modeling_prismatic.py

"""
modeling_prismatic.py

Core HuggingFace-style PrismaticPreTrainedModel and PrismaticForConditionalGeneration class definitions.
Inherits from the default `transformers.PretrainedModel`. Meant to be standalone and self-contained,
but exactly replicate the logic in `prismatic.models.vlms.prismatic.py`.
"""

import os
import logging
import math
from dataclasses import dataclass
from functools import partial
from typing import Any, Callable, ClassVar, Dict, List, Optional, Tuple, Union

import numpy as np
import timm
import tokenizers
import torch
import torch.nn as nn
import torch.nn.functional as F
import transformers
from timm.models.vision_transformer import LayerScale
from transformers import PretrainedConfig, PreTrainedModel
from transformers.cache_utils import Cache
from transformers.modeling_outputs import ModelOutput, BaseModelOutputWithPast, CausalLMOutputWithPast
from transformers.models.llama import LlamaForCausalLM, LlamaModel, LlamaPreTrainedModel

from prismatic.training.train_utils import (
    get_current_action_mask,
    get_next_actions_mask,
)
from prismatic.vla.constants import (
    ACTION_DIM,
    ACTION_PROPRIO_NORMALIZATION_TYPE,
    ACTION_TOKEN_BEGIN_IDX,
    IGNORE_INDEX,
    NUM_ACTIONS_CHUNK,
    STOP_INDEX,
    NormalizationType,
)

from .configuration_prismatic import OpenVLAConfig, PrismaticConfig

# Set up logger
logger = logging.getLogger(__name__)


class TokenPruner(nn.Module):
    def __init__(
        self,
        config,
        num_patches,
    ):
        super().__init__()
        self.num_patches = num_patches
        self.noise_scale = None
        self.scale_factor = 1 / math.sqrt(config.hidden_size)
        # Allow fully disabling pruning/gating for ablations
        self.disabled: bool = getattr(config, "prune_disabled", False)

        # === Pruning configuration ===
        self.selection_strategy = getattr(config, "prune_selection_strategy", "coverage")
        self.coverage_temperature = getattr(config, "prune_temperature", 0.1)
        self.coverage_target = getattr(config, "prune_target_coverage", 0.9)
        self.min_keep = getattr(config, "prune_min_keep", 64)
        self.max_keep = getattr(config, "prune_max_keep", None)
        keep_bins = getattr(config, "prune_keep_bins", (64, 96, 128, 160, 192))
        self.keep_bins = tuple(keep_bins) if keep_bins is not None else None
        self.top_k = getattr(config, "prune_top_k", None)
        self.debug = getattr(config, "prune_debug", False)
        self.debug_max_logs = getattr(config, "prune_debug_max_logs", 20)
        self._debug_counter = 0
        self._last_keep_counts: Optional[torch.Tensor] = None

        # Numerical helpers
        self._coverage_eps = 1e-6

        # Aggregation strategy for prompt tokens: "max" or "logsumexp"
        # - "max": emphasizes "hit any single word" (faster, more sparse)
        # - "logsumexp": smoother aggregation for multi-word semantic combinations
        #   (e.g., "red + cube + left side"), better captures joint evidence
        self.prompt_aggregation = getattr(config, "prune_prompt_aggregation", "logsumexp")
        self.logsumexp_temperature = getattr(config, "prune_logsumexp_temperature", 1.0)

        # Optional: mean-preserving rescale for soft gating (multiply by num_patches)
        # This can help avoid weakening the vision signal when using softmax weights
        # by keeping the expected per-token scale around 1 instead of ~1/num_patches.
        # Safety clip allows capping very peaked distributions.
        self.soft_rescale_mean_preserve = getattr(
            config, "prune_soft_rescale_mean_preserve", False
        )
        self.soft_rescale_clip = getattr(config, "prune_soft_rescale_clip", None)

        # Training-time gating behavior (default: soft gating). When enabled, uses
        # Straight-Through Top-K (Gumbel-Softmax + STE) during training to better
        # match evaluation-time hard pruning behavior while keeping gradients.
        self.train_use_st_topk: bool = getattr(config, "prune_train_use_st_topk", False)
        self.train_gumbel_tau: float = getattr(config, "prune_train_gumbel_tau", 1.0)
        self.train_gumbel_tau_min: Optional[float] = getattr(config, "prune_train_gumbel_tau_min", None)

    def set_noise_scale(self, noise_scale):
        self.noise_scale = noise_scale
    
    def set_coverage_target(self, coverage_target):
        """Dynamically set the coverage target for token pruning"""
        self.coverage_target = coverage_target

    def set_disabled(self, disabled: bool):
        """Enable/disable pruning and gating entirely (keep all tokens as-is)."""
        self.disabled = bool(disabled)

    def set_train_use_st_topk(self, use_st_topk: bool):
        self.train_use_st_topk = bool(use_st_topk)

    def set_train_gumbel_tau(self, tau: float):
        self.train_gumbel_tau = float(tau)

    def rms_norm(self, hidden_states, eps=1e-6):
        input_dtype = hidden_states.dtype
        hidden_states = hidden_states.to(torch.float32)
        variance = hidden_states.pow(2).mean(-1, keepdim=True)
        hidden_states = hidden_states * torch.rsqrt(variance + eps)
        return hidden_states.to(input_dtype)

    def get_score(
        self,
        patches,
        prompts,
        q_proj_weight,
        q_proj_bias,
        k_proj_weight,
        k_proj_bias,
        num_heads,
        prompt_mask=None,
    ):
        patches = self.rms_norm(patches)
        prompts = self.rms_norm(prompts)

        # Project into the first-layer attention space (reuse LLM Q/K weights)
        queries = F.linear(patches, q_proj_weight, q_proj_bias)
        keys = F.linear(prompts, k_proj_weight, k_proj_bias)

        bsz, num_patches, _ = queries.shape
        _, num_tokens, _ = keys.shape

        head_dim = queries.shape[-1] // num_heads
        queries = queries.view(bsz, num_patches, num_heads, head_dim).permute(0, 2, 1, 3)
        keys = keys.view(bsz, num_tokens, num_heads, head_dim).permute(0, 2, 1, 3)

        attn_logits = torch.matmul(queries, keys.transpose(-1, -2)) / math.sqrt(head_dim)

        if prompt_mask is not None:
            prompt_mask = prompt_mask.bool()
            expanded_mask = prompt_mask.unsqueeze(1).unsqueeze(2)
            attn_logits = attn_logits.masked_fill(~expanded_mask, float("-inf"))

        # Aggregate signal across prompt tokens (last dimension)
        if self.prompt_aggregation == "logsumexp":
            # Log-sum-exp aggregation: smoother, captures joint evidence from multiple tokens
            # LSE(x) = log(sum(exp(x))) - captures "soft OR" over prompt tokens
            # Temperature controls smoothness: lower = closer to max, higher = closer to mean
            token_scores = torch.logsumexp(
                attn_logits / self.logsumexp_temperature, dim=-1
            ) * self.logsumexp_temperature  # (B, H, P)
        else:
            # Max aggregation: emphasizes "hit any single word"
            token_scores = attn_logits.max(dim=-1).values  # (B, H, P)
        
        # Aggregate across attention heads
        score = token_scores.mean(dim=1)  # (B, P)

        score = torch.where(torch.isfinite(score), score, torch.zeros_like(score))

        return score

    def _budgeted_keep_counts(self, score):
        device = score.device
        bsz, num_patches = score.shape

        # Fallback: fixed Top-K when requested
        if self.selection_strategy == "topk" and self.top_k is not None:
            k = min(self.top_k, num_patches)
            keep_counts = torch.full((bsz,), k, device=device, dtype=torch.int64)
            sorted_indices = score.argsort(dim=-1, descending=True)
            return keep_counts, sorted_indices

        temperature = max(float(self.coverage_temperature), self._coverage_eps)
        probs = torch.softmax(score / temperature, dim=-1)
        sorted_probs, sorted_indices = probs.sort(dim=-1, descending=True)
        cumulative = torch.cumsum(sorted_probs, dim=-1)

        target = float(self.coverage_target)
        keep_counts = (cumulative < target).sum(dim=-1) + 1

        if self.min_keep is not None:
            keep_counts = torch.maximum(
                keep_counts,
                torch.full_like(keep_counts, min(self.min_keep, num_patches)),
            )

        if self.max_keep is not None:
            keep_counts = torch.minimum(
                keep_counts,
                torch.full_like(keep_counts, min(self.max_keep, num_patches)),
            )

        keep_counts = torch.clamp(keep_counts, min=1, max=num_patches)

        if self.keep_bins:
            valid_bins = [min(num_patches, int(bin_val)) for bin_val in self.keep_bins if bin_val > 0]
            if valid_bins:
                bins = torch.tensor(sorted(set(valid_bins)), device=device, dtype=torch.int64)
                search_idx = torch.searchsorted(bins, keep_counts, right=False)
                search_idx = torch.clamp(search_idx, max=bins.numel() - 1)
                keep_counts = bins[search_idx]

        if self.debug:
            self._last_keep_counts = keep_counts.detach().to("cpu")

        return keep_counts, sorted_indices

    def score_to_mask(self, score):
        bsz, num_patches = score.shape
        mask = torch.zeros(bsz, num_patches, dtype=torch.bool, device=score.device)

        keep_counts, sorted_indices = self._budgeted_keep_counts(score)

        for batch_idx in range(bsz):
            k = int(keep_counts[batch_idx].item())
            topk_indices = sorted_indices[batch_idx, :k]
            mask[batch_idx, topk_indices] = True

        return mask

    def score_to_indices(self, score, patches):
        if self.noise_scale is not None:
            score = score + torch.rand_like(score) * self.noise_scale
        hard_score = F.one_hot(score.argmax(dim=-1), num_classes=self.num_patches)
        soft_score = torch.softmax(score, dim=-1)
        score = hard_score + soft_score - soft_score.detach()
        return score.argmax(dim=-1), score @ patches

    def forward(
        self,
        tokens,
        position_ids,
        attention_mask,
        q_proj_weight,
        q_proj_bias,
        k_proj_weight,
        k_proj_bias,
        num_heads,
    ):
        # Short-circuit: fully disable pruning/gating (identity pass-through)
        if self.disabled:
            # Populate `_last_keep_counts` so callers can log expected keeps
            bsz = tokens.shape[0]
            self._last_keep_counts = (
                torch.full((bsz,), self.num_patches, dtype=torch.int64, device=tokens.device)
                .detach()
                .to("cpu")
            )
            return tokens, position_ids, attention_mask

        bsz, seq_len, dim = tokens.shape
        cls_token, patches, task = torch.split(tokens, [1, self.num_patches, seq_len-self.num_patches-1], dim=1)
        cls_token_id, patches_id, task_id = torch.split(position_ids, [1, self.num_patches, seq_len-self.num_patches-1], dim=1)
        if attention_mask is not None:
            cls_token_mask, patches_mask, task_mask = torch.split(attention_mask, [1, self.num_patches, seq_len-self.num_patches-1], dim=1)

        # task_score = self.get_task_score(attns)
        score = self.get_score(
            patches,
            task,
            q_proj_weight,
            q_proj_bias,
            k_proj_weight,
            k_proj_bias,
            num_heads,
            prompt_mask=task_mask if attention_mask is not None else None,
        )

        if self.training:
            if self.train_use_st_topk:
                # === Straight-Through Top-K Gating (Gumbel-Softmax + STE) ===
                # 1) Compute keep counts (respect coverage_target, min/max_keep, bins)
                keep_counts, sorted_indices = self._budgeted_keep_counts(score)

                # 2) Sample Gumbel noise and compute soft probabilities
                #    logits shape: (B, P)
                gumbel = -torch.log(-torch.log(torch.rand_like(score).clamp(min=1e-6, max=1.0 - 1e-6)))
                tau = float(self.train_gumbel_tau if self.train_gumbel_tau is not None else 1.0)
                tau = max(tau, self._coverage_eps)
                logits = (score + gumbel) / tau
                probs = torch.softmax(logits, dim=-1)

                # 3) Build hard Top-K mask per example
                bsz, num_patches = score.shape
                m_hard = torch.zeros_like(score)
                for b in range(bsz):
                    k = int(keep_counts[b].item())
                    topk_idx = sorted_indices[b, :k]
                    m_hard[b, topk_idx] = 1.0

                # 4) Straight-through estimator: forward uses hard mask; backward uses soft probs
                m = m_hard + probs - probs.detach()

                # 5) Optional mean-preserving rescale so average token scale remains ~1
                if self.soft_rescale_mean_preserve:
                    # scale per batch: num_patches / k
                    scale = (torch.ones_like(keep_counts, dtype=patches.dtype) * num_patches) / keep_counts.clamp(min=1).to(patches.dtype)
                    scale = scale.view(-1, 1)  # (B, 1)
                    if self.soft_rescale_clip is not None:
                        scale = torch.clamp(scale, max=float(self.soft_rescale_clip))
                    patches = patches * (m.unsqueeze(-1)) * scale.unsqueeze(-1)
                else:
                    patches = patches * (m.unsqueeze(-1))

                # Log keep counts for monitoring
                if self.debug:
                    self._last_keep_counts = keep_counts.detach().to("cpu")

                tokens = torch.cat([cls_token, patches, task], dim=1)
                position_ids = torch.cat([cls_token_id, patches_id, task_id], dim=1)
                if attention_mask is not None:
                    attention_mask = torch.cat([cls_token_mask, patches_mask, task_mask], dim=1)
            else:
                # === Default: Soft gating during training (no hard pruning) ===
                weights = torch.softmax(
                    score / max(self.coverage_temperature, self._coverage_eps), dim=-1
                )
                if self.soft_rescale_mean_preserve:
                    # Mean-preserving rescale: E[weights] ~ 1 so average token scale is unchanged.
                    weights = weights * patches.shape[1]
                    if self.soft_rescale_clip is not None:
                        # Optional safety: cap very large scales if distribution is extremely peaked
                        weights = torch.clamp(weights, max=float(self.soft_rescale_clip))
                patches = patches * weights.unsqueeze(-1)

                # Optionally compute and stash keep-counts for logging/monitoring
                if self.debug:
                    keep_counts, _ = self._budgeted_keep_counts(score)
                    self._last_keep_counts = keep_counts.detach().to("cpu")
                tokens = torch.cat([cls_token, patches, task], dim=1)
                position_ids = torch.cat([cls_token_id, patches_id, task_id], dim=1)
                if attention_mask is not None:
                    attention_mask = torch.cat([cls_token_mask, patches_mask, task_mask], dim=1)

        else:
            mask = self.score_to_mask(score)

            patches = patches[mask].view(bsz, -1, dim)
            tokens = torch.cat([cls_token, patches, task], dim=1)

            patches_id = patches_id[mask].view(bsz, -1)
            position_ids = torch.cat([cls_token_id, patches_id, task_id], dim=1)

            if attention_mask is not None:
                patches_mask = patches_mask[mask].view(bsz, -1)
                attention_mask = torch.cat([cls_token_mask, patches_mask, task_mask], dim=1)

            if self.debug and self._debug_counter < self.debug_max_logs:
                keep_counts = self._last_keep_counts
                if keep_counts is not None:
                    keep_counts = keep_counts.to(torch.float32)
                    logger.info(
                        "TokenPruner debug | keep_counts min=%.0f max=%.0f mean=%.2f | target=%.2f | temp=%.3f | bins=%s",
                        keep_counts.min().item(),
                        keep_counts.max().item(),
                        keep_counts.mean().item(),
                        float(self.coverage_target),
                        float(self.coverage_temperature),
                        self.keep_bins,
                    )
                self._debug_counter += 1

        return tokens, position_ids, attention_mask


class PrunedLlamaModel(LlamaModel):
    def __init__(self, config, num_patches):
        super().__init__(config)

        self.pruner = TokenPruner(
            config,
            num_patches,
        )

    def forward(
        self,
        input_ids: torch.LongTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple, BaseModelOutputWithPast]:
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if (input_ids is None) ^ (inputs_embeds is not None):
            raise ValueError(
                "You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one"
            )

        if self.gradient_checkpointing and self.training and use_cache:
            logger.warning_once(
                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
            )
            use_cache = False

        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input_ids)

        # embed positions
        hidden_states = inputs_embeds

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        next_decoder_cache = None

        past_seen_tokens = 0

        cache_position = torch.arange(
            past_seen_tokens, past_seen_tokens + hidden_states.shape[1], device=inputs_embeds.device
        )

        position_ids = cache_position.unsqueeze(0).expand(hidden_states.shape[0], -1)

        first_layer_attn = self.layers[0].self_attn
        hidden_states, position_ids, attention_mask = self.pruner(
            hidden_states,
            position_ids,
            attention_mask,
            first_layer_attn.q_proj.weight,
            first_layer_attn.q_proj.bias,
            first_layer_attn.k_proj.weight,
            first_layer_attn.k_proj.bias,
            first_layer_attn.num_heads,
        )

        past_seen_tokens = 0

        cache_position = torch.arange(
            past_seen_tokens, past_seen_tokens + hidden_states.shape[1], device=hidden_states.device
        )

        causal_mask = self._update_causal_mask(attention_mask, hidden_states, cache_position, past_seen_tokens)

        for decoder_layer in self.layers:

            if output_hidden_states:
                all_hidden_states += (hidden_states,)

            if self.gradient_checkpointing and self.training:
                layer_outputs = self._gradient_checkpointing_func(
                    decoder_layer.__call__,
                    hidden_states,
                    causal_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    use_cache,
                    cache_position,
                )
            else:
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=causal_mask,
                    position_ids=position_ids,
                    past_key_value=past_key_values,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                    cache_position=cache_position,
                )

            hidden_states = layer_outputs[0]

            if use_cache:
                next_decoder_cache = layer_outputs[2 if output_attentions else 1]

            if output_attentions:
                all_self_attns += (layer_outputs[1],)

        hidden_states = self.norm(hidden_states)

        # add hidden states from the last decoder layer
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        next_cache = None
        if use_cache:
            next_cache = (
                next_decoder_cache.to_legacy_cache() if isinstance(next_decoder_cache, Cache) else next_decoder_cache
            )
        if not return_dict:
            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
        )


class PrunedLlamaForCausalLM(LlamaForCausalLM):
    def __init__(self, config, num_patches):
        super(LlamaPreTrainedModel, self).__init__(config)
        self.model = PrunedLlamaModel(config, num_patches)
        self.vocab_size = config.vocab_size
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

        # Initialize weights and apply final processing
        self.post_init()

    def forward(
        self,
        input_ids: torch.LongTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple, CausalLMOutputWithPast]:
        r"""
        Args:
            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.

        Returns:

        Example:

        ```python
        >>> from transformers import AutoTokenizer, LlamaForCausalLM

        >>> model = LlamaForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
        >>> tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")

        >>> prompt = "Hey, are you conscious? Can you talk to me?"
        >>> inputs = tokenizer(prompt, return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
        ```"""
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
        outputs = self.model(
            input_ids=input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            cache_position=cache_position,
        )

        hidden_states = outputs[0]
        if self.config.pretraining_tp > 1:
            lm_head_slices = self.lm_head.weight.split(self.vocab_size // self.config.pretraining_tp, dim=0)
            logits = [F.linear(hidden_states, lm_head_slices[i]) for i in range(self.config.pretraining_tp)]
            logits = torch.cat(logits, dim=-1)
        else:
            logits = self.lm_head(hidden_states)
        logits = logits.float()

        loss = None
        if labels is not None:
            # Shift so that tokens < n predict n
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., -shift_logits.shape[-2]:].contiguous()
            # Flatten the tokens
            shift_logits = shift_logits.view(-1, self.config.vocab_size)
            shift_labels = shift_labels.view(-1)
            # Enable model parallelism
            shift_labels = shift_labels.to(shift_logits.device)
            loss = F.cross_entropy(shift_logits, shift_labels)

        if not return_dict:
            output = (logits,) + outputs[1:]
            return (loss,) + output if loss is not None else output

        return CausalLMOutputWithPast(
            loss=loss,
            logits=logits,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


# === Utility Functions for Monkey-Patching ===
def unpack_tuple(fn: Callable[[Any], Tuple[Any]]) -> Callable[[Any], Any]:
    def wrapper(*args: Any, **kwargs: Any) -> Any:
        result = fn(*args, **kwargs)
        return result[0] if isinstance(result, tuple) else result

    return wrapper


# HF Transformers overwrites parameters with names containing `gamma`; we're going to patch VisionBackbone.LayerScale.
#   =>> TIMM :: https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/vision_transformer.py#L109
#   =>> Transformers :: https://github.com/huggingface/transformers/blob/main/src/transformers/modeling_utils.py#L3960
def _ls_new_forward(self, x: torch.Tensor) -> torch.Tensor:
    return x.mul_(self.scale_factor) if self.inplace else x * self.scale_factor


def ls_apply_patch(ls_module: LayerScale):
    ls_module.scale_factor = nn.Parameter(ls_module.gamma.clone())
    ls_module.forward = _ls_new_forward.__get__(ls_module, LayerScale)
    del ls_module.gamma


# === Prismatic Vision Backbone (nn.Module) Definitions (w/ Fused Backbone Support) ===
class PrismaticVisionBackbone(nn.Module):
    """
    Vision backbone for Prismatic models that handles image feature extraction.

    Supports both single backbone (e.g., SigLIP) and fused backbone (e.g., SigLIP + DINOv2) configurations.
    For fused backbones, features from both models are concatenated along the feature dimension.
    """

    def __init__(
        self,
        use_fused_vision_backbone: bool,
        image_sizes: List[int],
        timm_model_ids: List[str],
        timm_override_act_layers: List[Optional[str]],
    ) -> None:
        """
        Initialize the vision backbone.

        Args:
            use_fused_vision_backbone: Whether to use two backbones and fuse their features
            image_sizes: List of image sizes for each backbone
            timm_model_ids: List of TIMM model IDs to use for each backbone
            timm_override_act_layers: List of activation layer overrides for each backbone
        """
        super().__init__()
        self.use_fused_vision_backbone = use_fused_vision_backbone
        self.num_images_in_input = 2  # Default value, can be overridden later

        # Validate number of (fused) vision backbones
        if len(timm_model_ids) > 2:
            raise ValueError("Prismatic models only support up to 2 (fused) vision backbones!")

        # Create primary featurizer
        self.featurizer = self._create_featurizer(
            model_id=timm_model_ids[0], img_size=image_sizes[0], act_layer=timm_override_act_layers[0]
        )
        self.embed_dim = self.featurizer.embed_dim

        # Create secondary featurizer if using fused backbone
        if self.use_fused_vision_backbone:
            self.fused_featurizer = self._create_featurizer(
                model_id=timm_model_ids[1], img_size=image_sizes[1], act_layer=timm_override_act_layers[1]
            )
            self.embed_dim += self.fused_featurizer.embed_dim

        # Patch LayerScale modules for HF compatibility
        self._patch_layer_scales()

    def _create_featurizer(self, model_id: str, img_size: int, act_layer: Optional[str]) -> nn.Module:
        """
        Create a TIMM-based featurizer model with appropriate configurations.

        Args:
            model_id: The TIMM model ID to load
            img_size: Input image size for the model
            act_layer: Override for the activation layer type

        Returns:
            A configured featurizer model
        """
        featurizer = timm.create_model(
            model_id,
            pretrained=False,
            num_classes=0,
            img_size=img_size,
            act_layer=act_layer,
        )

        # Monkey-patch the forward function to extract the second-to-last layer features
        num_blocks = len(featurizer.blocks)
        featurizer.forward = unpack_tuple(partial(featurizer.get_intermediate_layers, n={num_blocks - 2}))

        return featurizer

    def _patch_layer_scales(self) -> None:
        """
        Patch all LayerScale modules to be compatible with HF's parameter naming.

        HF Transformers overwrites parameters with names containing 'gamma',
        so we need to rename and modify the forward method.
        """
        # Patch primary featurizer
        for module in self.featurizer.modules():
            if isinstance(module, LayerScale):
                ls_apply_patch(module)

        # Patch secondary featurizer if it exists
        if self.use_fused_vision_backbone:
            for module in self.fused_featurizer.modules():
                if isinstance(module, LayerScale):
                    ls_apply_patch(module)

    def get_num_patches(self) -> int:
        """
        Returns the number of vision patches output by the vision backbone.

        Returns:
            Number of patches per image
        """
        return self.featurizer.patch_embed.num_patches

    def get_num_images_in_input(self) -> int:
        """
        Returns the number of input images for the vision backbone.

        Returns:
            Number of images expected in the input
        """
        return self.num_images_in_input

    def set_num_images_in_input(self, num_images_in_input: int) -> None:
        """
        Sets the number of input images for the vision backbone.

        Args:
            num_images_in_input: Number of images to expect in the input
        """
        self.num_images_in_input = num_images_in_input

    def forward(self, pixel_values: torch.Tensor) -> torch.Tensor:
        """
        Implements the forward pass for the vision backbone.

        If `self.use_fused_vision_backbone == True`, uses both SigLIP and DINOv2 transformers to extract visual features
        (otherwise uses SigLIP only). Allows multi-image inputs (but only for fused vision backbone).

        Args:
            pixel_values (torch.Tensor): Pixels for input image(s), (B, C, H, W).
        """
        if self.num_images_in_input == 1:
            if not self.use_fused_vision_backbone:
                return self.featurizer(pixel_values)

            # Split `pixel_values :: [bsz, 2 * 3, resolution, resolution]` =>> featurize =>> channel stack
            img, img_fused = torch.split(pixel_values, [3, 3], dim=1)
            patches, patches_fused = self.featurizer(img), self.fused_featurizer(img_fused)

            return torch.cat([patches, patches_fused], dim=2)

        else:
            assert self.use_fused_vision_backbone, "Multi-image inputs require using fused backbone!"

            # Split `pixel_values` into individual images (each with 6 channels: 3 for SigLIP + 3 for DINOv2)
            images = torch.split(pixel_values, [6] * self.num_images_in_input, dim=1)

            # Process each image and collect patches
            all_patches = []
            for img in images:
                # Split each image further into two stacks of channels (each with 3 channels)
                img_regular, img_fused = torch.split(img, [3, 3], dim=1)

                # Get patches from both SigLIP and DINOv2 vision transformers
                patches = self.featurizer(img_regular)
                patches_fused = self.fused_featurizer(img_fused)

                # Concatenate SigLIP and DINOv2 patches along the hidden dimension
                combined_patches = torch.cat([patches, patches_fused], dim=2)
                all_patches.append(combined_patches)

            # Concatenate all patches along the patch dimension
            return torch.cat(all_patches, dim=1)


# === Prismatic Projector (nn.Module) Definitions ===
class PrismaticProjector(nn.Module):
    def __init__(self, use_fused_vision_backbone: bool, vision_dim: int, llm_dim: int) -> None:
        super().__init__()
        self.use_fused_vision_backbone = use_fused_vision_backbone
        self.vision_dim, self.llm_dim = vision_dim, llm_dim

        # Switch on `use_fused_vision_backbone` =>> use slightly different MLPs and projection factors!
        if not self.use_fused_vision_backbone:
            self.fc1 = nn.Linear(self.vision_dim, self.llm_dim, bias=True)
            self.fc2 = nn.Linear(self.llm_dim, self.llm_dim, bias=True)
            self.act_fn1 = nn.GELU()
        else:
            initial_projection_dim = 4 * vision_dim
            self.fc1 = nn.Linear(self.vision_dim, initial_projection_dim, bias=True)
            self.fc2 = nn.Linear(initial_projection_dim, self.llm_dim, bias=True)
            self.fc3 = nn.Linear(self.llm_dim, self.llm_dim, bias=True)
            self.act_fn1 = nn.GELU()
            self.act_fn2 = nn.GELU()

    def forward(self, img_patches: torch.Tensor) -> torch.Tensor:
        if not self.use_fused_vision_backbone:
            projected_features = self.fc1(img_patches)
            projected_features = self.act_fn1(projected_features)
            projected_features = self.fc2(projected_features)
        else:
            projected_features = self.fc1(img_patches)
            projected_features = self.act_fn1(projected_features)
            projected_features = self.fc2(projected_features)
            projected_features = self.act_fn2(projected_features)
            projected_features = self.fc3(projected_features)

        return projected_features


# === Main HF Class Definitions ===
@dataclass
class PrismaticCausalLMOutputWithPast(ModelOutput):
    """Base class for Prismatic casual (visually-conditioned) language model outputs; also exposes visual features."""

    loss: Optional[torch.FloatTensor] = None
    logits: torch.FloatTensor = None
    past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None
    hidden_states: Optional[Tuple[torch.FloatTensor, ...]] = None
    attentions: Optional[Tuple[torch.FloatTensor]] = None

    # Additions for VLMs
    projector_features: Optional[torch.FloatTensor] = None


class PrismaticPreTrainedModel(PreTrainedModel):
    config_class: PretrainedConfig = PrismaticConfig
    base_model_prefix: str = "model"
    supports_gradient_checkpointing: bool = True

    _no_split_modules: ClassVar[List[str]] = ["PrismaticProjector"]
    _skip_keys_device_placement: str = "past_key_values"
    _supports_flash_attn_2: bool = False

    def _init_weights(self, module: nn.Module) -> None:
        # Important :: this HF ported version is *not* meant for training from scratch; only inference and fine-tuning!
        #   => As such, this init_weights code is not correct; if training VLMs from scratch, use the main codebase at
        #      https://github.com/TRI-ML/prismatic-vlms
        std = (
            self.config.initializer_range
            if hasattr(self.config, "initializer_range")
            else self.config.text_config.initializer_range
        )

        if hasattr(module, "class_embedding"):
            module.class_embedding.data.normal_(mean=0.0, std=std)

        if isinstance(module, (nn.Linear, nn.Conv2d)):
            module.weight.data.normal_(mean=0.0, std=std)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.normal_(mean=0.0, std=std)
            if module.padding_idx is not None:
                module.weight.data[module.padding_idx].zero_()

    @property
    def _supports_sdpa(self) -> bool:
        """Check LLM supports SDPA Attention"""
        return self.language_model._supports_sdpa


class PrismaticForConditionalGeneration(PrismaticPreTrainedModel):
    def __init__(self, config: PrismaticConfig) -> None:
        super().__init__(config)

        # [Validation] Lightweight Validate on `config` Fields + Dependency Versions
        if config.use_fused_vision_backbone is None:
            raise ValueError("Missing config field `use_fused_vision_backbone`")

        if timm.__version__ not in {"0.9.10", "0.9.11", "0.9.12", "0.9.16"}:
            raise NotImplementedError(
                "TIMM Version must be >= 0.9.10 and < 1.0.0 (breaking); please raise a GitHub Issue "
                "if you urgently need support for latest TIMM versions."
            )

        if (transformers.__version__ != "4.40.1") or (tokenizers.__version__ != "0.19.1"):
            logger.warning(
                f"Expected `transformers==4.40.1` and `tokenizers==0.19.1` but got "
                f"`transformers=={transformers.__version__}` and `tokenizers=={tokenizers.__version__}`; "
                f"there might be inference-time regressions due to dependency changes. If in doubt, please"
                f"use the above versions."
            )

        # Instantiate PrismaticVisionBackbone (w/ Potential Fused Backbone)
        self.vision_backbone = PrismaticVisionBackbone(
            config.use_fused_vision_backbone, config.image_sizes, config.timm_model_ids, config.timm_override_act_layers
        )

        # Create Multimodal Projector
        self.projector = PrismaticProjector(
            config.use_fused_vision_backbone,
            vision_dim=self.vision_backbone.embed_dim,
            llm_dim=config.text_config.hidden_size,
        )

        # Instantiate LLM Backbone
        num_patches = self.vision_backbone.get_num_patches() * self.vision_backbone.get_num_images_in_input()
        # Bridge prune-related config fields from outer config to text_config so TokenPruner can see them
        for key in (
            "prune_selection_strategy",
            "prune_temperature",
            "prune_target_coverage",
            "prune_min_keep",
            "prune_max_keep",
            "prune_keep_bins",
            "prune_top_k",
            "prune_debug",
            "prune_debug_max_logs",
            "prune_prompt_aggregation",
            "prune_logsumexp_temperature",
            "prune_soft_rescale_mean_preserve",
            "prune_soft_rescale_clip",
            "prune_disabled",
        ):
            if hasattr(config, key) and not hasattr(config.text_config, key):
                setattr(config.text_config, key, getattr(config, key))
        self.language_model = PrunedLlamaForCausalLM(config.text_config, num_patches)
        self.vocab_size = config.text_config.vocab_size
        self.pad_token_id = config.pad_token_id
        self.llm_dim = config.text_config.hidden_size

        # HF Boilerplate =>> initializes weights via `_init_weights()` and sets gradient checkpointing
        self.post_init()

    # === `PreTrainedModel` Boilerplate ===
    def get_input_embeddings(self) -> nn.Module:
        return self.language_model.get_input_embeddings()

    def set_input_embeddings(self, value: nn.Module) -> None:
        self.language_model.set_input_embeddings(value)

    def get_output_embeddings(self) -> nn.Module:
        return self.language_model.get_output_embeddings()

    def set_output_embeddings(self, new_embeddings: nn.Module) -> None:
        self.language_model.set_output_embeddings(new_embeddings)

    def get_decoder(self) -> nn.Module:
        return self.language_model.get_decoder()

    def set_decoder(self, decoder: nn.Module) -> None:
        self.language_model.set_decoder(decoder)

    def tie_weights(self) -> None:
        self.language_model.tie_weights()  # Note: `Llama-2` and `Mistral` don't tie weights (no-op)

    def resize_token_embeddings(
        self, new_num_tokens: Optional[int] = None, pad_to_multiple_of: Optional[int] = None
    ) -> nn.Embedding:
        updated_embeddings = self.language_model.resize_token_embeddings(new_num_tokens, pad_to_multiple_of)

        # Update config/instance variables
        self.config.text_config.vocab_size = updated_embeddings.num_embeddings
        self.vocab_size = updated_embeddings.num_embeddings

        return updated_embeddings

    def _replace_input_embeddings(self, input_embeddings, all_actions_mask, noisy_action_features):
        """
        Replace embeddings in input_embeddings at positions where all_actions_mask is True
        with embeddings from noisy_action_features, using vectorized operations.

        Args:
            input_embeddings: Tensor of shape (B, S, D)
            all_actions_mask: Boolean tensor of shape (B, S)
            noisy_action_features: Tensor of shape (B, K, D) where K is the number of True values in mask per sample

        Returns:
            Modified input_embeddings tensor
        """
        # Clone input to avoid modifying the original tensor
        new_input_embeddings = input_embeddings.clone()

        # Create a tensor with the same shape of input_embeddings to hold the noisy action features
        repositioned_noisy_action_features = torch.zeros_like(input_embeddings)

        # Create batch indices for splicing
        batch_indices = torch.arange(input_embeddings.shape[0], device=input_embeddings.device)
        batch_indices = batch_indices.unsqueeze(1).expand(-1, noisy_action_features.shape[1])

        # Get indices where mask is True for each sample
        masked_indices = torch.stack([torch.where(mask)[0] for mask in all_actions_mask])

        # Move the noisy action features into their correct positions
        repositioned_noisy_action_features[batch_indices, masked_indices] = noisy_action_features

        # Combine original input embeddings and noisy action embeddings using the mask
        new_input_embeddings = torch.where(
            all_actions_mask.unsqueeze(-1), repositioned_noisy_action_features, new_input_embeddings
        )

        return new_input_embeddings

    def _process_action_masks(self, labels):
        """Helper to get action masks from labels"""
        current_action_mask = get_current_action_mask(labels)
        next_actions_mask = get_next_actions_mask(labels)
        all_actions_mask = current_action_mask | next_actions_mask  # (B, seq_len)
        return all_actions_mask

    def _process_vision_features(self, pixel_values, language_embeddings=None, use_film=False):
        """Process vision features with optional FiLM conditioning"""
        if use_film:
            # FiLM: Infuse language inputs into visual features
            patch_features = self.vision_backbone(pixel_values, language_embeddings)  # (bsz, 256 * num_images, D)
        else:
            patch_features = self.vision_backbone(pixel_values)  # (bsz, 256 * num_images, D)

        # Project patch embeddings into language embedding space
        return self.projector(patch_features)

    def _process_proprio_features(self, projected_patch_embeddings, proprio, proprio_projector):
        """Process proprioceptive features and append to vision features"""
        if proprio_projector is not None and proprio is not None:
            # projected_patch_embeddings: (bsz, num_patches * num_images, llm_dim)
            # proprio: (bsz, proprio_dim) or (propro_dim,)
            proprio = proprio.reshape(projected_patch_embeddings.shape[0], -1)  # (bsz, proprio_dim)
            proprio_features = proprio_projector(proprio)  # (bsz, llm_dim)
            proprio_features = proprio_features.unsqueeze(dim=1)  # (bsz, 1, llm_dim)
            # For simplicity, just append proprio token to the end of projected vision patch tokens
            return torch.cat((projected_patch_embeddings, proprio_features), dim=1)
        return projected_patch_embeddings

    def _build_multimodal_attention(self, input_embeddings, projected_patch_embeddings, attention_mask):
        """Build multimodal embeddings and attention mask"""
        # Update attention mask
        projected_patch_attention_mask = None
        if attention_mask is not None:
            projected_patch_attention_mask = torch.full(
                (projected_patch_embeddings.shape[0], projected_patch_embeddings.shape[1]),
                fill_value=True,
                dtype=attention_mask.dtype,
                device=attention_mask.device,
            )

        # Build multimodal embeddings & attention mask; insert embeddings after <BOS> token (1:)
        multimodal_embeddings = torch.cat(
            [input_embeddings[:, :1, :], projected_patch_embeddings, input_embeddings[:, 1:, :]], dim=1
        )

        multimodal_attention_mask = None
        if attention_mask is not None:
            multimodal_attention_mask = torch.cat(
                [attention_mask[:, :1], projected_patch_attention_mask, attention_mask[:, 1:]], dim=1
            )

        return multimodal_embeddings, multimodal_attention_mask

    def _build_multimodal_labels(self, labels, projected_patch_embeddings):
        """Build multimodal labels with IGNORE_INDEX for patch embeddings"""
        if labels is not None:
            projected_patch_labels = torch.full(
                (projected_patch_embeddings.shape[0], projected_patch_embeddings.shape[1]),
                fill_value=IGNORE_INDEX,
                dtype=labels.dtype,
                device=labels.device,
            )
            return torch.cat([labels[:, :1], projected_patch_labels, labels[:, 1:]], dim=1)
        return None

    # === Core Prismatic VLM `forward()` Logic ===
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        output_projector_features: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        proprio=None,
        proprio_projector=None,
        noisy_actions=None,
        noisy_action_projector=None,
        diffusion_timestep_embeddings=None,
        use_film: bool = False,
    ) -> Union[Tuple, PrismaticCausalLMOutputWithPast]:
        """Run a forward pass through the VLM, returning a PrismaticCausalLMOutputWithPast instance."""
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        output_projector_features = output_projector_features if output_projector_features is not None else False
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        # Respect `use_cache` only if not training (even if `gradient_checkpointing` is off)
        use_cache = use_cache and not self.training

        # Instantiate Placeholder for Projector Features
        projected_patch_embeddings = None

        # === Handle Generation with Cache (`input_ids.shape[1] == 1`) =>> requires `past_keys_values` ===
        if input_ids.shape[1] == 1:
            assert input_ids.shape[0] == 1, "Generation is only currently supported for batch size of 1!"
            assert past_key_values is not None, "You must provide `past_key_values` during cached generation!"
            assert labels is None, "Unexpected key `labels` provided during cached generation!"

            language_model_output = self.language_model(
                input_ids=input_ids,
                attention_mask=None,
                position_ids=None,
                past_key_values=past_key_values,
                inputs_embeds=None,
                labels=None,
                use_cache=use_cache,
                output_attentions=output_attentions,
                output_hidden_states=output_hidden_states,
                return_dict=return_dict,
            )

        # === Handle Unimodal Forward ===
        elif pixel_values is None:
            assert (input_ids is not None) and (inputs_embeds is None), "Missing `input_ids` in language-only forward!"
            assert past_key_values is None, "Unexpected key `past_key_values` provided during language-only forward!"

            language_model_output = self.language_model(
                input_ids=input_ids,
                attention_mask=attention_mask,
                position_ids=None,
                past_key_values=None,
                inputs_embeds=None,
                labels=labels,
                use_cache=use_cache,
                output_attentions=output_attentions,
                output_hidden_states=output_hidden_states,
                return_dict=return_dict,
            )

        # === Handle Multimodal Forward ===
        elif (input_ids.shape[0] == pixel_values.shape[0]) or (inputs_embeds.shape[0] == pixel_values.shape[0]):
            assert past_key_values is None, "Unexpected key `past_key_values` provided during multimodal forward!"

            # Get input embeddings (from language model embeddings)
            input_embeddings = self.get_input_embeddings()(input_ids)  # (B, seq_len, D)

            # Extract action masks
            all_actions_mask = self._process_action_masks(labels)

            # Extract the language portion of the input embeddings (i.e. remove the action tokens portion)
            language_embeddings = input_embeddings[~all_actions_mask].reshape(
                input_embeddings.shape[0], -1, input_embeddings.shape[2]
            )  # (B, lang_seq_len, llm_dim)

            # Get visual features
            projected_patch_embeddings = self._process_vision_features(pixel_values, language_embeddings, use_film)

            # Add proprioceptive state if provided
            projected_patch_embeddings = self._process_proprio_features(
                projected_patch_embeddings, proprio, proprio_projector
            )

            # [Diffusion] Add diffusion timestep embedding if provided
            if diffusion_timestep_embeddings is not None:
                # For simplicity, just append diffusion timestep embedding to the end of projected vision patch tokens
                projected_patch_embeddings = torch.cat(
                    (projected_patch_embeddings, diffusion_timestep_embeddings), dim=1
                )

            # Process action embeddings
            if noisy_actions is not None:
                # Get mask corresponding to all action tokens
                all_actions_mask = self._process_action_masks(labels)

                # Reshape noisy actions into individual action tokens
                # noisy_actions: (B, chunk_len, action_dim) -> (B, chunk_len * action_dim, 1)
                B = noisy_actions.shape[0]
                noisy_actions = noisy_actions.reshape(B, -1).unsqueeze(-1)

                # Project noisy action tokens into language model embedding space
                noisy_action_features = noisy_action_projector(noisy_actions)  # (B, chunk_len * action_dim, llm_dim)

                # Replace embeddings of the action tokens with noisy action embeddings
                input_embeddings = self._replace_input_embeddings(
                    input_embeddings, all_actions_mask, noisy_action_features
                )
            else:
                # Replace the embeddings of the action tokens with zeros
                # (Later on, the positional embeddings will be added to them)
                all_actions_mask = all_actions_mask.unsqueeze(-1)  # (B, seq_len, 1)
                input_embeddings = input_embeddings * ~all_actions_mask

            # Build multimodal embeddings & attention mask
            multimodal_embeddings, multimodal_attention_mask = self._build_multimodal_attention(
                input_embeddings, projected_patch_embeddings, attention_mask
            )

            # Build labels for multimodal sequence if needed
            multimodal_labels = self._build_multimodal_labels(labels, projected_patch_embeddings)

            # Dispatch to language model
            language_model_output = self.language_model(
                input_ids=None,
                attention_mask=multimodal_attention_mask,
                position_ids=None,
                past_key_values=None,
                inputs_embeds=multimodal_embeddings,
                labels=multimodal_labels,
                use_cache=use_cache,
                output_attentions=output_attentions,
                output_hidden_states=output_hidden_states,
                return_dict=return_dict,
            )

        # === Otherwise =>> Assume Invalid! ===
        elif (input_ids.shape[0] != pixel_values.shape[0]) or (inputs_embeds.shape[0] != pixel_values.shape[0]):
            raise ValueError("Non-homogenous batch of (text, image) input -- forward() does not support mixed batches!")

        else:
            raise ValueError(
                "Invalid PrismaticForConditionalGeneration `forward()` call with provided arguments:\n"
                f"=> `input_ids` = {input_ids is not None}\n"
                f"=> `attention_mask` = {attention_mask is not None}\n"
                f"=> `pixel_values` = {pixel_values is not None}\n"
                f"=> `labels` = {labels is not None}\n"
                f"=> `input_embeds` = {inputs_embeds is not None}\n"
                f"=> `past_key_values` = {past_key_values is not None}\n"
                f"=> `use_cache` = {use_cache}"
            )

        # Unpack `language_model_output` and return PrismaticCausalLMOutputWithPast (or tuple if not `return_dict`)
        if not return_dict:
            if output_projector_features and (projected_patch_embeddings is not None):
                return *language_model_output, projected_patch_embeddings

            return language_model_output

        return PrismaticCausalLMOutputWithPast(
            loss=language_model_output.loss,
            logits=language_model_output.logits,
            past_key_values=language_model_output.past_key_values,
            hidden_states=language_model_output.hidden_states,
            attentions=language_model_output.attentions,
            projector_features=projected_patch_embeddings,
        )

    # === GenerationMixin Methods ===
    def prepare_inputs_for_generation(
        self,
        input_ids: Optional[torch.Tensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        **kwargs: str,
    ) -> Dict[str, torch.Tensor]:
        """Borrowed from `LlamaForCausalLM` and simplified for batch size = 1; mirrors original PrismaticVLM logic."""
        if ((input_ids is not None) and (input_ids.shape[0] > 1)) or (
            (inputs_embeds is not None) and (inputs_embeds.shape[0] > 1)
        ):
            raise ValueError("Generation with batch size > 1 is not currently supported!")

        # Handle `past_key_values` (cache) =>> assume `input_ids` just has unprocessed tokens
        if past_key_values is not None:
            input_ids = input_ids[:, -1:]

        # If `input_embeds` are passed, we only want to use them in the 1st generation step
        if inputs_embeds is not None and past_key_values is None:
            model_inputs = {"input_embeds": inputs_embeds}
        else:
            model_inputs = {"input_ids": input_ids}

        # Make sure `pixel_values` are preserved in `model_inputs`
        model_inputs.update(
            {
                "attention_mask": attention_mask,
                "pixel_values": pixel_values,
                "past_key_values": past_key_values,
                "use_cache": kwargs.get("use_cache"),
            }
        )

        return model_inputs

    # Defer to Language Model (all handle this differently, with different return types)
    def _reorder_cache(self, *args, **kwargs) -> Any:
        return self.language_model._reorder_cache(*args, **kwargs)


class OpenVLAForActionPrediction(PrismaticForConditionalGeneration):
    config_class: PretrainedConfig = OpenVLAConfig

    def __init__(self, config: OpenVLAConfig) -> None:
        super().__init__(config)
        self.norm_stats = config.norm_stats

        # Compute action bins
        self.bins = np.linspace(-1, 1, config.n_action_bins)
        self.bin_centers = (self.bins[:-1] + self.bins[1:]) / 2.0

        # Compute vocab size for de-tokenization -- revert added "multiple of"
        self.vocab_size = self.config.text_config.vocab_size - self.config.pad_to_multiple_of
    
    def set_num_images_in_input(self, num_images_in_input):
        self.vision_backbone.set_num_images_in_input(num_images_in_input)
        self.language_model.model.pruner.num_patches = self.vision_backbone.get_num_patches() * self.vision_backbone.get_num_images_in_input()

    def get_num_patches(self):
        return self.vision_backbone.get_num_patches() * self.vision_backbone.get_num_images_in_input()

    def _prepare_input_for_action_prediction(self, input_ids, attention_mask):
        """Prepares input for action prediction by adding necessary tokens"""
        # Add (ACTION_DIM * NUM_ACTIONS_CHUNK) placeholder tokens to input_ids to simulate action tokens
        placeholder_action_token_ids = (
            torch.ones((input_ids.shape[0], ACTION_DIM * NUM_ACTIONS_CHUNK)).to(input_ids.device).to(input_ids.dtype)
        )
        input_ids = torch.cat([input_ids, placeholder_action_token_ids], dim=-1)

        # Add stop token to sequence (needed in non-causal bi-directional self-attention, as it appears at train time)
        stop_token_id = torch.ones((input_ids.shape[0], 1)).to(input_ids.device).to(input_ids.dtype) * STOP_INDEX
        input_ids = torch.cat([input_ids, stop_token_id], dim=-1)

        # Extend the attention mask to fit the new shape of input
        # Note: Only batch size == 1 supported right now
        mask_extension = (
            torch.ones((attention_mask.shape[0], input_ids.shape[-1] - attention_mask.shape[-1]))
            .to(attention_mask.device)
            .to(attention_mask.dtype)
        )
        attention_mask = torch.cat([attention_mask, mask_extension], dim=-1)

        return input_ids, attention_mask

    def _prepare_labels_for_action_prediction(self, labels, input_ids):
        """Creates labels tensor for action prediction if not provided"""
        # Extend labels tensor with fake action labels
        ARBITRARY_ACTION_TOKEN_IDX = ACTION_TOKEN_BEGIN_IDX + 1
        labels_extension = (
            torch.ones((labels.shape[0], input_ids.shape[-1] - labels.shape[-1])).to(labels.device).to(labels.dtype)
            * ARBITRARY_ACTION_TOKEN_IDX
        )
        labels = torch.cat([labels, labels_extension], dim=-1)

        # Replace last label token with stop token
        labels[:, -1] = STOP_INDEX

        return labels

    def _unnormalize_actions(self, normalized_actions, unnorm_key=None):
        """Unnormalize actions using dataset statistics"""
        action_norm_stats = self.get_action_stats(unnorm_key)

        if ACTION_PROPRIO_NORMALIZATION_TYPE == NormalizationType.BOUNDS:
            mask = action_norm_stats.get("mask", np.ones_like(action_norm_stats["min"], dtype=bool))
            action_high, action_low = np.array(action_norm_stats["max"]), np.array(action_norm_stats["min"])
        elif ACTION_PROPRIO_NORMALIZATION_TYPE == NormalizationType.BOUNDS_Q99:
            mask = action_norm_stats.get("mask", np.ones_like(action_norm_stats["q01"], dtype=bool))
            action_high, action_low = np.array(action_norm_stats["q99"]), np.array(action_norm_stats["q01"])
        else:
            raise ValueError("Unsupported action/proprio normalization type detected!")

        actions = np.where(
            mask,
            0.5 * (normalized_actions + 1) * (action_high - action_low + 1e-8) + action_low,
            normalized_actions,
        )

        return actions

    def _run_diffusion_prediction(
        self,
        input_embeddings,
        all_actions_mask,
        noise,
        action_head,
        projected_patch_embeddings,
        labels,
        attention_mask,
        NUM_PROMPT_TOKENS,
        noisy_action_projector,
    ):
        """Run diffusion-based action prediction"""
        # Clone embedding for reuse in each timestep
        orig_projected_patch_embeddings = projected_patch_embeddings.clone()
        curr_noisy_actions = noise

        # Reverse diffusion: Iteratively denoise to generate action prediction
        for t in action_head.noise_scheduler.timesteps:
            # Get diffusion model's noise prediction (conditioned on VLA latent embedding, current noisy action
            # embedding, and diffusion timestep embedding)
            timesteps = torch.Tensor([t]).to(labels.device)
            diffusion_timestep_embeddings = (
                action_head.time_encoder(timesteps).to(curr_noisy_actions.dtype).to(curr_noisy_actions.device)
            )  # (B, llm_dim)
            diffusion_timestep_embeddings = diffusion_timestep_embeddings.unsqueeze(1)  # (B, 1, llm_dim)

            # [Diffusion] Replace the embeddings of the action tokens with noisy actions
            # (Later on, the positional embeddings will be added to them)

            # For simplicity, append diffusion timestep embedding to the end of projected vision tokens
            projected_patch_embeddings = torch.cat(
                (orig_projected_patch_embeddings, diffusion_timestep_embeddings), dim=1
            )

            # Reshape and project noisy actions into language embedding space
            B = curr_noisy_actions.shape[0]
            orig_curr_noisy_actions_shape = curr_noisy_actions.shape
            curr_noisy_actions = curr_noisy_actions.reshape(B, -1).unsqueeze(-1)
            noisy_action_features = noisy_action_projector(curr_noisy_actions)
            curr_noisy_actions = curr_noisy_actions.reshape(orig_curr_noisy_actions_shape)

            # Replace action token embeddings with noisy action embeddings
            input_embeddings = self._replace_input_embeddings(
                input_embeddings.clone(), all_actions_mask, noisy_action_features
            )

            # Build multimodal embeddings and attention mask
            multimodal_embeddings, multimodal_attention_mask = self._build_multimodal_attention(
                input_embeddings, projected_patch_embeddings, attention_mask
            )

            # Forward pass through language model
            language_model_output = self.language_model(
                input_ids=None,
                attention_mask=multimodal_attention_mask,
                position_ids=None,
                past_key_values=None,
                inputs_embeds=multimodal_embeddings,
                labels=None,
                use_cache=None,
                output_attentions=False,
                output_hidden_states=True,
                return_dict=True,
            )

            # Extract hidden states for action portion of response
            last_hidden_states = language_model_output.hidden_states[-1]  # (B, seq_len, D)
            actions_hidden_states = last_hidden_states[
                :,
                -ACTION_DIM * NUM_ACTIONS_CHUNK:,
                :,
            ]  # (B, act_chunk_len, D)

            # Predict noise and update noisy actions: x_t -> x_{t-1}
            noise_pred = action_head.predict_noise(actions_hidden_states)
            curr_noisy_actions = action_head.noise_scheduler.step(noise_pred, t, curr_noisy_actions).prev_sample

        curr_noisy_actions = curr_noisy_actions.reshape(NUM_ACTIONS_CHUNK, ACTION_DIM)

        # Return final actions
        return curr_noisy_actions.float().cpu().detach().numpy(), actions_hidden_states

    def _regression_or_discrete_prediction(
        self,
        input_embeddings,
        all_actions_mask,
        projected_patch_embeddings,
        attention_mask,
        labels,
        NUM_PROMPT_TOKENS,
        action_head=None,
    ):
        """Run L1 regression-based continuous action prediction or discrete action tokens prediction."""
        # Zero out action token embeddings
        all_actions_mask = all_actions_mask.unsqueeze(-1)  # (B, seq_len, 1)
        input_embeddings = input_embeddings * ~all_actions_mask

        # Build multimodal embeddings and attention mask
        multimodal_embeddings, multimodal_attention_mask = self._build_multimodal_attention(
            input_embeddings, projected_patch_embeddings, attention_mask
        )

        # Forward pass through language model
        language_model_output = self.language_model(
            input_ids=None,
            attention_mask=multimodal_attention_mask,
            position_ids=None,
            past_key_values=None,
            inputs_embeds=multimodal_embeddings,
            labels=None,
            use_cache=None,
            output_attentions=False,
            output_hidden_states=True,
            return_dict=True,
        )

        # Extract hidden states for action tokens
        last_hidden_states = language_model_output.hidden_states[-1]  # (B, seq_len, D)
        actions_hidden_states = last_hidden_states[
            :,
            -ACTION_DIM * NUM_ACTIONS_CHUNK:,
            :,
        ]  # (B, act_chunk_len, D)

        # Handle different prediction methods
        if action_head is not None:
            # L1 regression prediction
            normalized_actions = action_head.predict_action(actions_hidden_states)
            normalized_actions = normalized_actions.reshape(NUM_ACTIONS_CHUNK, ACTION_DIM)
            normalized_actions = normalized_actions.float().cpu().detach().numpy()
        else:
            # Discrete token-based prediction
            predicted_action_token_ids = (
                language_model_output.logits[
                    :,
                    -ACTION_DIM * NUM_ACTIONS_CHUNK:,
                ]
                .argmax(dim=2)
                .cpu()
                .numpy()
            )
            discretized_actions = self.vocab_size - predicted_action_token_ids
            discretized_actions = np.clip(discretized_actions - 1, a_min=0, a_max=self.bin_centers.shape[0] - 1)
            normalized_actions = self.bin_centers[discretized_actions]
            normalized_actions = normalized_actions.reshape(NUM_ACTIONS_CHUNK, ACTION_DIM)

        return normalized_actions, actions_hidden_states

    def predict_action(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        unnorm_key: Optional[str] = None,
        proprio=None,
        proprio_projector=None,
        action_head=None,
        noisy_action_projector=None,
        use_film: bool = False,
        **kwargs: str,
    ) -> np.ndarray:
        """Predict actions from input sequence, with options for different prediction methods.

        Args:
            input_ids: Input token ids
            unnorm_key: Key for unnormalization statistics
            proprio: Proprioceptive features
            proprio_projector: Projector for proprioceptive features
            action_head: Optional head for L1 regression or diffusion-based prediction
            noisy_action_projector: Projector for noisy actions in diffusion-based prediction
            use_film: Whether to use FiLM conditioning
            **kwargs: Additional arguments including pixel_values and attention_mask

        Returns:
            Tuple of (unnormalized_actions, action_hidden_states)
        """
        # If the special empty token ('') does not already appear after the colon (':') token in the prompt
        # (after "OUT:" or "ASSISTANT:"), insert it to match the inputs seen at training time
        if not torch.all(input_ids[:, -1] == 29871):
            input_ids = torch.cat(
                (input_ids, torch.unsqueeze(torch.Tensor([29871]).long(), dim=0).to(input_ids.device)), dim=1
            )

        pixel_values = kwargs["pixel_values"]
        attention_mask = kwargs["attention_mask"]

        # Create fake labels tensor (needed for action mask)
        labels = input_ids.clone()
        labels[:] = IGNORE_INDEX

        # Get number of tokens in prompt (excluding the start token)
        NUM_PROMPT_TOKENS = input_ids.shape[-1] - 1  # Subtract action tokens and stop token

        # Prepare inputs by adding necessary tokens
        input_ids, attention_mask = self._prepare_input_for_action_prediction(input_ids, attention_mask)

        # Update labels tensor for action mask computation later
        labels = self._prepare_labels_for_action_prediction(labels, input_ids)

        # Get input embeddings and action masks
        input_embeddings = self.get_input_embeddings()(input_ids)
        all_actions_mask = self._process_action_masks(labels)

        # Extract language embeddings
        language_embeddings = input_embeddings[~all_actions_mask].reshape(
            input_embeddings.shape[0], -1, input_embeddings.shape[2]
        )

        # Process vision features
        projected_patch_embeddings = self._process_vision_features(pixel_values, language_embeddings, use_film)

        # Add proprioceptive features if provided
        use_proprio = proprio_projector is not None and proprio is not None
        if use_proprio:
            proprio = torch.Tensor(proprio).to(projected_patch_embeddings.device, dtype=projected_patch_embeddings.dtype)
            projected_patch_embeddings = self._process_proprio_features(
                projected_patch_embeddings, proprio, proprio_projector
            )

        # Use diffusion if provided, otherwise use regression or discrete prediction
        use_diffusion = noisy_action_projector is not None and hasattr(action_head, "noise_scheduler")

        if use_diffusion:
            # Sample random noise with shape equal to output action, used as the starting state for reverse diffusion
            noise = torch.randn(
                size=(1, NUM_ACTIONS_CHUNK, ACTION_DIM), device=input_embeddings.device, dtype=input_embeddings.dtype
            )

            # Run diffusion-based prediction
            normalized_actions, actions_hidden_states = self._run_diffusion_prediction(
                input_embeddings,
                all_actions_mask,
                noise,
                action_head,
                projected_patch_embeddings,
                labels,
                attention_mask,
                NUM_PROMPT_TOKENS,
                noisy_action_projector,
            )
        else:
            # Run regression or discrete token-based prediction
            normalized_actions, actions_hidden_states = self._regression_or_discrete_prediction(
                input_embeddings,
                all_actions_mask,
                projected_patch_embeddings,
                attention_mask,
                labels,
                NUM_PROMPT_TOKENS,
                action_head,
            )

        # Unnormalize predicted actions
        actions = self._unnormalize_actions(normalized_actions, unnorm_key)

        return actions, actions_hidden_states

    @staticmethod
    def _check_unnorm_key(norm_stats: Dict[str, Dict[str, Any]], unnorm_key: Optional[str]) -> str:
        """Validate and resolve the unnormalization key for action statistics"""
        if unnorm_key is None:
            assert len(norm_stats) == 1, (
                f"Your model was trained on more than one dataset, "
                f"please pass a `unnorm_key` from the following options to choose the statistics "
                f"used for un-normalizing actions: {norm_stats.keys()}"
            )
            unnorm_key = next(iter(norm_stats.keys()))

        assert unnorm_key in norm_stats, (
            f"The `unnorm_key` you chose is not in the set of available dataset statistics, "
            f"please choose from: {norm_stats.keys()}"
        )
        return unnorm_key

    def get_action_dim(self, unnorm_key: Optional[str] = None) -> int:
        """Get the dimensionality of the policy's action space."""
        unnorm_key = self._check_unnorm_key(self.norm_stats, unnorm_key)
        return len(self.norm_stats[unnorm_key]["action"]["min"])

    def get_action_stats(self, unnorm_key: Optional[str] = None) -> Dict[str, Any]:
        """Get all the logged statistics for the given dataset."""
        unnorm_key = self._check_unnorm_key(self.norm_stats, unnorm_key)
        return self.norm_stats[unnorm_key]["action"]