add LlamaRotary because generation is not good otherwise

src/nanotron/models/llama.py

@@ -117,6 +117,73 @@ def forward(
         return x_out.type(dtype)
 
 
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+### llama
+class LlamaRotaryEmbedding(nn.Module):
+    def __init__(self, dim: int, end: int, theta: float = 500000.0):
+        super().__init__()
+        self.dim = dim
+        self.end = end
+        self.theta = theta
+        self.init_rotary_embeddings()
+
+    def init_rotary_embeddings(self):
+        inv_freq = 1.0 / (self.theta ** (torch.arange(0, self.dim, 2, dtype=torch.float, device="cuda") / self.dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+
+    @torch.no_grad()
+    def forward(
+        self,
+        x: torch.Tensor,  # [batch_size, seq_length, num_heads, d_qk]
+        position_ids: Optional[torch.LongTensor],  # [batch_size, seq_length]
+    ):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        # print("rotary")
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+        # Force float32 since bfloat16 loses precision on long contexts
+        # See https://github.com/huggingface/transformers/pull/29285
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, unsqueeze_dim=2):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
 class GLUActivation(nn.Module):
     def __init__(self, act_fn_name: str):
         super().__init__()
@@ -303,10 +370,17 @@ def __init__(
             contiguous_chunks=qkv_contiguous_chunks,
         )
         # TODO(kunhao): We want to have only one version per device and not one version per layer.
-        self.rotary_embedding = RotaryEmbedding(
-            dim=self.d_qk,
-            end=config.max_position_embeddings,
-        )
+        if config.rope_interleaved:
+            self.rotary_embedding = RotaryEmbedding(
+                dim=self.d_qk,
+                end=config.max_position_embeddings,
+            )
+        else:
+            self.rotary_embedding = LlamaRotaryEmbedding(
+                dim=self.d_qk,
+                end=config.max_position_embeddings,
+            )
+        self.rope_interleaved = config.rope_interleaved
 
         # NOTE: Only supported for training (TODO(fmom): position_ids not supported yet)
         self.flash_rotary_embedding = FlashRotaryEmbedding(
@@ -336,6 +410,7 @@ def forward(
         self,
         hidden_states,  # [seq_length, batch_size, hidden_size]
         sequence_mask,  # [batch_size, seq_length]
+        position_ids: Optional[torch.LongTensor] = None,
     ):
         from flash_attn import bert_padding
         from flash_attn.flash_attn_interface import (
@@ -390,8 +465,19 @@ def forward(
             # Compute rotary embeddings
             # Note: keep track of old rotary embedding end to check if we need to enlarge k_cache and v_cache
             old_rotary_embed_end = self.rotary_embedding.end
-            query_states = self.rotary_embedding(query_states, position_ids=position_ids)
-            key_states = self.rotary_embedding(key_states, position_ids=position_ids)
+
+            # Rotate half rotary_embedding
+            # cos, sin = self.rotary_embedding(value_states, position_ids)
+            # query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+            # interleaved
+            if self.rope_interleaved:
+                query_states = self.rotary_embedding(query_states, position_ids=position_ids)
+                key_states = self.rotary_embedding(key_states, position_ids=position_ids)
+            # llama rotary position embedding
+            else:
+                cos, sin = self.rotary_embedding(value_states, position_ids)
+                query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
             if "key" not in store:
                 # First inference iteration (Prefill)