refactor : align config field names with Hugging Face
- dim -> hidden_size, n_layers -> num_hidden_layers - dim_ffn -> intermediate_size, n_heads -> num_attention_heads - n_kv_heads -> num_key_value_heads, max_len -> max_position_embeddings - norm_eps -> rms_norm_eps, tie_weight -> tie_word_embeddings - update model, inference, training, scripts, tests, docs
This commit is contained in:
@@ -480,7 +480,7 @@ def train(
|
||||
config.neftune_alpha = neftune_alpha
|
||||
|
||||
if window_size is None:
|
||||
window_size = config.max_len
|
||||
window_size = config.max_position_embeddings
|
||||
|
||||
strategy_kwargs = {
|
||||
"beta": kwargs.pop("dpo_beta"),
|
||||
|
||||
Reference in New Issue
Block a user