From 0f98acfac6cc561dc57586bfff778405e42b576b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Steffen=20R=C3=B6cker?= Date: Sat, 18 May 2024 10:04:55 +0200 Subject: [PATCH] llama : add support for larger Granite Code Models (20B, 34B) (#7324) Tie the weights for ARCH_STARCODER to support the larger Granite code models. Partially addresses ggerganov/issues/7116 There still remains to be a few things to fix. Currently requires `--override-kv tokenizer.ggml.add_bos_token=bool:false` --- llama.cpp | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/llama.cpp b/llama.cpp index b752ddc6b401f..2b91eec86c64f 100644 --- a/llama.cpp +++ b/llama.cpp @@ -5188,7 +5188,14 @@ static bool llm_load_tensors( { model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}); - model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}); + model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false); + if (!model.output) { + // needs to be on GPU + model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); + ml.n_created--; // artificial tensor + ml.size_data += ggml_nbytes(model.output); + } + } for (int i = 0; i < n_layer; ++i) {