|
from dataclasses import dataclass |
|
from typing import Optional, Tuple |
|
import torch |
|
import torch.nn as nn |
|
from transformers import LlamaPreTrainedModel, LlamaModel |
|
from transformers.utils import ModelOutput |
|
|
|
@dataclass |
|
class MultiAspectRewardOutput(ModelOutput): |
|
""" |
|
Custom output class to return multi-aspect predictions plus final reward. |
|
|
|
Args: |
|
aspect_scores (torch.FloatTensor): shape (batch, 5) |
|
final_reward (torch.FloatTensor): shape (batch,) |
|
logits (torch.FloatTensor): shape (batch,) same as final_reward |
|
loss (torch.FloatTensor): optional scalar |
|
hidden_states (tuple(torch.FloatTensor)): optional hidden states |
|
attentions (tuple(torch.FloatTensor)): optional attentions |
|
""" |
|
aspect_scores: torch.FloatTensor = None |
|
final_reward: torch.FloatTensor = None |
|
logits: torch.FloatTensor = None |
|
loss: Optional[torch.FloatTensor] = None |
|
hidden_states: Optional[Tuple[torch.FloatTensor, ...]] = None |
|
attentions: Optional[Tuple[torch.FloatTensor, ...]] = None |
|
|
|
class LlamaFixedWeightReward(LlamaPreTrainedModel): |
|
""" |
|
A single final class that: |
|
1) Optionally takes a pretrained Llama backbone (base_llama), |
|
2) Predicts 5 aspect scores, computing MSE if 5-dim labels are provided, |
|
3) Aggregates the 5 aspect scores via fixed weights -> 1 scalar reward, |
|
4) Returns MultiAspectRewardOutput with shape [batch] in 'final_reward' and 'logits'. |
|
""" |
|
def __init__(self, config, base_llama=None, rule_weights=None): |
|
""" |
|
Args: |
|
config: LlamaConfig with num_labels=5 for multi-aspect predictions. |
|
base_llama: (optional) an already loaded LlamaModel |
|
rule_weights: (optional) A list or torch.Tensor of shape (5,) for aggregation. |
|
If None, defaults to [0.2, 0.2, 0.2, 0.2, 0.2]. |
|
""" |
|
super().__init__(config) |
|
|
|
|
|
if base_llama is not None: |
|
self.llama = base_llama |
|
else: |
|
self.llama = LlamaModel(config) |
|
|
|
|
|
|
|
self.aspect_head = nn.Linear(config.hidden_size, config.num_labels) |
|
|
|
|
|
if rule_weights is not None: |
|
w = torch.tensor(rule_weights, dtype=torch.float) |
|
else: |
|
weights = [1/config.num_labels] * config.num_labels |
|
|
|
w = torch.tensor(weights, dtype=torch.float) |
|
self.register_buffer("rule_weights", w.view(1, -1), persistent=True) |
|
|
|
self.post_init() |
|
|
|
def forward( |
|
self, |
|
input_ids=None, |
|
attention_mask=None, |
|
labels=None, |
|
**kwargs |
|
): |
|
|
|
outputs = self.llama( |
|
input_ids=input_ids, |
|
attention_mask=attention_mask, |
|
**kwargs |
|
) |
|
|
|
last_hidden = outputs.last_hidden_state |
|
|
|
|
|
pooled = last_hidden[:, -1, :] |
|
|
|
|
|
aspect_scores = self.aspect_head(pooled) |
|
|
|
aspect_scores = torch.sigmoid(aspect_scores) |
|
|
|
|
|
loss = None |
|
if labels is not None: |
|
mse_fn = nn.MSELoss() |
|
loss = mse_fn(aspect_scores, labels.float()) |
|
|
|
|
|
reward = (aspect_scores * self.rule_weights).sum(dim=-1) |
|
|
|
|
|
return MultiAspectRewardOutput( |
|
loss=loss, |
|
aspect_scores=aspect_scores, |
|
final_reward=reward, |
|
logits=reward, |
|
hidden_states=outputs.hidden_states, |
|
attentions=outputs.attentions |
|
) |