| { |
| "metadata": { |
| "total_size": 18483411968.0 |
| }, |
| "weight_map": { |
| "/decoder_layers/gemma2_decoder_layer/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_1/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_2/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_3/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_4/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_5/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_6/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_7/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_8/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_9/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_10/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_11/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/mlp/gate/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/mlp/up/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/post_attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/post_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_12/pre_feedforward_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/attention/key/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/attention/output_proj/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/attention/query/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/attention/value/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/attention_norm/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/mlp/down/vars": "model_00000.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_13/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_14/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_15/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_16/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_17/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_18/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_19/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_20/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_21/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_22/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_23/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_24/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_25/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/attention/key/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/attention/output_proj/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/attention/query/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/attention/value/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/mlp/down/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/mlp/gate/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/mlp/up/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/post_attention_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/post_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_26/pre_feedforward_norm/vars": "model_00001.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_27/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_28/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_29/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_30/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_31/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_32/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_33/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_34/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_35/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_36/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_37/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_38/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/mlp/gate/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/mlp/up/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/post_attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/post_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_39/pre_feedforward_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/attention/key/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/attention/output_proj/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/attention/query/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/attention/value/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/attention_norm/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/mlp/down/vars": "model_00002.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/mlp/gate/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/mlp/up/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/post_attention_norm/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/post_feedforward_norm/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_40/pre_feedforward_norm/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/attention/key/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/attention/output_proj/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/attention/query/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/attention/value/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/attention_norm/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/mlp/down/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/mlp/gate/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/mlp/up/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/post_attention_norm/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/post_feedforward_norm/vars": "model_00003.weights.h5", |
| "/decoder_layers/gemma2_decoder_layer_41/pre_feedforward_norm/vars": "model_00003.weights.h5", |
| "/final_norm/vars": "model_00003.weights.h5", |
| "/layers/embedding/vars": "model_00003.weights.h5" |
| } |
| } |
