From 8123210d7ebb35ca5766eb5873ae0b78000b694c Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Sun, 2 Feb 2025 03:56:25 -0800 Subject: [PATCH] Update llama.py --- unsloth/models/llama.py | 1 + 1 file changed, 1 insertion(+) diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index 475b234e17..401b8986a8 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -936,6 +936,7 @@ def LlamaModel_fast_forward_inference( next_decoder_cache = [] residual = torch.empty_like(X) + print(bsz, q_len, hd) _XX = torch.empty((2, bsz, q_len, hd), dtype = torch.float32) XX, XX2 = _XX[0], _XX[1] variance = torch.empty((bsz, q_len, 1), dtype = torch.float32, device = "cuda:0")