a/unsloth/__init__.py +++ b/unsloth/__init__.py @@ -198,7 +198,7 @@ pass # Check for unsloth_zoo try: unsloth_zoo_version = importlib_version("unsloth_zoo") - if Version(unsloth_zoo_version) < Version("2025.2.6"): + if Version(unsloth_zoo_version) < Version("2025.3.1"): try: os.system("pip install --upgrade --no-cache-dir --no-deps unsloth_zoo") except: diff --git a/unsloth/kernels/utils.py b/unsloth/kernels/utils.py index 273eddcc20..5eb9b8f5ce 100644 --- a/unsloth/kernels/utils.py +++ b/unsloth/kernels/utils.py @@ -473,7 +473,7 @@ pass def matmul_lora(X, W, W_quant, A, B, s, out = None): dtype = X.dtype W = fast_dequantize(W.t(), W_quant, use_global_buffer = True) - + if X.dim() == 3: batch, seq_len, d = X.shape X = X.view(-1, X.shape[-1])