dreamer4 0.0.101__tar.gz → 0.1.0__tar.gz

{dreamer4-0.0.101 → dreamer4-0.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dreamer4
-Version: 0.0.101
+Version: 0.1.0
 Summary: Dreamer 4
 Project-URL: Homepage, https://pypi.org/project/dreamer4/
 Project-URL: Repository, https://github.com/lucidrains/dreamer4
@@ -53,11 +53,75 @@ Description-Content-Type: text/markdown
 
 <img src="./dreamer4-fig2.png" width="400px"></img>
 
-## Dreamer 4 (wip)
+## Dreamer 4
 
 Implementation of Danijar's [latest iteration](https://arxiv.org/abs/2509.24527v1) for his [Dreamer](https://danijar.com/project/dreamer4/) line of work
 
-[Temporary Discord](https://discord.gg/MkACrrkrYR)
+## Install
+
+```bash
+$ pip install dreamer4-pytorch
+```
+
+## Usage
+
+```python
+import torch
+from dreamer4 import VideoTokenizer, DynamicsWorldModel
+
+# video tokenizer, learned through MAE + lpips
+
+tokenizer = VideoTokenizer(
+    dim = 512,
+    dim_latent = 32,
+    patch_size = 32,
+    image_height = 256,
+    image_width = 256
+)
+
+# dynamics world model
+
+dynamics = DynamicsWorldModel(
+    dim = 512,
+    dim_latent = 32,
+    video_tokenizer = tokenizer,
+    num_discrete_actions = 4,
+    num_residual_streams = 1
+)
+
+# state, action, rewards
+
+video = torch.randn(2, 3, 10, 256, 256)
+discrete_actions = torch.randint(0, 4, (2, 10, 1))
+rewards = torch.randn(2, 10)
+
+# learn dynamics / behavior cloned model
+
+loss = dynamics(
+    video = video,
+    rewards = rewards,
+    discrete_actions = discrete_actions
+)
+
+loss.backward()
+
+# do the above with much data
+
+# then generate dreams
+
+dreams = dynamics.generate(
+    10,
+    batch_size = 2,
+    return_decoded_video = True,
+    return_for_policy_optimization = True
+)
+
+# learn from the dreams
+
+actor_loss, critic_loss = dynamics.learn_from_experience(dreams)
+
+(actor_loss + critic_loss).backward()
+```
 
 ## Citation
 
@@ -72,3 +136,5 @@ Implementation of Danijar's [latest iteration](https://arxiv.org/abs/2509.24527v
     url     = {https://arxiv.org/abs/2509.24527}, 
 }
 ```
+
+*the conquest of nature is to be achieved through number and measure* - angels to Descartes, in a dream, the story goes.

dreamer4-0.1.0/README.md

@@ -0,0 +1,87 @@
+<img src="./dreamer4-fig2.png" width="400px"></img>
+
+## Dreamer 4
+
+Implementation of Danijar's [latest iteration](https://arxiv.org/abs/2509.24527v1) for his [Dreamer](https://danijar.com/project/dreamer4/) line of work
+
+## Install
+
+```bash
+$ pip install dreamer4-pytorch
+```
+
+## Usage
+
+```python
+import torch
+from dreamer4 import VideoTokenizer, DynamicsWorldModel
+
+# video tokenizer, learned through MAE + lpips
+
+tokenizer = VideoTokenizer(
+    dim = 512,
+    dim_latent = 32,
+    patch_size = 32,
+    image_height = 256,
+    image_width = 256
+)
+
+# dynamics world model
+
+dynamics = DynamicsWorldModel(
+    dim = 512,
+    dim_latent = 32,
+    video_tokenizer = tokenizer,
+    num_discrete_actions = 4,
+    num_residual_streams = 1
+)
+
+# state, action, rewards
+
+video = torch.randn(2, 3, 10, 256, 256)
+discrete_actions = torch.randint(0, 4, (2, 10, 1))
+rewards = torch.randn(2, 10)
+
+# learn dynamics / behavior cloned model
+
+loss = dynamics(
+    video = video,
+    rewards = rewards,
+    discrete_actions = discrete_actions
+)
+
+loss.backward()
+
+# do the above with much data
+
+# then generate dreams
+
+dreams = dynamics.generate(
+    10,
+    batch_size = 2,
+    return_decoded_video = True,
+    return_for_policy_optimization = True
+)
+
+# learn from the dreams
+
+actor_loss, critic_loss = dynamics.learn_from_experience(dreams)
+
+(actor_loss + critic_loss).backward()
+```
+
+## Citation
+
+```bibtex
+@misc{hafner2025trainingagentsinsidescalable,
+    title   = {Training Agents Inside of Scalable World Models}, 
+    author  = {Danijar Hafner and Wilson Yan and Timothy Lillicrap},
+    year    = {2025},
+    eprint  = {2509.24527},
+    archivePrefix = {arXiv},
+    primaryClass = {cs.AI},
+    url     = {https://arxiv.org/abs/2509.24527}, 
+}
+```
+
+*the conquest of nature is to be achieved through number and measure* - angels to Descartes, in a dream, the story goes.

{dreamer4-0.0.101 → dreamer4-0.1.0}/dreamer4/dreamer4.py

@@ -1902,6 +1902,7 @@ class DynamicsWorldModel(Module):
         pmpo_pos_to_neg_weight = 0.5, # pos and neg equal weight
         pmpo_reverse_kl = True,
         pmpo_kl_div_loss_weight = .3,
+        normalize_advantages = None,
         value_clip = 0.4,
         policy_entropy_weight = .01,
         gae_use_accelerated = False
@@ -2425,8 +2426,10 @@ class DynamicsWorldModel(Module):
         value_optim: Optimizer | None = None,
         only_learn_policy_value_heads = True, # in the paper, they do not finetune the entire dynamics model, they just learn the heads
         use_pmpo = True,
+        normalize_advantages = None,
         eps = 1e-6
     ):
+        assert isinstance(experience, Experience)
 
         latents = experience.latents
         actions = experience.actions
@@ -2439,7 +2442,7 @@ class DynamicsWorldModel(Module):
         step_size = experience.step_size
         agent_index = experience.agent_index
 
-        assert all([*map(exists, (old_log_probs, actions, old_values, rewards, step_size))]), 'the generations need to contain the log probs, values, and rewards for policy optimization'
+        assert all([*map(exists, (old_log_probs, actions, old_values, rewards, step_size))]), 'the generations need to contain the log probs, values, and rewards for policy optimization - world_model.generate(..., return_log_probs_and_values = True)'
 
         batch, time = latents.shape[0], latents.shape[1]
 
@@ -2507,16 +2510,19 @@ class DynamicsWorldModel(Module):
         else:
             advantage = returns - old_values
 
-        # apparently they just use the sign of the advantage
+        # if using pmpo, do not normalize advantages, but can be overridden
+
+        normalize_advantages = default(normalize_advantages, not use_pmpo)
+
+        if normalize_advantages:
+            advantage = F.layer_norm(advantage, advantage.shape, eps = eps)
+
         # https://arxiv.org/abs/2410.04166v1
 
         if use_pmpo:
             pos_advantage_mask = advantage >= 0.
             neg_advantage_mask = ~pos_advantage_mask
 
-        else:
-            advantage = F.layer_norm(advantage, advantage.shape, eps = eps)
-
         # replay for the action logits and values
         # but only do so if fine tuning the entire world model for RL
 
@@ -2689,12 +2695,22 @@ class DynamicsWorldModel(Module):
         return_rewards_per_frame = False,
         return_agent_actions = False,
         return_log_probs_and_values = False,
+        return_for_policy_optimization = False,
         return_time_kv_cache = False,
         store_agent_embed = True,
         store_old_action_unembeds = True
 
     ): # (b t n d) | (b c t h w)
 
+        # handy flag for returning generations for rl
+
+        if return_for_policy_optimization:
+            return_agent_actions |= True
+            return_log_probs_and_values |= True
+            return_rewards_per_frame |= True
+
+        # more variables
+
         has_proprio = self.has_proprio
         was_training = self.training
         self.eval()
@@ -2764,6 +2780,19 @@ class DynamicsWorldModel(Module):
 
             curr_time_steps = latents.shape[1]
 
+            # determine whether to take an extra step if
+            # (1) using time kv cache
+            # (2) decoding anything off agent embedding (rewards, actions, etc)
+
+            take_extra_step = (
+                use_time_kv_cache or
+                return_rewards_per_frame or
+                store_agent_embed or
+                return_agent_actions
+            )
+
+            # prepare noised latent / proprio inputs
+
             noised_latent = randn((batch_size, 1, self.num_video_views, *latent_shape), device = self.device)
 
             noised_proprio = None
@@ -2771,7 +2800,10 @@ class DynamicsWorldModel(Module):
             if has_proprio:
                 noised_proprio = randn((batch_size, 1, self.dim_proprio), device = self.device)
 
-            for step in range(num_steps):
+            # denoising steps
+
+            for step in range(num_steps + int(take_extra_step)):
+
                 is_last_step = (step + 1) == num_steps
 
                 signal_levels = full((batch_size, 1), step * step_size, dtype = torch.long, device = self.device)
@@ -2814,6 +2846,11 @@ class DynamicsWorldModel(Module):
                 if use_time_kv_cache and is_last_step:
                     time_kv_cache = next_time_kv_cache
 
+                # early break if taking an extra step for agent embedding off cleaned latents for decoding
+
+                if take_extra_step and is_last_step:
+                    break
+
                 # maybe proprio
 
                 if has_proprio:
@@ -3016,7 +3053,7 @@ class DynamicsWorldModel(Module):
         latent_is_noised = False,
         return_all_losses = False,
         return_intermediates = False,
-        add_autoregressive_action_loss = False,
+        add_autoregressive_action_loss = True,
         update_loss_ema = None,
         latent_has_view_dim = False
     ):

{dreamer4-0.0.101 → dreamer4-0.1.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "dreamer4"
-version = "0.0.101"
+version = "0.1.0"
 description = "Dreamer 4"
 authors = [
     { name = "Phil Wang", email = "lucidrains@gmail.com" }

dreamer4-0.0.101/README.md

@@ -1,21 +0,0 @@
-<img src="./dreamer4-fig2.png" width="400px"></img>
-
-## Dreamer 4 (wip)
-
-Implementation of Danijar's [latest iteration](https://arxiv.org/abs/2509.24527v1) for his [Dreamer](https://danijar.com/project/dreamer4/) line of work
-
-[Temporary Discord](https://discord.gg/MkACrrkrYR)
-
-## Citation
-
-```bibtex
-@misc{hafner2025trainingagentsinsidescalable,
-    title   = {Training Agents Inside of Scalable World Models}, 
-    author  = {Danijar Hafner and Wilson Yan and Timothy Lillicrap},
-    year    = {2025},
-    eprint  = {2509.24527},
-    archivePrefix = {arXiv},
-    primaryClass = {cs.AI},
-    url     = {https://arxiv.org/abs/2509.24527}, 
-}
-```