From 2e4bd2a665713e3bb7d6bdee86b0fed920c6b104 Mon Sep 17 00:00:00 2001
From: Matthias Reso <13337103+mreso@users.noreply.github.com>
Date: Sat, 16 Sep 2023 19:13:55 +0000
Subject: [PATCH] Resize vocab size to fix idx error
---
src/llama_recipes/finetuning.py | 19 +++++++++++--------
1 file changed, 11 insertions(+), 8 deletions(-)
diff --git a/src/llama_recipes/finetuning.py b/src/llama_recipes/finetuning.py
index 51647d89..44c74c78 100644
--- a/src/llama_recipes/finetuning.py
+++ b/src/llama_recipes/finetuning.py
@@ -108,6 +108,17 @@ def main(**kwargs):
model = BetterTransformer.transform(model)
except ImportError:
print("Module 'optimum' not found. Please install 'optimum' it before proceeding.")
+
+ # Load the tokenizer and add special tokens
+ tokenizer = LlamaTokenizer.from_pretrained(train_config.model_name)
+ tokenizer.add_special_tokens(
+ {
+
+ "pad_token": "<PAD>",
+ }
+ )
+ model.resize_token_embeddings(model.config.vocab_size + 1)
+
print_model_size(model, train_config, rank if train_config.enable_fsdp else 0)
# Prepare the model for int8 training if quantization is enabled
@@ -118,14 +129,6 @@ def main(**kwargs):
if train_config.enable_fsdp and fsdp_config.pure_bf16:
model.to(torch.bfloat16)
- # Load the tokenizer and add special tokens
- tokenizer = LlamaTokenizer.from_pretrained(train_config.model_name)
- tokenizer.add_special_tokens(
- {
-
- "pad_token": "<PAD>",
- }
- )
if train_config.use_peft:
peft_config = generate_peft_config(train_config, kwargs)
model = get_peft_model(model, peft_config)
--
GitLab