-
Notifications
You must be signed in to change notification settings - Fork 273
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
【Hackathon 5th No.12】Add AdaptiveLogSoftmaxWithLoss API to Paddle #770
Merged
Merged
Changes from all commits
Commits
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,12 +1,12 @@ | ||
# paddle.nn.AdaptiveLogSoftmaxWithLoss 设计文档 | ||
|
||
|API名称 | paddle.nn.AdaptiveLogSoftmaxWithLoss | | ||
|API名称 | paddle.nn.AdaptiveLogSoftmaxWithLoss | | ||
|---|------------------------------------| | ||
|提交作者<input type="checkbox" class="rowselector hidden"> | PeachML | | ||
|提交时间<input type="checkbox" class="rowselector hidden"> | 2022-03-22 | | ||
|版本号 | V1.0 | | ||
|依赖飞桨版本<input type="checkbox" class="rowselector hidden"> | develop | | ||
|文件名 | 20200322_api_design_for_AdaptiveLogSoftmaxWithLoss.md<br> | | ||
|提交作者<input type="checkbox" class="rowselector hidden"> | netpunk | | ||
|提交时间<input type="checkbox" class="rowselector hidden"> | 2023-12-02 | | ||
|版本号 | V1.0 | | ||
|依赖飞桨版本<input type="checkbox" class="rowselector hidden"> | develop | | ||
|文件名 | 20200322_api_design_for_AdaptiveLogSoftmaxWithLoss.md<br> | | ||
|
||
# 一、概述 | ||
|
||
|
@@ -17,10 +17,33 @@ Paddle需要扩充API,新增 AdaptiveLogSoftmaxWithLoss API, | |
调用路径为:`paddle.nn.AdaptiveLogSoftmaxWithLoss` 和 `paddle.nn.functional.adaptive_log_softmax_with_loss`。 | ||
实现Softmax快速近似计算的功能。 | ||
|
||
## 2、功能目标 | ||
|
||
为飞桨补充 AdaptiveLogSoftmaxWithLoss API,该API实现 softmax 函数近似计算 | ||
|
||
adaptive_log_softmax_with_loss的计算分步骤如下 | ||
|
||
1. ![image](https://github.com/PaddlePaddle/community/assets/69072522/3f17c9fd-212a-444c-9a87-2a975c452940) | ||
|
||
(将输入 `input` 通过线性变换映射到一个高维空间,其中 `head_weight` 是学习到的权重,`head_bias` 是偏置项。这个映射允许模型学习类别之间的复杂关系。) | ||
|
||
2. ![image](https://github.com/PaddlePaddle/community/assets/69072522/893286a4-9c78-4e7f-b5f0-ec152ef69267) | ||
|
||
( 将线性变换后的结果进行 softmax 操作,得到每个类别的概率分布,然后取对数。这有助于解决数值稳定性的问题,并且对数概率更容易处理。) | ||
|
||
3. ![image](https://github.com/PaddlePaddle/community/assets/69072522/b6987bfb-e1a6-4193-9c12-818b9cc2a76c) | ||
|
||
(从 `head_logprob` 中选择与给定类别索引 `gather_inds` 相对应的对数概率,然后将其累加到 `output` 中。这一步是为了计算 adaptive softmax 损失,其中仅关注一小部分类别的对数概率。) | ||
|
||
4. ![image](https://github.com/PaddlePaddle/community/assets/69072522/e0b6e756-a6d3-46d5-b0bb-240e3847a05f) | ||
|
||
(将累加的对数概率取负值并求平均,得到损失值。这是一个常见的负对数似然损失,用于衡量模型输出与真实标签之间的差异。) | ||
|
||
这个函数不止输出`loss`,还输出`output`,表示经过 log softmax 转换后的对数概率的累加值,即每个类别的对数概率的总和。可能用于其他需要基于类别概率进行决策或分析的需求 | ||
|
||
## 3、意义 | ||
在自然语言处理中,当字典维度过大时,embedding 将占据模型大部分参数量。 | ||
例如机器翻译任务中,词表维度大约是2^17,embedding维度取1024,那么就会产生将近1亿参数量, | ||
如果不共享embedding矩阵和softmax映射的矩阵,将会再多出1亿参数量。 | ||
|
||
在自然语言处理中,当字典维度过大时,embedding 可能会占据模型较大部分的参数量。 | ||
|
||
这样会引起常见的两个问题: | ||
|
||
|
@@ -144,113 +167,115 @@ Efficient softmax approximation as described in | |
|
||
2. 训练 | ||
```python | ||
def forward(self, input, target): | ||
# input的shape为[batch_size * bptt, hidden_size] | ||
# target的shape为[batch_size * bptt, 1] | ||
if input.size(0) != target.size(0): | ||
raise RuntimeError('Input and target should have the same size ' | ||
'in the batch dimension.') | ||
# 用来统计多个cluster计算的batch,然后求和,保证最终等于batch_size | ||
def forward(self, input_: Tensor, target_: Tensor) -> _ASMoutput: | ||
targ_dim = target_.dim() | ||
|
||
if targ_dim == 1: | ||
if input_.size(0) != target_.size(0): | ||
raise RuntimeError('Input and target should have the same size ' | ||
'in the batch dimension.') | ||
if input_.dim() != 2: | ||
raise RuntimeError('1D target tensor expects 2D input tensors, ' | ||
'but found inputs with size', input_.size()) | ||
elif targ_dim == 0: | ||
if input_.dim() != 1: | ||
raise RuntimeError('0D target tensor expects 1D input tensors, ' | ||
'but found inputs with size', input_.size()) | ||
else: | ||
raise RuntimeError('0D or 1D target tensor expected, ' | ||
'multi-target not supported') | ||
|
||
is_batched = targ_dim > 0 | ||
input = input_ if is_batched else input_.unsqueeze(0) | ||
target = target_ if is_batched else target_.unsqueeze(0) | ||
|
||
used_rows = 0 | ||
batch_size = target.size(0) | ||
# 用来记录在target位置的 logprob | ||
|
||
output = input.new_zeros(batch_size) | ||
# 用来记录batch样本在第一层对应的类别 | ||
gather_inds = target.new_empty(batch_size) | ||
|
||
cutoff_values = [0] + self.cutoffs | ||
for i in range(len(cutoff_values) - 1): | ||
|
||
low_idx = cutoff_values[i] | ||
high_idx = cutoff_values[i + 1] | ||
# 找到当前cluster的样本对应的index | ||
|
||
target_mask = (target >= low_idx) & (target < high_idx) | ||
row_indices = target_mask.nonzero().squeeze() | ||
# 如果当前cluster没有样本,则没有loss | ||
|
||
if row_indices.numel() == 0: | ||
continue | ||
# target对应高频词,这里只用来记录batch对应的target,高频词的预测在后面 self.head | ||
|
||
if i == 0: | ||
gather_inds.index_copy_(0, row_indices, target[target_mask]) | ||
# target对应低频词 | ||
|
||
else: | ||
# 获取低频cluster对应的target的相对位置 | ||
relative_target = target[target_mask] - low_idx | ||
# 获取对应cluster的input | ||
input_subset = input.index_select(0, row_indices) | ||
# 经过线性变换 得到 [batch_size_i, target_i] | ||
|
||
cluster_output = self.tail[i - 1](input_subset) | ||
# 当前cluster对应第一层权重元素的类别 | ||
cluster_index = self.shortlist_size + i - 1 | ||
# 记录对应第一层的类别 | ||
|
||
gather_inds.index_fill_(0, row_indices, cluster_index) | ||
# 计算当前cluster的log_prob | ||
cluster_logprob = log_softmax(cluster_output, dim=1) | ||
# 获取对应target位置的log_prob | ||
local_logprob = cluster_logprob.gather(1, relative_target.unsqueeze(1)) | ||
# 将结果记录到对应的batch中 | ||
output.index_copy_(0, row_indices, local_logprob.squeeze(1)) | ||
|
||
used_rows += row_indices.numel() | ||
|
||
if used_rows != batch_size: | ||
raise RuntimeError("Target values should be in [0, {}], " | ||
"but values in range [{}, {}] " | ||
"were found. ".format(self.n_classes - 1, | ||
target.min().item(), | ||
target.max().item())) | ||
# 第一层的线性变换,因为无论高频和低频词都需要计算第一层,所以放到了这里统一计算 | ||
raise RuntimeError(f"Target values should be in [0, {self.n_classes - 1}], " | ||
f"but values in range [{target.min().item()}, {target.max().item()}] " | ||
"were found. ") | ||
|
||
head_output = self.head(input) | ||
# 取log_prob | ||
head_logprob = log_softmax(head_output, dim=1) | ||
# 这里是第一层的log_prob和第二层的log_prob加起来作为最后的输出 | ||
# tips: 对于属于第一层的样本,只需要计算第一层的log_prob就好 | ||
# 对于属于第二层的样本,需要将第一层计算得到的cluster对应类别的log_prob和 | ||
第二层cluster内计算得到的log_prob加起来,所以是output += | ||
output += head_logprob.gather(1, gather_inds.unsqueeze(1)).squeeze() | ||
loss = (-output).mean() | ||
# 返回一个nametuple | ||
|
||
if not is_batched: | ||
output = output.squeeze(0) | ||
|
||
return _ASMoutput(output, loss) | ||
``` | ||
|
||
|
||
3. 预测 | ||
```python | ||
def predict(self, input): | ||
""" | ||
def predict(self, input: Tensor) -> Tensor: | ||
r""" This is equivalent to `self.log_prob(input).argmax(dim=1)`, | ||
but is more efficient in some cases. | ||
|
||
Args: | ||
input (Tensor): a minibatch of examples | ||
|
||
Returns: | ||
output (Tensor): a class with the highest probability for each example | ||
|
||
Shape: | ||
- Input: :math:`(N, in\_features)` | ||
- Input: :math:`(N, \texttt{in\_features})` | ||
- Output: :math:`(N)` | ||
""" | ||
# 第一层的线性转化 | ||
|
||
head_output = self.head(input) | ||
# 记录预测target的位置 | ||
output = torch.argmax(head_output, dim=1) | ||
# 判断预测的位置是否都是低频词 | ||
not_in_shortlist = (output >= self.shortlist_size) | ||
# 获取预测高频词的样本index | ||
all_in_shortlist = not (not_in_shortlist.any()) | ||
# 如果预测的结果都为高频词,则直接返回结果 | ||
|
||
if all_in_shortlist: | ||
return output | ||
# 如果预测的结果都为低频词 | ||
|
||
elif not_in_shortlist.all(): | ||
# 计算低频词对应cluster中target对应的log_prob | ||
log_prob = self._get_full_log_prob(input, head_output) | ||
return torch.argmax(log_prob, dim=1) | ||
# 如果预测的结果既有高频词,也有低频词 | ||
|
||
else: | ||
# 只对低频词进行对应cluser的预测 | ||
log_prob = self._get_full_log_prob(input[not_in_shortlist], | ||
head_output[not_in_shortlist]) | ||
output[not_in_shortlist] = torch.argmax(log_prob, dim=1) | ||
return output | ||
|
||
# 计算低频词对应cluster中target对应的log_prob | ||
def _get_full_log_prob(self, input, head_output): | ||
""" Given input tensor, and output of `self.head`, | ||
compute the log of the full distribution """ | ||
|
@@ -268,41 +293,89 @@ def _get_full_log_prob(self, input, head_output): | |
out[:, start_idx:stop_idx] = output_logprob | ||
|
||
return out | ||
|
||
def log_prob(self, input: Tensor) -> Tensor: | ||
r""" Computes log probabilities for all :math:`\texttt{n\_classes}` | ||
|
||
Args: | ||
input (Tensor): a minibatch of examples | ||
|
||
Returns: | ||
log-probabilities of for each class :math:`c` | ||
in range :math:`0 <= c <= \texttt{n\_classes}`, where :math:`\texttt{n\_classes}` is a | ||
parameter passed to ``AdaptiveLogSoftmaxWithLoss`` constructor. | ||
|
||
Shape: | ||
- Input: :math:`(N, \texttt{in\_features})` | ||
- Output: :math:`(N, \texttt{n\_classes})` | ||
|
||
""" | ||
|
||
head_output = self.head(input) | ||
return self._get_full_log_prob(input, head_output) | ||
``` | ||
|
||
|
||
# 四、对比分析 | ||
无其它框架实现 | ||
|
||
# 五、方案设计 | ||
# 五、设计思路与实现方案 | ||
## 命名与参数设计 | ||
API设计为`paddle.nn.AdaptiveLogSoftmaxWithLoss(in_features, n_classes, cutoffs, div_value=4.0, head_bias=False, name=None)`及 | ||
`paddle.nn.functional.adaptive_log_softmax_with_loss(input, label, | ||
in_features, n_classes, cutoffs, div_value=4.0, head_bias=False, name=None)`, 返回为`NamedTuple` 包含 `output` 和 `loss`字段 | ||
|
||
layer层类API:`paddle.nn.AdaptiveLogSoftmaxWithLoss(in_features, n_classes, cutoffs, div_value=4.0, head_bias=False, name=None)`,包含两个主要方法: | ||
- forward(self, input, label),用于训练,返回为`output` 和 `loss` | ||
- predict(self, input),用于预测 | ||
|
||
- in_features (int): 输入tensor的特征数量。 | ||
- n_classes (int): 数据集中类型的个数。 | ||
- cutoffs (Sequence): 用于将label分配到不同存储桶的截断值。 | ||
- div_value (float, 可选): 用于计算簇大小的指数值. 默认值:4.0。 | ||
- head_bias (bool, 可选): 如果为 ``True``,向自适应 softmax 的头部添加偏置项. 默认值:``False``. | ||
- name (str, 可选): 具体用法请参见 :ref:`api_guide_Name`,一般无需设置,默认值为 None。 | ||
|
||
function API:`paddle.nn.functional.adaptive_log_softmax_with_loss(input, label, head_weight, tail_weights, cutoffs, head_bias=None)` 用于训练计算 | ||
|
||
- input (Tensor): 输入张量,数据类型为 float32 或 float64。 | ||
- label (Tensor): 标签张量,数据类型为 float32 或 float64。 | ||
- head_weight (Tensor): 用于线性计算的权重矩阵,数据类型为 float32 或 float64。 | ||
- tail_weights (Tensor): 用于线性计算的权重矩阵,数据类型为 float32 或 float64。 | ||
- cutoffs (Sequence): 用于将label分配到不同存储桶的截断值。 | ||
- head_bias (Tensor, 可选): 用于线性计算的偏置矩阵,数据类型为 float32 或 float64。 | ||
- name (str, 可选): 具体用法请参见 :ref:`api_guide_Name`,一般无需设置,默认值为 None。 | ||
|
||
## 底层OP设计 | ||
使用已有API组合实现,不再单独设计OP。 | ||
|
||
## API实现方案 | ||
主要参考pytorch实现,替换掉部分paddle没有的api | ||
|
||
|
||
计算逻辑参考pytorch实现,并基于paddle API进行重组与封装: | ||
- function API:`paddle.nn.functional.adaptive_log_softmax_with_loss(input, label, head_weight, tail_weights, cutoffs, head_bias=None)`,使用已有api进行组合实现, | ||
|
||
- layer API:`paddle.nn.AdaptiveLogSoftmaxWithLoss(self, in_features, n_classes, cutoffs, div_value=4.0, head_bias=False, name=None)`,包含两个主要方法: | ||
- `forward(self, input, label)`,用于训练,返回为`output` 和 `loss` | ||
- `predict(self, input)`,用于预测,其计算与forward共享权重但是计算逻辑存在差异,故使用已有API组合实现的方式单独实现 | ||
|
||
# 六、测试和验收的考量 | ||
测试考虑的case如下: | ||
|
||
- 数值正确性 | ||
- 数值正确性(CPU、GPU、动态图、静态图) | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. 这个正确性准备怎么验证呢 There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. 和torch一样用计算等价的方式验证,numpy一部分缺失部分API,并且该API函数逻辑比较多,所以完全复现会比较繁琐 |
||
- 对`log_prob`(前置函数),log_prob的各类总和概率为1,即`paddle.exp(logprob_out).sum(1)=paddle.ones([4])` | ||
- 对`forward` | ||
- `output`为各类别概率即`output=log_prob.gather(y.unsqueeze(1), 1).slice([1], [0], [1]).squeeze()` | ||
- `loss`为`loss=nll_loss(log_prob, y)`,其中`nll_loss`已经实现 | ||
- 对`predict`,有`predict=log_prob.argmax(axis=1)` | ||
- 错误检查:`cutoff`的唯一性,数据类型,数值大于零小于`n_classes - 1` | ||
- 错误检查:`input`尺寸与`in_features`一致 | ||
|
||
|
||
# 七、可行性分析及规划排期 | ||
方案主要依赖paddle现有api组合而成 | ||
|
||
paddle.gather与torch.gather存在差异,使用paddle.take_along_axis替换实现。实现无明显难点,可以按期完成。 | ||
|
||
# 八、影响面 | ||
为独立新增API,对其他模块没有影响 | ||
|
||
# 名词解释 | ||
无 | ||
# 附件及参考资料 | ||
无 | ||
无 |
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
这个格式好像也有点问题