https://zh.d2l.ai/chapter_linear-networks/softmax-regression-scratch.html
# 练习 1
print(tf.math.exp(50))
---------------------------------------------------------------------------
InvalidArgumentError Traceback (most recent call last)
Cell In[37], line 2
1 # 练习 1
----> 2 print(tf.math.exp(50))
File /usr/local/lib/python3.11/dist-packages/tensorflow/python/ops/weak_tensor_ops.py:88, in weak_tensor_unary_op_wrapper.<locals>.wrapper(*args, **kwargs)
86 def wrapper(*args, **kwargs):
87 if not ops.is_auto_dtype_conversion_enabled():
---> 88 return op(*args, **kwargs)
89 bound_arguments = signature.bind(*args, **kwargs)
90 bound_arguments.apply_defaults()
File /usr/local/lib/python3.11/dist-packages/tensorflow/python/util/traceback_utils.py:153, in filter_traceback.<locals>.error_handler(*args, **kwargs)
151 except Exception as e:
152 filtered_tb = _process_traceback_frames(e.__traceback__)
--> 153 raise e.with_traceback(filtered_tb) from None
154 finally:
155 del filtered_tb
File /usr/local/lib/python3.11/dist-packages/tensorflow/python/framework/ops.py:6002, in raise_from_not_ok_status(e, name)
6000 def raise_from_not_ok_status(e, name) -> NoReturn:
6001 e.message += (" name: " + str(name if name is not None else ""))
-> 6002 raise core._status_to_exception(e) from None
InvalidArgumentError: Value for attr 'T' of int32 is not in the list of allowed values: bfloat16, half, float, double, complex64, complex128
; NodeDef: {{node Exp}}; Op<name=Exp; signature=x:T -> y:T; attr=T:type,allowed=[DT_BFLOAT16, DT_HALF, DT_FLOAT, DT_DOUBLE, DT_COMPLEX64, DT_COMPLEX128]> [Op:Exp] name:
# 练习 2
print(-tf.math.log(0))
---------------------------------------------------------------------------
InvalidArgumentError Traceback (most recent call last)
Cell In[38], line 2
1 # 练习 2
----> 2 print(-tf.math.log(0))
File /usr/local/lib/python3.11/dist-packages/tensorflow/python/ops/weak_tensor_ops.py:88, in weak_tensor_unary_op_wrapper.<locals>.wrapper(*args, **kwargs)
86 def wrapper(*args, **kwargs):
87 if not ops.is_auto_dtype_conversion_enabled():
---> 88 return op(*args, **kwargs)
89 bound_arguments = signature.bind(*args, **kwargs)
90 bound_arguments.apply_defaults()
File /usr/local/lib/python3.11/dist-packages/tensorflow/python/ops/gen_math_ops.py:5681, in log(x, name)
5679 return _result
5680 except _core._NotOkStatusException as e:
-> 5681 _ops.raise_from_not_ok_status(e, name)
5682 except _core._FallbackException:
5683 pass
File /usr/local/lib/python3.11/dist-packages/tensorflow/python/framework/ops.py:6002, in raise_from_not_ok_status(e, name)
6000 def raise_from_not_ok_status(e, name) -> NoReturn:
6001 e.message += (" name: " + str(name if name is not None else ""))
-> 6002 raise core._status_to_exception(e) from None
InvalidArgumentError: Value for attr 'T' of int32 is not in the list of allowed values: bfloat16, half, float, double, complex64, complex128
; NodeDef: {{node Log}}; Op<name=Log; signature=x:T -> y:T; attr=T:type,allowed=[DT_BFLOAT16, DT_HALF, DT_FLOAT, DT_DOUBLE, DT_COMPLEX64, DT_COMPLEX128]> [Op:Log] name:
# 练习 3
def cross_entropy(y_hat, y):
# 对预测概率进行数值裁剪，避免对数计算时出现数值不稳定问题
y_hat = tf.clip_by_value(y_hat, 1e-10, 1.0)
# 获取每个样本对应真实标签的索引
indices = tf.stack([tf.range(y.shape[0]), y], axis=1)
# 根据索引从预测概率张量中提取相应的元素
y_hat_selected = tf.gather_nd(y_hat, indices)
# 计算交叉熵损失
return -tf.math.log(y_hat_selected)
cross_entropy(y_hat, y)
<tf.Tensor: shape=(2,), dtype=float32, numpy=array([2.3025851, 0.6931472], dtype=float32)>
练习 4
并不是，不同场景下可能会对分类设置不同的阈值。比如要判断一个人是否需要进一步接受检查，则可能对于某种病源检出的概率不足50%时，就非常有必要进行进一步检查。我们可能要根据这个概率给出“疑病”的诊断，而非“大概率不患病”。
练习 5
当单词数量过多时，one-hot 编码会变得非常稀疏，导致模型难以学习，并消耗大量内存。此时可能需要借助其他方法，如词嵌入（word embedding）来解决这个问题。