From 9d7bb8e91705d3b3f63df2a5480a265dcd62aa14 Mon Sep 17 00:00:00 2001 From: younesbelkada Date: Tue, 22 Nov 2022 11:02:56 +0000 Subject: [PATCH] revert `keys_to_ignore` for M2M100 --- src/transformers/models/m2m_100/modeling_m2m_100.py | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) diff --git a/src/transformers/models/m2m_100/modeling_m2m_100.py b/src/transformers/models/m2m_100/modeling_m2m_100.py index f7ef189a155d..60606f9b9ce6 100755 --- a/src/transformers/models/m2m_100/modeling_m2m_100.py +++ b/src/transformers/models/m2m_100/modeling_m2m_100.py @@ -1128,7 +1128,14 @@ def custom_forward(*inputs): M2M_100_START_DOCSTRING, ) class M2M100Model(M2M100PreTrainedModel): - _keys_to_ignore_on_load_missing = ["encoder.embed_tokens.weight", "decoder.embed_tokens.weight"] + _keys_to_ignore_on_load_missing = [ + "encoder.embed_tokens.weight", + "decoder.embed_tokens.weight", + "encoder.embed_positions.weights", + "encoder.embed_positions.bias", + "decoder.embed_positions.weights", + "decoder.embed_positions.bias", + ] def __init__(self, config: M2M100Config): super().__init__(config) @@ -1248,6 +1255,10 @@ class M2M100ForConditionalGeneration(M2M100PreTrainedModel): r"lm_head.weight", r"encoder.embed_tokens.weight", r"decoder.embed_tokens.weight", + r"encoder.embed_positions.weights", + r"encoder.embed_positions.bias", + r"decoder.embed_positions.weights", + r"decoder.embed_positions.bias", ] def __init__(self, config: M2M100Config):