andrewor14

/

Llama-3.1-8B-Instruct-float8dq

like 0

• "torch.float8_e4m3fn",
• "torchao.quantization.linear_activation_quantized_tensor.LinearActivationQuantizedTensor",
• "torch.storage.UntypedStorage",
• "collections.OrderedDict",
• "torchao.quantization.granularity.PerTensor",
• "torchao.dtypes.affine_quantized_tensor.AffineQuantizedTensor",
• "torch.bfloat16",
• "torchao.dtypes.floatx.float8_layout.Float8Layout",
• "torchao.float8.inference.Float8MMConfig",
• "torch._utils._rebuild_tensor_v3",
• "torchao.dtypes.floatx.float8_layout.Float8AQTTensorImpl",
• "torchao.quantization.quant_api._input_activation_quant_func_fp8",
• "torch._utils._rebuild_tensor_v2",
• "torch.Size",
• "torch._tensor._rebuild_from_type_v2",
• "torch.serialization._get_layout",
• "torch.BFloat16Storage",
• "torch._utils._rebuild_wrapper_subclass",
• "torch.FloatStorage",
• "torch.device",
• "torchao.quantization.quant_primitives.ZeroPointDomain"

How to fix it?

• "torch.BFloat16Storage",
• "torchao.dtypes.affine_quantized_tensor.AffineQuantizedTensor",
• "torchao.dtypes.floatx.float8_layout.Float8Layout",
• "torch.device",
• "torchao.quantization.granularity.PerTensor",
• "torchao.dtypes.floatx.float8_layout.Float8AQTTensorImpl",
• "torch.bfloat16",
• "torchao.float8.inference.Float8MMConfig",
• "torch.Size",
• "torch._utils._rebuild_tensor_v3",
• "torchao.quantization.quant_api._input_activation_quant_func_fp8",
• "collections.OrderedDict",
• "torch._utils._rebuild_tensor_v2",
• "torchao.quantization.linear_activation_quantized_tensor.LinearActivationQuantizedTensor",
• "torch.serialization._get_layout",
• "torch.FloatStorage",
• "torch._utils._rebuild_wrapper_subclass",
• "torch._tensor._rebuild_from_type_v2",
• "torchao.quantization.quant_primitives.ZeroPointDomain",
• "torch.storage.UntypedStorage",
• "torch.float8_e4m3fn"

How to fix it?