From 30cc4ffd67abd9ea227a982d1eeb87fabfb2b79f Mon Sep 17 00:00:00 2001 From: Daniel Han-Chen Date: Sat, 24 Feb 2024 03:27:21 +1100 Subject: [PATCH] Update llama.py --- unsloth/models/llama.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index 623855d290..9fef920a9a 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -203,7 +203,9 @@ def LlamaAttention_fast_forward_inference( A = torch.matmul(A, Vnn, out = Qn) A = A.transpose(1, 2) A = A.reshape(bsz, 1, attention_size) + print(self.temp_QA[1].shape) A = fast_linear_forward(self.o_proj, A)#, out = self.temp_QA[1][:,:,self.hidden_size]) + print(A.shape) return A, (Kn, Vn) pass