More cleanup

transformer_lens/HookedTransformer.py

@@ -114,7 +114,6 @@ def __init__(
         tokenizer: Optional[PreTrainedTokenizerBase] = None,
         move_to_device: bool = True,
         default_padding_side: Literal["left", "right"] = "right",
-        zero_pos_embed: bool = False,
     ):
         """Model initialization.
 
@@ -139,7 +138,6 @@ def __init__(
             )
 
         self.cfg = HookedTransformerConfig.unwrap(cfg)
-        self.zero_pos_embed = zero_pos_embed
 
         if tokenizer is not None:
             self.set_tokenizer(tokenizer, default_padding_side=default_padding_side)
@@ -189,7 +187,7 @@ def __init__(
             self.hook_tokens = HookPoint()  # [batch, pos]
 
         self.blocks = nn.ModuleList(
-            [TransformerBlock(self.cfg, block_index, self.zero_pos_embed) for block_index in range(self.cfg.n_layers)]
+            [TransformerBlock(self.cfg, block_index) for block_index in range(self.cfg.n_layers)]
         )
 
         if self.cfg.normalization_type == "RMS":
@@ -358,7 +356,7 @@ def input_to_embed(
             pos_embed = self.hook_pos_embed(
                 self.pos_embed(tokens, pos_offset, attention_mask)
             )  # [batch, pos, d_model]
-            residual = embed + pos_embed * (1. if not self.zero_pos_embed else 0.)  # [batch, pos, d_model]
+            residual = embed + pos_embed  # [batch, pos, d_model]
             shortformer_pos_embed = None
         elif self.cfg.positional_embedding_type == "shortformer":
             # If we're using shortformer style attention, we don't add the positional embedding to
@@ -367,7 +365,7 @@ def input_to_embed(
                 self.pos_embed(tokens, pos_offset, attention_mask)
             )  # [batch, pos, d_model]
             residual = embed
-            shortformer_pos_embed = pos_embed * (1. if not self.zero_pos_embed else 0.)
+            shortformer_pos_embed = pos_embed
         elif self.cfg.positional_embedding_type == "rotary":
             # Rotary doesn't use positional embeddings, instead they're applied when dot producting
             # keys and queries. See HookedTransformerConfig for details
@@ -573,7 +571,7 @@ def forward(
                 if shortformer_pos_embed is not None:
                     shortformer_pos_embed = shortformer_pos_embed.to(
                         devices.get_device_for_block_index(i, self.cfg)
-                    ) * (1. if not self.zero_pos_embed else 0.)
+                    )
 
                 residual = block(
                     residual,
@@ -1088,7 +1086,6 @@ def from_pretrained(
         default_prepend_bos: Optional[bool] = None,
         default_padding_side: Literal["left", "right"] = "right",
         dtype="float32",
-        zero_pos_embed: bool = False,
         first_n_layers: Optional[int] = None,
         **from_pretrained_kwargs,
     ) -> T:
@@ -1332,7 +1329,6 @@ def from_pretrained(
             tokenizer,
             move_to_device=False,
             default_padding_side=default_padding_side,
-            zero_pos_embed=zero_pos_embed,
         )
 
         model.load_and_process_state_dict(
@@ -2238,7 +2234,7 @@ def forward_(*model_args, **model_kwargs):
                     )
                 final_logits = logits[:, -1, :]
 
-                # SAMPLING
+                # sampling
                 if do_sample:
                     sampled_tokens = utils.sample_logits(
                         final_logits,
@@ -2265,10 +2261,11 @@ def forward_(*model_args, **model_kwargs):
                         )
                     )
 
-                # update the tokens!
+                # concatenate the new tokens
                 tokens = torch.cat([tokens, sampled_tokens.unsqueeze(-1)], dim=-1)
 
-                # APPEND – we need to clone on the first pass to prevent overwrite
+                # concatenate the cache
+                # we need to clone on the first pass to prevent overwrite
                 token_tape = torch.cat([token_tape, sampled_tokens.unsqueeze(-1)], dim=-1) if token_tape is not None else torch.clone(tokens[:, -ctx_length:])  # appends to the prompt tokens
                 if return_cache:
                     def cat_cache_var(key, var_tape, var):