zhangyiss/mlx-examples
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx-examples.git synced 2025-08-30 02:53:41 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Don't dupe BOS

Browse Source 
Ensure completion batching doesn't allow BOS dupes for instruction models with chat models whose tokenizer configurations have ```add_bos_token = True``` (see: 1095)
This commit is contained in:
Chime Ogbuji 2024-11-10 10:08:44 -05:00 committed by Awni Hannun
parent 69282ab7fc
commit 3f08dfc762
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
llms/mlx_lm/tuner/trainer.py
View File

@ -166,7 +166,7 @@ def iterate_completion_batches(
for j in batch_idx[i]:
prompt, completion = dataset.get_prompt_and_completion(j)
prompt_lengths.append(input_length(prompt, completion, tokenizer))
full_sequence = tokenizer.encode(dataset[j])
full_sequence = tokenizer.encode(dataset[j], add_special_tokens=False)
if full_sequence[-1] != tokenizer.eos_token_id:
full_sequence.append(tokenizer.eos_token_id)
batch.append(full_sequence)