vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix n_ctx issue for Baichuan & Baichuan2 13B model

Browse source 
Signed-off-by: caiyesd <caiyesd@gmail.com>
This commit is contained in:
caiyesd 2023-11-23 10:46:58 +08:00
parent 8e672efe63
commit 7209a6ae0b
1 changed files with 2 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

2
convert.py
View file

@ -229,6 +229,8 @@ class Params:
n_ctx = config["max_sequence_length"] n_ctx = config["max_sequence_length"]
elif "max_position_embeddings" in config: elif "max_position_embeddings" in config:
n_ctx = config["max_position_embeddings"] n_ctx = config["max_position_embeddings"]
elif "model_max_length" in config:
n_ctx = config["model_max_length"]
else: else:
raise Exception("failed to guess 'n_ctx'. This model is unknown or unsupported.\n" raise Exception("failed to guess 'n_ctx'. This model is unknown or unsupported.\n"
"Suggestion: provide 'config.json' of the model in the same directory containing model files.") "Suggestion: provide 'config.json' of the model in the same directory containing model files.")