-
Notifications
You must be signed in to change notification settings - Fork 25
Open
Description
Versions:
MAX 25.7.0
OS: 26.2 (25C56)
✨ Pixi task (s05): python checks/check_step_05.py Running checks for Step 05: Layer Normalization...
✅ Successfully imported LayerNorm
✅ LayerNorm can be instantiated
✅ Has attribute: weight
✅ Has attribute: bias
✅ Has attribute: eps
✅ weight has correct shape: (768,)
✅ bias has correct shape: (768,)
✅ Has forward method
error: invalid call to 'execute': value passed to 'gamma' cannot be converted from 'ManagedTensorSlice[FusedInput, static_spec=rebuild_static_tensor_specs_with_input_lambda[DType.bfloat16, 1](StaticTensorSpec[DType.bfloat16, 1](param__26, param__27, 256, AddressSpace.GENERIC, True, OptionalReg[fn[simd_width: Int, element_alignment: Int = 1](IndexList[1]) capturing -> SIMD[DType.bfloat16, simd_width]](None), OptionalReg[fn[simd_width: Int, element_alignment: Int = 1](IndexList[1], SIMD[DType.bfloat16, simd_width]) capturing -> None](None), OptionalReg[fn[simd_width: Int, element_alignment: Int = 1](IndexList[1], SIMD[DType.bfloat16, simd_width]) capturing -> SIMD[DType.bfloat16, simd_width]](None)), input_lambda__23[::Int,::Int](::IndexList[::Int(1), ::DType(int64)]))]' to 'ManagedTensorSlice[FusedInput, static_spec=gamma.static_spec]'
MOGGKernelAPI.LayerNorm.execute[target = "gpu"](var__39, var__18, var__30, var__41, var__40, arg__5)
~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~ ~~~~~~~
note: function declared here
from MOGGPrimitives import *
^
note: .dtype of left value is 'DType.bfloat16' but the right value is 'DType.float32'
MOGGKernelAPI.LayerNorm.execute[target = "gpu"](var__39, var__18, var__30, var__41, var__40, arg__5)
^
❌ ERRORS:
- Forward pass failed: Failed to compile the model. Please file an issue, all models should be correct by construction and this error should have been caught during construction.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels