Update gemma.py

unsloth/models/gemma.py

@@ -20,6 +20,7 @@ from transformers.models.gemma.modeling_gemma import (
     GemmaDecoderLayer,
     GemmaModel,
     GemmaForCausalLM,
+    GemmaRotaryEmbedding,
     apply_rotary_pos_emb,
     repeat_kv,
 )
@@ -56,6 +57,31 @@ def fast_geglu_inference(self, X):
 pass
 
 
+class FastGemmaRotaryEmbedding(nn.Module):
+    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
+        super().__init__()
+
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        self.register_buffer("inv_freq", None, persistent=False)
+
+    def forward(self, x, position_ids, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        if self.inv_freq is None:
+            self.inv_freq = 1.0 / (
+                self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64, device=x.device).float() / self.dim)
+            )
+        print(position_ids)
+
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+        freqs = (inv_freq_expanded @ position_ids_expanded).transpose(1, 2)
+        emb = torch.cat((freqs, freqs), dim=-1)
+        return emb.cos().to(dtype=x.dtype), emb.sin().to(dtype=x.dtype)
+pass
+
+
 # https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/modeling_llama.py#L320
 def GemmaAttention_fast_forward(
     self,
@@ -513,7 +539,7 @@ class FastGemmaModel(FastLlamaModel):
         GemmaModel          .forward = GemmaModel_fast_forward
         GemmaForCausalLM    .forward = GemmaForCausalLM_fast_forward
         PeftModelForCausalLM.forward = PeftModelForCausalLM_fast_forward
-
+        GemmaRotaryEmbedding = FastGemmaRotaryEmbedding
         # Solves https://github.com/unslothai/unsloth/issues/168
         # Static KV Cache was introduced in 4.38.0, causing training to be much slower.
         # Inferene can now be CUDAGraphed, but we shall retain the old rotary embeddings.