Hugging Face's logo

AsadIsmail
/
whisper-medium-ternary

Automatic Speech Recognition
Transformers
ternary-quant
quantization
ternary
audio
speech-to-text
whisper
Model card Files
xet
 Community
whisper-medium-ternary / metadata.json
AsadIsmail's picture
AsadIsmail
Add files using upload-large-folder tool
5abf93b verified
raw
history blame contribute delete
180 kB
{
 "model_name": "openai/whisper-medium",
 "model_config": {
 "transformers_version": "5.5.3",
 "architectures": [
 "WhisperForConditionalGeneration"
 ],
 "output_hidden_states": false,
 "return_dict": true,
 "dtype": "float32",
 "chunk_size_feed_forward": 0,
 "is_encoder_decoder": true,
 "id2label": {
 "0": "LABEL_0",
 "1": "LABEL_1"
 },
 "label2id": {
 "LABEL_0": 0,
 "LABEL_1": 1
 },
 "problem_type": null,
 "vocab_size": 51865,
 "num_mel_bins": 80,
 "encoder_layers": 24,
 "encoder_attention_heads": 16,
 "decoder_layers": 24,
 "decoder_attention_heads": 16,
 "decoder_ffn_dim": 4096,
 "encoder_ffn_dim": 4096,
 "encoder_layerdrop": 0.0,
 "decoder_layerdrop": 0.0,
 "decoder_start_token_id": 50258,
 "use_cache": true,
 "activation_function": "gelu",
 "d_model": 1024,
 "dropout": 0.0,
 "attention_dropout": 0.0,
 "activation_dropout": 0.0,
 "init_std": 0.02,
 "scale_embedding": false,
 "max_source_positions": 1500,
 "max_target_positions": 448,
 "pad_token_id": 50257,
 "bos_token_id": 50257,
 "eos_token_id": 50257,
 "suppress_tokens": [
 1,
 2,
 7,
 8,
 9,
 10,
 14,
 25,
 26,
 27,
 28,
 29,
 31,
 58,
 59,
 60,
 61,
 62,
 63,
 90,
 91,
 92,
 93,
 359,
 503,
 522,
 542,
 873,
 893,
 902,
 918,
 922,
 931,
 1350,
 1853,
 1982,
 2460,
 2627,
 3246,
 3253,
 3268,
 3536,
 3846,
 3961,
 4183,
 4667,
 6585,
 6647,
 7273,
 9061,
 9383,
 10428,
 10929,
 11938,
 12033,
 12331,
 12562,
 13793,
 14157,
 14635,
 15265,
 15618,
 16553,
 16604,
 18362,
 18956,
 20075,
 21675,
 22520,
 26130,
 26161,
 26435,
 28279,
 29464,
 31650,
 32302,
 32470,
 36865,
 42863,
 47425,
 49870,
 50254,
 50258,
 50358,
 50359,
 50360,
 50361,
 50362
 ],
 "begin_suppress_tokens": [
 220,
 50257
 ],
 "use_weighted_layer_sum": false,
 "classifier_proj_size": 256,
 "apply_spec_augment": false,
 "mask_time_prob": 0.05,
 "mask_time_length": 10,
 "mask_time_min_masks": 2,
 "mask_feature_prob": 0.0,
 "mask_feature_length": 10,
 "mask_feature_min_masks": 0,
 "median_filter_width": 7,
 "tie_word_embeddings": true,
 "_name_or_path": "openai/whisper-medium",
 "forced_decoder_ids": [
 [
 1,
 50259
 ],
 [
 2,
 50359
 ],
 [
 3,
 50363
 ]
 ],
 "model_type": "whisper",
 "output_attentions": false
 },
 "quant_config": {
 "components": [
 "decoder"
 ],
 "scheme": "tritplane3",
 "group_size": 32,
 "n_iter": 10,
 "salient_fraction": 0.0,
 "rescue_fraction": 0.0,
 "n_planes": 3,
 "allow_all_linear": false,
 "target_module_names": [
 "Wqkv",
 "att_proj",
 "attn.proj",
 "attn.qkv",
 "c_attn",
 "c_fc",
 "c_proj",
 "dense",
 "dense_4h_to_h",
 "dense_h_to_4h",
 "down_proj",
 "fc1",
 "fc2",
 "ff_proj",
 "gate_proj",
 "gate_up_proj",
 "k",
 "k_proj",
 "linear",
 "o",
 "o_proj",
 "out_proj",
 "per_layer_input_gate",
 "per_layer_projection",
 "proj",
 "q",
 "q_proj",
 "qkv",
 "qkv_proj",
 "query_key_value",
 "up_proj",
 "v",
 "v_proj",
 "w1",
 "w2",
 "w3",
 "wi",
 "wi_0",
 "wi_1",
 "wo"
 ],
 "max_length": 160,
 "calibration_batch_size": 2,
 "calibration_prompts": null,
 "vlm_use_demo_image": true
 },
 "plan": {},
 "layer_info": {
 "model.decoder.layers.0.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.0.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.1.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.2.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.3.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.4.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.5.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.6.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.7.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.8.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.9.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.10.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.11.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.12.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.13.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.14.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.15.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.16.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.17.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.18.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.19.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.20.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.21.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.22.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.self_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.self_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.self_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.self_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.encoder_attn.k_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.encoder_attn.v_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.encoder_attn.q_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.encoder_attn.out_proj": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 1048576,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 1179648,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.fc1": {
 "scheme": "tritplane_small_v1",
 "shape": [
 4096,
 1024
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 },
 "model.decoder.layers.23.fc2": {
 "scheme": "tritplane_small_v1",
 "shape": [
 1024,
 4096
 ],
 "dtype": "torch.float32",
 "num_elements": 4194304,
 "n_planes": 3,
 "group_sizes": [
 32,
 32,
 32
 ],
 "rescued_rows": 0,
 "stored_bytes": 4718592,
 "effective_bits": 9.0
 }
 },
 "stats": {
 "model.decoder.layers.0.self_attn.k_proj": {
 "mse": 8.22391393739963e-06,
 "rmse": 0.0028677367273513147,
 "relative_error": 0.11225825031610676,
 "max_error": 0.07249832153320312,
 "sparsity": 0.39663855234781903,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.self_attn.v_proj": {
 "mse": 2.537128693802515e-06,
 "rmse": 0.001592836681459375,
 "relative_error": 0.1253849898161065,
 "max_error": 0.052634596824645996,
 "sparsity": 0.3911902109781901,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.self_attn.q_proj": {
 "mse": 7.867356544011272e-06,
 "rmse": 0.002804880843103905,
 "relative_error": 0.11485303222807783,
 "max_error": 0.12257003784179688,
 "sparsity": 0.3965638478597005,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.self_attn.out_proj": {
 "mse": 2.3530305952590425e-06,
 "rmse": 0.0015339591243768663,
 "relative_error": 0.12316348329498344,
 "max_error": 0.05019688606262207,
 "sparsity": 0.3834683100382487,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.encoder_attn.k_proj": {
 "mse": 2.932305505964905e-06,
 "rmse": 0.0017123975899203156,
 "relative_error": 0.14878440781545277,
 "max_error": 0.052741289138793945,
 "sparsity": 0.38413047790527344,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.encoder_attn.v_proj": {
 "mse": 2.9527541300922167e-06,
 "rmse": 0.0017183579749552235,
 "relative_error": 0.15262952638414398,
 "max_error": 0.04509568214416504,
 "sparsity": 0.3845523198445638,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.encoder_attn.q_proj": {
 "mse": 1.7530026070744498e-06,
 "rmse": 0.001324010047950713,
 "relative_error": 0.11584667248381872,
 "max_error": 0.0369831919670105,
 "sparsity": 0.38655567169189453,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.encoder_attn.out_proj": {
 "mse": 3.997231488028774e-06,
 "rmse": 0.0019993077522054412,
 "relative_error": 0.1804483932496295,
 "max_error": 0.04954719543457031,
 "sparsity": 0.3609612782796224,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.fc1": {
 "mse": 3.4595873330545146e-06,
 "rmse": 0.0018599965949040106,
 "relative_error": 0.12480082603415626,
 "max_error": 0.06718039512634277,
 "sparsity": 0.39367198944091797,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.0.fc2": {
 "mse": 9.965755452867597e-06,
 "rmse": 0.0031568584784351034,
 "relative_error": 0.258659206797803,
 "max_error": 0.10831642150878906,
 "sparsity": 0.3432497978210449,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.self_attn.k_proj": {
 "mse": 2.03306717594387e-06,
 "rmse": 0.0014258566463511927,
 "relative_error": 0.10631741618956413,
 "max_error": 0.050121307373046875,
 "sparsity": 0.38936614990234375,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.self_attn.v_proj": {
 "mse": 1.2314576451899484e-06,
 "rmse": 0.0011097106132636329,
 "relative_error": 0.12502401296497923,
 "max_error": 0.036118507385253906,
 "sparsity": 0.39406808217366535,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.self_attn.q_proj": {
 "mse": 2.091504939016886e-06,
 "rmse": 0.0014462036298588405,
 "relative_error": 0.114342204814357,
 "max_error": 0.052300095558166504,
 "sparsity": 0.3865979512532552,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.self_attn.out_proj": {
 "mse": 2.3585996586916735e-06,
 "rmse": 0.0015357733096689997,
 "relative_error": 0.16983263226843545,
 "max_error": 0.055203378200531006,
 "sparsity": 0.36684226989746094,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.encoder_attn.k_proj": {
 "mse": 2.569649268480134e-06,
 "rmse": 0.0016030125603001787,
 "relative_error": 0.145535704383827,
 "max_error": 0.061995506286621094,
 "sparsity": 0.3825670878092448,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.encoder_attn.v_proj": {
 "mse": 2.6175509901804617e-06,
 "rmse": 0.001617884727099079,
 "relative_error": 0.152717885751071,
 "max_error": 0.05068695545196533,
 "sparsity": 0.385009765625,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.encoder_attn.q_proj": {
 "mse": 1.3694378822037834e-06,
 "rmse": 0.0011702298416139383,
 "relative_error": 0.10105443997993861,
 "max_error": 0.051283836364746094,
 "sparsity": 0.3925361633300781,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.encoder_attn.out_proj": {
 "mse": 3.2205025490839034e-06,
 "rmse": 0.001794575868857013,
 "relative_error": 0.1678963406077039,
 "max_error": 0.07158446311950684,
 "sparsity": 0.3641344706217448,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.fc1": {
 "mse": 2.8058025236532558e-06,
 "rmse": 0.001675052991297068,
 "relative_error": 0.12601454584557661,
 "max_error": 0.08018875122070312,
 "sparsity": 0.38286805152893066,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.1.fc2": {
 "mse": 7.15729493094841e-06,
 "rmse": 0.0026753121184169165,
 "relative_error": 0.2385123906472085,
 "max_error": 0.1420459747314453,
 "sparsity": 0.37098320325215656,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.self_attn.k_proj": {
 "mse": 1.408790922141634e-06,
 "rmse": 0.0011869249858949106,
 "relative_error": 0.15586319641563348,
 "max_error": 0.05228686332702637,
 "sparsity": 0.38851292928059894,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.self_attn.v_proj": {
 "mse": 2.4211458367062733e-06,
 "rmse": 0.0015560031608921215,
 "relative_error": 0.19859374281883027,
 "max_error": 0.0744476318359375,
 "sparsity": 0.39120133717854816,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.self_attn.q_proj": {
 "mse": 1.3298435987962876e-06,
 "rmse": 0.0011531884489519861,
 "relative_error": 0.1712732837930552,
 "max_error": 0.044757843017578125,
 "sparsity": 0.38156604766845703,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.self_attn.out_proj": {
 "mse": 3.029804929610691e-06,
 "rmse": 0.0017406334851457646,
 "relative_error": 0.2584208819885451,
 "max_error": 0.07974696159362793,
 "sparsity": 0.35305945078531903,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.encoder_attn.k_proj": {
 "mse": 2.2631470528722275e-06,
 "rmse": 0.001504375967925647,
 "relative_error": 0.1425945197962898,
 "max_error": 0.04911303520202637,
 "sparsity": 0.3848470052083333,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.encoder_attn.v_proj": {
 "mse": 3.066921181016369e-06,
 "rmse": 0.0017512627390018805,
 "relative_error": 0.15523498710637637,
 "max_error": 0.043721556663513184,
 "sparsity": 0.3850396474202474,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.encoder_attn.q_proj": {
 "mse": 1.306824174207577e-06,
 "rmse": 0.0011431641064202362,
 "relative_error": 0.10384071954394945,
 "max_error": 0.0584259033203125,
 "sparsity": 0.39062023162841797,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.encoder_attn.out_proj": {
 "mse": 2.39322457673552e-06,
 "rmse": 0.0015470050344893902,
 "relative_error": 0.13837125067483497,
 "max_error": 0.03503692150115967,
 "sparsity": 0.3704687754313151,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.fc1": {
 "mse": 2.6717477794591105e-06,
 "rmse": 0.0016345481881728388,
 "relative_error": 0.13424853534381614,
 "max_error": 0.12715303897857666,
 "sparsity": 0.37990784645080566,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.2.fc2": {
 "mse": 6.862672307761386e-06,
 "rmse": 0.002619670266991895,
 "relative_error": 0.24381065264040552,
 "max_error": 0.22676372528076172,
 "sparsity": 0.3882887363433838,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.self_attn.k_proj": {
 "mse": 1.7996961787503096e-06,
 "rmse": 0.0013415275542270123,
 "relative_error": 0.1192658384835929,
 "max_error": 0.04256439208984375,
 "sparsity": 0.37576770782470703,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.self_attn.v_proj": {
 "mse": 2.0297898117860314e-06,
 "rmse": 0.0014247069213652438,
 "relative_error": 0.13746902832804708,
 "max_error": 0.03883171081542969,
 "sparsity": 0.38527584075927734,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.self_attn.q_proj": {
 "mse": 2.1074170035717543e-06,
 "rmse": 0.0014516945283260367,
 "relative_error": 0.11590249312665402,
 "max_error": 0.07066059112548828,
 "sparsity": 0.38370641072591144,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.self_attn.out_proj": {
 "mse": 1.333709519713011e-06,
 "rmse": 0.0011548634203718685,
 "relative_error": 0.117272895602355,
 "max_error": 0.07680273056030273,
 "sparsity": 0.4009917577107747,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.encoder_attn.k_proj": {
 "mse": 2.471836751283263e-06,
 "rmse": 0.001572207604384123,
 "relative_error": 0.14458506102228255,
 "max_error": 0.04527485370635986,
 "sparsity": 0.3834635416666667,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.encoder_attn.v_proj": {
 "mse": 3.16734735861246e-06,
 "rmse": 0.0017797042896538907,
 "relative_error": 0.15301164247019403,
 "max_error": 0.059139907360076904,
 "sparsity": 0.3846127192179362,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.encoder_attn.q_proj": {
 "mse": 1.4525229516948457e-06,
 "rmse": 0.001205206601249282,
 "relative_error": 0.10425491084889804,
 "max_error": 0.09468531608581543,
 "sparsity": 0.3902708689371745,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.encoder_attn.out_proj": {
 "mse": 2.678065584404976e-06,
 "rmse": 0.0016364796315276814,
 "relative_error": 0.1430544824579208,
 "max_error": 0.051494598388671875,
 "sparsity": 0.37117163340250653,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.fc1": {
 "mse": 2.3618308659933973e-06,
 "rmse": 0.0015368249301704463,
 "relative_error": 0.1245729986149135,
 "max_error": 0.05836009979248047,
 "sparsity": 0.37219882011413574,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.3.fc2": {
 "mse": 1.5611518392688595e-05,
 "rmse": 0.003951141403783037,
 "relative_error": 0.3720975771798556,
 "max_error": 0.5879669189453125,
 "sparsity": 0.38265403111775714,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.self_attn.k_proj": {
 "mse": 2.8701028895739e-06,
 "rmse": 0.0016941378012351592,
 "relative_error": 0.11716970889271452,
 "max_error": 0.055669307708740234,
 "sparsity": 0.3828910191853841,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.self_attn.v_proj": {
 "mse": 1.9905580757040298e-06,
 "rmse": 0.0014108713887892226,
 "relative_error": 0.12301872679402027,
 "max_error": 0.053325772285461426,
 "sparsity": 0.3898334503173828,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.self_attn.q_proj": {
 "mse": 2.990354914800264e-06,
 "rmse": 0.0017292642697980733,
 "relative_error": 0.11543029947746948,
 "max_error": 0.06011366844177246,
 "sparsity": 0.3869177500406901,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.self_attn.out_proj": {
 "mse": 1.7686877527012257e-06,
 "rmse": 0.001329920205388739,
 "relative_error": 0.1234683605164969,
 "max_error": 0.06157732009887695,
 "sparsity": 0.3810313542683919,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.encoder_attn.k_proj": {
 "mse": 2.4499702249158872e-06,
 "rmse": 0.0015652380729192243,
 "relative_error": 0.14753199522550947,
 "max_error": 0.04569506645202637,
 "sparsity": 0.38436158498128253,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.encoder_attn.v_proj": {
 "mse": 3.2714881399442675e-06,
 "rmse": 0.0018087255568339459,
 "relative_error": 0.15390755351674237,
 "max_error": 0.04762768745422363,
 "sparsity": 0.38466930389404297,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.encoder_attn.q_proj": {
 "mse": 1.2864924201494432e-06,
 "rmse": 0.001134236492160891,
 "relative_error": 0.10177971901165225,
 "max_error": 0.0303804874420166,
 "sparsity": 0.39114030202229816,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.encoder_attn.out_proj": {
 "mse": 2.4311630113516003e-06,
 "rmse": 0.0015592187182533438,
 "relative_error": 0.13463845742763503,
 "max_error": 0.04355788230895996,
 "sparsity": 0.3733196258544922,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.fc1": {
 "mse": 2.469627816026332e-06,
 "rmse": 0.001571504952593638,
 "relative_error": 0.12524985734406285,
 "max_error": 0.08176088333129883,
 "sparsity": 0.3696654637654622,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.4.fc2": {
 "mse": 5.365253400668735e-06,
 "rmse": 0.002316301664435946,
 "relative_error": 0.21985876196410048,
 "max_error": 0.16083449125289917,
 "sparsity": 0.3907614549001058,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.self_attn.k_proj": {
 "mse": 2.7977457648376003e-06,
 "rmse": 0.0016726463358515452,
 "relative_error": 0.10678882412118908,
 "max_error": 0.06360578536987305,
 "sparsity": 0.3896319071451823,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.self_attn.v_proj": {
 "mse": 2.054051037703175e-06,
 "rmse": 0.0014331960918531612,
 "relative_error": 0.11478326691502795,
 "max_error": 0.04179668426513672,
 "sparsity": 0.3933906555175781,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.self_attn.q_proj": {
 "mse": 2.887436266973964e-06,
 "rmse": 0.0016992457935725378,
 "relative_error": 0.10747942741596392,
 "max_error": 0.07212746143341064,
 "sparsity": 0.39190673828125,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.self_attn.out_proj": {
 "mse": 2.422082388875424e-06,
 "rmse": 0.0015563040798235491,
 "relative_error": 0.12821936986008606,
 "max_error": 0.03042781352996826,
 "sparsity": 0.37738768259684247,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.encoder_attn.k_proj": {
 "mse": 2.1594835288851755e-06,
 "rmse": 0.0014695181281240376,
 "relative_error": 0.14820830447000122,
 "max_error": 0.05715823173522949,
 "sparsity": 0.38480599721272785,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.encoder_attn.v_proj": {
 "mse": 2.402921154498472e-06,
 "rmse": 0.0015501358503365026,
 "relative_error": 0.15172739547983966,
 "max_error": 0.049695730209350586,
 "sparsity": 0.38500118255615234,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.encoder_attn.q_proj": {
 "mse": 1.2580545671880827e-06,
 "rmse": 0.0011216303166320365,
 "relative_error": 0.10791561656075024,
 "max_error": 0.07042884826660156,
 "sparsity": 0.39124329884847003,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.encoder_attn.out_proj": {
 "mse": 2.0925813259964343e-06,
 "rmse": 0.0014465757242524272,
 "relative_error": 0.1422147514369969,
 "max_error": 0.05846059322357178,
 "sparsity": 0.37059179941813153,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.fc1": {
 "mse": 3.071619175898377e-06,
 "rmse": 0.0017526035421333533,
 "relative_error": 0.12686562610632215,
 "max_error": 0.0797877311706543,
 "sparsity": 0.3853236834208171,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.5.fc2": {
 "mse": 5.129960754857166e-06,
 "rmse": 0.002264941666987732,
 "relative_error": 0.1860462585742565,
 "max_error": 0.05859053134918213,
 "sparsity": 0.38653842608133954,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.self_attn.k_proj": {
 "mse": 2.7227015380049124e-06,
 "rmse": 0.001650061070992499,
 "relative_error": 0.1084165589968581,
 "max_error": 0.05523967742919922,
 "sparsity": 0.3882598876953125,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.self_attn.v_proj": {
 "mse": 1.960485178642557e-06,
 "rmse": 0.0014001732673646348,
 "relative_error": 0.11180013694606943,
 "max_error": 0.06603431701660156,
 "sparsity": 0.3924814860026042,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.self_attn.q_proj": {
 "mse": 2.7193116238777293e-06,
 "rmse": 0.0016490335423749662,
 "relative_error": 0.10691578744139109,
 "max_error": 0.04374641180038452,
 "sparsity": 0.3899396260579427,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.self_attn.out_proj": {
 "mse": 2.0261477402527817e-06,
 "rmse": 0.0014234281647672923,
 "relative_error": 0.11932234164492085,
 "max_error": 0.05954170227050781,
 "sparsity": 0.3812958399454753,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.encoder_attn.k_proj": {
 "mse": 1.914561835292261e-06,
 "rmse": 0.0013836769259087402,
 "relative_error": 0.1409572228796846,
 "max_error": 0.05499941110610962,
 "sparsity": 0.3855908711751302,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.encoder_attn.v_proj": {
 "mse": 2.064477939711651e-06,
 "rmse": 0.0014368291268315975,
 "relative_error": 0.15090229510718225,
 "max_error": 0.041681647300720215,
 "sparsity": 0.38470935821533203,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.encoder_attn.q_proj": {
 "mse": 1.2734592473861994e-06,
 "rmse": 0.0011284765160986733,
 "relative_error": 0.10325489365368369,
 "max_error": 0.03181350231170654,
 "sparsity": 0.3910401662190755,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.encoder_attn.out_proj": {
 "mse": 1.8945763713418273e-06,
 "rmse": 0.0013764361123357043,
 "relative_error": 0.1406672765822477,
 "max_error": 0.03437471389770508,
 "sparsity": 0.37631797790527344,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.fc1": {
 "mse": 3.0933047128200997e-06,
 "rmse": 0.0017587793246510773,
 "relative_error": 0.12332660056166439,
 "max_error": 0.09690666198730469,
 "sparsity": 0.3895715077718099,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.6.fc2": {
 "mse": 5.023699941375526e-06,
 "rmse": 0.002241361180482861,
 "relative_error": 0.17801352059730607,
 "max_error": 0.07062911987304688,
 "sparsity": 0.38503503799438477,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.self_attn.k_proj": {
 "mse": 2.669979267011513e-06,
 "rmse": 0.001634007119633055,
 "relative_error": 0.10356912057004246,
 "max_error": 0.05674147605895996,
 "sparsity": 0.39308420817057294,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.self_attn.v_proj": {
 "mse": 1.693527110546711e-06,
 "rmse": 0.0013013558739048711,
 "relative_error": 0.10513660505308145,
 "max_error": 0.04603254795074463,
 "sparsity": 0.39510377248128253,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.self_attn.q_proj": {
 "mse": 2.754891966105788e-06,
 "rmse": 0.0016597867230779345,
 "relative_error": 0.10414344160911279,
 "max_error": 0.0881616473197937,
 "sparsity": 0.3932339350382487,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.self_attn.out_proj": {
 "mse": 2.3079376205714652e-06,
 "rmse": 0.0015191897908330826,
 "relative_error": 0.12670858213626052,
 "max_error": 0.0412135124206543,
 "sparsity": 0.37942059834798175,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.encoder_attn.k_proj": {
 "mse": 3.342863919897354e-06,
 "rmse": 0.0018283500539823753,
 "relative_error": 0.15063672352428492,
 "max_error": 0.0661466121673584,
 "sparsity": 0.3860619862874349,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.encoder_attn.v_proj": {
 "mse": 2.972983565996401e-06,
 "rmse": 0.00172423419696873,
 "relative_error": 0.15345657818512687,
 "max_error": 0.21555328369140625,
 "sparsity": 0.38444073994954425,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.encoder_attn.q_proj": {
 "mse": 1.6744468211982166e-06,
 "rmse": 0.0012940041812908552,
 "relative_error": 0.10236945613803666,
 "max_error": 0.060260772705078125,
 "sparsity": 0.39227835337320965,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.encoder_attn.out_proj": {
 "mse": 2.9717300549236825e-06,
 "rmse": 0.0017238706607294186,
 "relative_error": 0.15377582478905313,
 "max_error": 0.08759880065917969,
 "sparsity": 0.3660281499226888,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.fc1": {
 "mse": 2.8657561870204518e-06,
 "rmse": 0.001692854449449347,
 "relative_error": 0.11476676538134392,
 "max_error": 0.1609211564064026,
 "sparsity": 0.39171473185221356,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.7.fc2": {
 "mse": 5.369829978008056e-06,
 "rmse": 0.0023172893600083818,
 "relative_error": 0.17915047105444631,
 "max_error": 0.06115150451660156,
 "sparsity": 0.3791286150614421,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.self_attn.k_proj": {
 "mse": 2.874462779800524e-06,
 "rmse": 0.001695424070785986,
 "relative_error": 0.10408217734095601,
 "max_error": 0.04898780584335327,
 "sparsity": 0.39197508494059247,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.self_attn.v_proj": {
 "mse": 1.954801291503827e-06,
 "rmse": 0.0013981420855921,
 "relative_error": 0.10715554077548149,
 "max_error": 0.04911994934082031,
 "sparsity": 0.39489396413167316,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.self_attn.q_proj": {
 "mse": 2.988884716614848e-06,
 "rmse": 0.0017288391239831565,
 "relative_error": 0.10488478135736144,
 "max_error": 0.052776336669921875,
 "sparsity": 0.3927752176920573,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.self_attn.out_proj": {
 "mse": 2.2270126009971136e-06,
 "rmse": 0.0014923178619171969,
 "relative_error": 0.11837970535523554,
 "max_error": 0.04682779312133789,
 "sparsity": 0.38154157002766925,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.encoder_attn.k_proj": {
 "mse": 4.145021193835419e-06,
 "rmse": 0.002035932512102358,
 "relative_error": 0.14794230558981988,
 "max_error": 0.06369626522064209,
 "sparsity": 0.38465213775634766,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.encoder_attn.v_proj": {
 "mse": 3.111481191808707e-06,
 "rmse": 0.0017639391122736372,
 "relative_error": 0.15156701666066377,
 "max_error": 0.04685235023498535,
 "sparsity": 0.38491566975911456,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.encoder_attn.q_proj": {
 "mse": 2.2250369511311874e-06,
 "rmse": 0.0014916557750135207,
 "relative_error": 0.10434207297549174,
 "max_error": 0.050441741943359375,
 "sparsity": 0.39132245381673175,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.encoder_attn.out_proj": {
 "mse": 3.1934757771523437e-06,
 "rmse": 0.0017870298758421313,
 "relative_error": 0.15283987169747038,
 "max_error": 0.08969151973724365,
 "sparsity": 0.3676007588704427,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.fc1": {
 "mse": 2.79184632745455e-06,
 "rmse": 0.0016708819011092765,
 "relative_error": 0.11036536299564526,
 "max_error": 0.12213349342346191,
 "sparsity": 0.3927458127339681,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.8.fc2": {
 "mse": 5.2020336624991614e-06,
 "rmse": 0.002280796716610045,
 "relative_error": 0.17299622967154382,
 "max_error": 0.07349264621734619,
 "sparsity": 0.3765496412913005,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.self_attn.k_proj": {
 "mse": 3.0386931939574424e-06,
 "rmse": 0.0017431847847997762,
 "relative_error": 0.1058218625977078,
 "max_error": 0.0761750340461731,
 "sparsity": 0.3904520670572917,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.self_attn.v_proj": {
 "mse": 2.23638562601991e-06,
 "rmse": 0.0014954549896335596,
 "relative_error": 0.11097779406996414,
 "max_error": 0.041104793548583984,
 "sparsity": 0.3938833872477214,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.self_attn.q_proj": {
 "mse": 3.189991730323527e-06,
 "rmse": 0.0017860547948827122,
 "relative_error": 0.10832728289648673,
 "max_error": 0.06943631172180176,
 "sparsity": 0.3921623229980469,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.self_attn.out_proj": {
 "mse": 2.696985120564932e-06,
 "rmse": 0.001642250017678469,
 "relative_error": 0.1281785123233537,
 "max_error": 0.18055415153503418,
 "sparsity": 0.37871678670247394,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.encoder_attn.k_proj": {
 "mse": 3.2500684028491378e-06,
 "rmse": 0.0018027946091690916,
 "relative_error": 0.14403063462528215,
 "max_error": 0.056221961975097656,
 "sparsity": 0.3845011393229167,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.encoder_attn.v_proj": {
 "mse": 3.189758899679873e-06,
 "rmse": 0.0017859896135419917,
 "relative_error": 0.15269143882686897,
 "max_error": 0.04081469774246216,
 "sparsity": 0.38484986623128253,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.encoder_attn.q_proj": {
 "mse": 1.9499902919051237e-06,
 "rmse": 0.001396420528316998,
 "relative_error": 0.10472704022264365,
 "max_error": 0.03749656677246094,
 "sparsity": 0.3913602828979492,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.encoder_attn.out_proj": {
 "mse": 2.9134057513147127e-06,
 "rmse": 0.0017068701624068283,
 "relative_error": 0.14487531428174485,
 "max_error": 0.10330677032470703,
 "sparsity": 0.3698616027832031,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.fc1": {
 "mse": 2.8058291263732826e-06,
 "rmse": 0.0016750609321374798,
 "relative_error": 0.1102062185498671,
 "max_error": 0.19041013717651367,
 "sparsity": 0.3934587637583415,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.9.fc2": {
 "mse": 5.31405476067448e-06,
 "rmse": 0.002305223364594954,
 "relative_error": 0.17306741105103132,
 "max_error": 0.0740041732788086,
 "sparsity": 0.37359078725179035,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.self_attn.k_proj": {
 "mse": 2.8235283480171347e-06,
 "rmse": 0.0016803357843053676,
 "relative_error": 0.10581116696721167,
 "max_error": 0.12559711933135986,
 "sparsity": 0.39281590779622394,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.self_attn.v_proj": {
 "mse": 2.129301947206841e-06,
 "rmse": 0.0014592127833893318,
 "relative_error": 0.10461520351896639,
 "max_error": 0.06537044048309326,
 "sparsity": 0.39534536997477215,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.self_attn.q_proj": {
 "mse": 2.9481207093340345e-06,
 "rmse": 0.0017170092339105327,
 "relative_error": 0.10566365640016753,
 "max_error": 0.0826263427734375,
 "sparsity": 0.3939990997314453,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.self_attn.out_proj": {
 "mse": 3.0388287086680066e-06,
 "rmse": 0.0017432236542302903,
 "relative_error": 0.12941463793508615,
 "max_error": 0.04368305206298828,
 "sparsity": 0.37840938568115234,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.encoder_attn.k_proj": {
 "mse": 4.262691163603449e-06,
 "rmse": 0.002064628577638954,
 "relative_error": 0.1455730293477049,
 "max_error": 0.06619501113891602,
 "sparsity": 0.3850278854370117,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.encoder_attn.v_proj": {
 "mse": 4.141321824135957e-06,
 "rmse": 0.002035023789574942,
 "relative_error": 0.15507668501761301,
 "max_error": 0.046338558197021484,
 "sparsity": 0.38466326395670575,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.encoder_attn.q_proj": {
 "mse": 2.3088205125532113e-06,
 "rmse": 0.0015194803429308362,
 "relative_error": 0.10376620421198642,
 "max_error": 0.04837995767593384,
 "sparsity": 0.3922373453776042,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.encoder_attn.out_proj": {
 "mse": 3.749622464965796e-06,
 "rmse": 0.0019363941915234604,
 "relative_error": 0.15077165945466575,
 "max_error": 0.043268442153930664,
 "sparsity": 0.3687426249186198,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.fc1": {
 "mse": 2.742124252108624e-06,
 "rmse": 0.0016559360652237225,
 "relative_error": 0.10881992614733726,
 "max_error": 0.09426462650299072,
 "sparsity": 0.39169470469156903,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.10.fc2": {
 "mse": 5.248308752925368e-06,
 "rmse": 0.0022909187573821486,
 "relative_error": 0.16801593877256882,
 "max_error": 0.09567546844482422,
 "sparsity": 0.37460168202718097,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.self_attn.k_proj": {
 "mse": 3.0592714210797567e-06,
 "rmse": 0.0017490773056328176,
 "relative_error": 0.10734568285249818,
 "max_error": 0.09254008531570435,
 "sparsity": 0.3920777638753255,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.self_attn.v_proj": {
 "mse": 2.1376101813075365e-06,
 "rmse": 0.0014620568324478828,
 "relative_error": 0.10559583271255113,
 "max_error": 0.029812395572662354,
 "sparsity": 0.3952624003092448,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.self_attn.q_proj": {
 "mse": 3.1045001378515735e-06,
 "rmse": 0.0017619591759889255,
 "relative_error": 0.10693716839240887,
 "max_error": 0.090118408203125,
 "sparsity": 0.3924741744995117,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.self_attn.out_proj": {
 "mse": 3.1066649626154685e-06,
 "rmse": 0.0017625733921217206,
 "relative_error": 0.13225657965413115,
 "max_error": 0.06195068359375,
 "sparsity": 0.3758252461751302,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.encoder_attn.k_proj": {
 "mse": 4.285246177460067e-06,
 "rmse": 0.0020700836160551747,
 "relative_error": 0.14720766834565957,
 "max_error": 0.06798171997070312,
 "sparsity": 0.3851782480875651,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.encoder_attn.v_proj": {
 "mse": 4.828083092434099e-06,
 "rmse": 0.0021972899427326604,
 "relative_error": 0.1586933413077182,
 "max_error": 0.05201244354248047,
 "sparsity": 0.3848873774210612,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.encoder_attn.q_proj": {
 "mse": 2.283047706441721e-06,
 "rmse": 0.001510975746476998,
 "relative_error": 0.10365660136115314,
 "max_error": 0.048856914043426514,
 "sparsity": 0.3916358947753906,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.encoder_attn.out_proj": {
 "mse": 4.28708926847321e-06,
 "rmse": 0.00207052874128161,
 "relative_error": 0.15415400301442833,
 "max_error": 0.04470396041870117,
 "sparsity": 0.3667583465576172,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.fc1": {
 "mse": 2.7612836674961727e-06,
 "rmse": 0.001661711066189358,
 "relative_error": 0.10893671218267384,
 "max_error": 0.1811065673828125,
 "sparsity": 0.3911479314168294,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.11.fc2": {
 "mse": 5.29205226484919e-06,
 "rmse": 0.002300446101270184,
 "relative_error": 0.16487893138147464,
 "max_error": 0.11721420288085938,
 "sparsity": 0.373733123143514,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.self_attn.k_proj": {
 "mse": 2.9784519028908107e-06,
 "rmse": 0.0017258191976249455,
 "relative_error": 0.10706666687076272,
 "max_error": 0.10212254524230957,
 "sparsity": 0.3907483418782552,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.self_attn.v_proj": {
 "mse": 2.395839828750468e-06,
 "rmse": 0.0015478500666248225,
 "relative_error": 0.10913041404713451,
 "max_error": 0.05898064374923706,
 "sparsity": 0.39325809478759766,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.self_attn.q_proj": {
 "mse": 3.0919227356207557e-06,
 "rmse": 0.0017583864011134628,
 "relative_error": 0.10877376011568239,
 "max_error": 0.09144771099090576,
 "sparsity": 0.39208539326985675,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.self_attn.out_proj": {
 "mse": 3.2778616514406167e-06,
 "rmse": 0.0018104865786413929,
 "relative_error": 0.13259861849468188,
 "max_error": 0.05882430076599121,
 "sparsity": 0.3744872411092122,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.encoder_attn.k_proj": {
 "mse": 4.489171260502189e-06,
 "rmse": 0.002118766447842279,
 "relative_error": 0.14508222072774984,
 "max_error": 0.06339359283447266,
 "sparsity": 0.3842233022054036,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.encoder_attn.v_proj": {
 "mse": 4.473969056562055e-06,
 "rmse": 0.0021151758925824714,
 "relative_error": 0.1564127825375384,
 "max_error": 0.05270886421203613,
 "sparsity": 0.38474464416503906,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.encoder_attn.q_proj": {
 "mse": 2.4208638933487236e-06,
 "rmse": 0.0015559125596731726,
 "relative_error": 0.10374680012869537,
 "max_error": 0.03216981887817383,
 "sparsity": 0.39106400807698566,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.encoder_attn.out_proj": {
 "mse": 4.03118428948801e-06,
 "rmse": 0.002007780936628299,
 "relative_error": 0.15214832136551018,
 "max_error": 0.05477547645568848,
 "sparsity": 0.3683900833129883,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.fc1": {
 "mse": 2.6280795282218605e-06,
 "rmse": 0.0016211352590767558,
 "relative_error": 0.10722597587911696,
 "max_error": 0.1983776092529297,
 "sparsity": 0.3908070723215739,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.12.fc2": {
 "mse": 4.76806872029556e-06,
 "rmse": 0.0021835907859064528,
 "relative_error": 0.15451343321763653,
 "max_error": 0.06719207763671875,
 "sparsity": 0.37807003657023114,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.self_attn.k_proj": {
 "mse": 3.1032777769723907e-06,
 "rmse": 0.0017616122663549976,
 "relative_error": 0.10859283459981478,
 "max_error": 0.08640223741531372,
 "sparsity": 0.3908227284749349,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.self_attn.v_proj": {
 "mse": 2.4164837668649852e-06,
 "rmse": 0.0015545043476507183,
 "relative_error": 0.11088639104509466,
 "max_error": 0.09391403198242188,
 "sparsity": 0.3939778010050456,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.self_attn.q_proj": {
 "mse": 3.1459003366762772e-06,
 "rmse": 0.0017736686095988386,
 "relative_error": 0.10887852983292154,
 "max_error": 0.0973324179649353,
 "sparsity": 0.39158089955647785,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.self_attn.out_proj": {
 "mse": 3.010093223565491e-06,
 "rmse": 0.001734962023666654,
 "relative_error": 0.12833051491091624,
 "max_error": 0.08125495910644531,
 "sparsity": 0.37650807698567706,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.encoder_attn.k_proj": {
 "mse": 4.259429260855541e-06,
 "rmse": 0.002063838477414243,
 "relative_error": 0.14530331557484114,
 "max_error": 0.061621904373168945,
 "sparsity": 0.38376013437906903,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.encoder_attn.v_proj": {
 "mse": 4.011252713098656e-06,
 "rmse": 0.002002811202559706,
 "relative_error": 0.15357150644740222,
 "max_error": 0.05202364921569824,
 "sparsity": 0.38488419850667316,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.encoder_attn.q_proj": {
 "mse": 2.3202264856081456e-06,
 "rmse": 0.0015232289669016097,
 "relative_error": 0.10355130654722193,
 "max_error": 0.04495662450790405,
 "sparsity": 0.3904116948445638,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.encoder_attn.out_proj": {
 "mse": 3.6354515486891614e-06,
 "rmse": 0.001906686012087245,
 "relative_error": 0.14821881625212166,
 "max_error": 0.07686328887939453,
 "sparsity": 0.36955706278483075,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.fc1": {
 "mse": 2.724333626247244e-06,
 "rmse": 0.0016505555507910796,
 "relative_error": 0.10972522648773587,
 "max_error": 0.21759366989135742,
 "sparsity": 0.39084283510843915,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.13.fc2": {
 "mse": 4.931787771056406e-06,
 "rmse": 0.0022207628804211416,
 "relative_error": 0.15521514437074152,
 "max_error": 0.09429550170898438,
 "sparsity": 0.37915460268656415,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.self_attn.k_proj": {
 "mse": 3.087111963395728e-06,
 "rmse": 0.0017570179177787938,
 "relative_error": 0.1092595745005226,
 "max_error": 0.09504544734954834,
 "sparsity": 0.390896479288737,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.self_attn.v_proj": {
 "mse": 2.5206579721270828e-06,
 "rmse": 0.001587658014853036,
 "relative_error": 0.11081719637216424,
 "max_error": 0.051631927490234375,
 "sparsity": 0.39329179128011066,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.self_attn.q_proj": {
 "mse": 3.190672487107804e-06,
 "rmse": 0.0017862453602760747,
 "relative_error": 0.11018739252519799,
 "max_error": 0.08269834518432617,
 "sparsity": 0.3911660512288411,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.self_attn.out_proj": {
 "mse": 3.2235734579444397e-06,
 "rmse": 0.0017954312735230051,
 "relative_error": 0.12984946467693861,
 "max_error": 0.07321488857269287,
 "sparsity": 0.37707042694091797,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.encoder_attn.k_proj": {
 "mse": 4.632142918126192e-06,
 "rmse": 0.002152241370786788,
 "relative_error": 0.14750340403551016,
 "max_error": 0.06115221977233887,
 "sparsity": 0.3844248453776042,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.encoder_attn.v_proj": {
 "mse": 4.610026280715829e-06,
 "rmse": 0.0021470971754244915,
 "relative_error": 0.15607740424439667,
 "max_error": 0.05488014221191406,
 "sparsity": 0.3850485483805339,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.encoder_attn.q_proj": {
 "mse": 2.4240475795522798e-06,
 "rmse": 0.0015569353164317007,
 "relative_error": 0.10406009770556121,
 "max_error": 0.042209625244140625,
 "sparsity": 0.3910134633382161,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.encoder_attn.out_proj": {
 "mse": 3.991907306044595e-06,
 "rmse": 0.001997975802166932,
 "relative_error": 0.14917388540435966,
 "max_error": 0.1012420654296875,
 "sparsity": 0.3692118326822917,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.fc1": {
 "mse": 2.651945123943733e-06,
 "rmse": 0.001628479390088721,
 "relative_error": 0.10843630002577104,
 "max_error": 0.17707645893096924,
 "sparsity": 0.3903570969899495,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.14.fc2": {
 "mse": 5.093220352136996e-06,
 "rmse": 0.0022568164196799427,
 "relative_error": 0.15510645443953647,
 "max_error": 0.09729671478271484,
 "sparsity": 0.378409465154012,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.self_attn.k_proj": {
 "mse": 2.952674776679487e-06,
 "rmse": 0.00171833488490442,
 "relative_error": 0.11039714548231819,
 "max_error": 0.05137348175048828,
 "sparsity": 0.39000829060872394,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.self_attn.v_proj": {
 "mse": 2.5764302336028777e-06,
 "rmse": 0.0016051262360334397,
 "relative_error": 0.11392904939485686,
 "max_error": 0.07793331146240234,
 "sparsity": 0.3936421076456706,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.self_attn.q_proj": {
 "mse": 3.0681758289574645e-06,
 "rmse": 0.0017516209147408193,
 "relative_error": 0.111066066762652,
 "max_error": 0.061458587646484375,
 "sparsity": 0.3910897572835286,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.self_attn.out_proj": {
 "mse": 3.2459129215567373e-06,
 "rmse": 0.0018016417295224757,
 "relative_error": 0.13252200909794726,
 "max_error": 0.17857837677001953,
 "sparsity": 0.3764670689900716,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.encoder_attn.k_proj": {
 "mse": 3.870277396345045e-06,
 "rmse": 0.001967302060270625,
 "relative_error": 0.1451358525235946,
 "max_error": 0.05497002601623535,
 "sparsity": 0.38480281829833984,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.encoder_attn.v_proj": {
 "mse": 4.471807187655941e-06,
 "rmse": 0.002114664793213322,
 "relative_error": 0.15706173162807832,
 "max_error": 0.07093095779418945,
 "sparsity": 0.3840920130411784,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.encoder_attn.q_proj": {
 "mse": 2.2638698737864615e-06,
 "rmse": 0.0015046161881976617,
 "relative_error": 0.10758771192543139,
 "max_error": 0.06101179122924805,
 "sparsity": 0.389739990234375,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.encoder_attn.out_proj": {
 "mse": 4.6661739361297805e-06,
 "rmse": 0.00216013285150006,
 "relative_error": 0.164261855187405,
 "max_error": 0.16168975830078125,
 "sparsity": 0.36803849538167316,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.fc1": {
 "mse": 2.6208585950371344e-06,
 "rmse": 0.0016189066047913741,
 "relative_error": 0.10883047265246418,
 "max_error": 0.2597169876098633,
 "sparsity": 0.3904755115509033,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.15.fc2": {
 "mse": 5.432469151855912e-06,
 "rmse": 0.002330765786572283,
 "relative_error": 0.15863985173106748,
 "max_error": 0.09306836128234863,
 "sparsity": 0.37816667556762695,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.self_attn.k_proj": {
 "mse": 2.975872803290258e-06,
 "rmse": 0.0017250718255453186,
 "relative_error": 0.11149507911807895,
 "max_error": 0.0687759518623352,
 "sparsity": 0.38975874582926434,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.self_attn.v_proj": {
 "mse": 2.708741931201075e-06,
 "rmse": 0.001645825607772912,
 "relative_error": 0.11401836328792606,
 "max_error": 0.10725784301757812,
 "sparsity": 0.3934755325317383,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.self_attn.q_proj": {
 "mse": 3.0313924526126357e-06,
 "rmse": 0.0017410894441735713,
 "relative_error": 0.11183478459241353,
 "max_error": 0.08448153734207153,
 "sparsity": 0.3898779551188151,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.self_attn.out_proj": {
 "mse": 2.974707740577287e-06,
 "rmse": 0.001724734107211105,
 "relative_error": 0.12364482519876267,
 "max_error": 0.10401153564453125,
 "sparsity": 0.3794857660929362,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.encoder_attn.k_proj": {
 "mse": 5.051839252701029e-06,
 "rmse": 0.0022476296965249923,
 "relative_error": 0.1462108008993025,
 "max_error": 0.06560587882995605,
 "sparsity": 0.38422679901123047,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.encoder_attn.v_proj": {
 "mse": 4.101235390407965e-06,
 "rmse": 0.00202515070807285,
 "relative_error": 0.15203837714653734,
 "max_error": 0.062149763107299805,
 "sparsity": 0.38334306081136066,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.encoder_attn.q_proj": {
 "mse": 2.8545221084641526e-06,
 "rmse": 0.0016895331036899373,
 "relative_error": 0.10754797639395516,
 "max_error": 0.05772113800048828,
 "sparsity": 0.39005088806152344,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.encoder_attn.out_proj": {
 "mse": 3.7151462493056897e-06,
 "rmse": 0.001927471465237732,
 "relative_error": 0.14317596856972012,
 "max_error": 0.07494163513183594,
 "sparsity": 0.37266890207926434,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.fc1": {
 "mse": 2.5293929866165854e-06,
 "rmse": 0.0015904065475898248,
 "relative_error": 0.1073629947850975,
 "max_error": 0.15852278470993042,
 "sparsity": 0.3905927340189616,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.16.fc2": {
 "mse": 6.293004844337702e-06,
 "rmse": 0.002508586224218275,
 "relative_error": 0.16603359634059295,
 "max_error": 0.10226941108703613,
 "sparsity": 0.37468798955281574,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.self_attn.k_proj": {
 "mse": 3.057536559936125e-06,
 "rmse": 0.0017485812992069098,
 "relative_error": 0.11201393995950695,
 "max_error": 0.09579706192016602,
 "sparsity": 0.3894125620524089,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.self_attn.v_proj": {
 "mse": 2.4118428427755134e-06,
 "rmse": 0.0015530108958972289,
 "relative_error": 0.11385278737880668,
 "max_error": 0.04882359504699707,
 "sparsity": 0.3921810785929362,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.self_attn.q_proj": {
 "mse": 3.1471593047172064e-06,
 "rmse": 0.0017740234791899475,
 "relative_error": 0.11315092671556473,
 "max_error": 0.08912181854248047,
 "sparsity": 0.38965892791748047,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.self_attn.out_proj": {
 "mse": 3.0684352623211453e-06,
 "rmse": 0.0017516949684009328,
 "relative_error": 0.13184404672823571,
 "max_error": 0.062131404876708984,
 "sparsity": 0.3765624364217122,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.encoder_attn.k_proj": {
 "mse": 4.313059889682336e-06,
 "rmse": 0.002076790766948451,
 "relative_error": 0.1438282830189719,
 "max_error": 0.0576934814453125,
 "sparsity": 0.38422075907389325,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.encoder_attn.v_proj": {
 "mse": 5.9649946706485935e-06,
 "rmse": 0.0024423338573275753,
 "relative_error": 0.15639097085047413,
 "max_error": 0.13858509063720703,
 "sparsity": 0.38591798146565753,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.encoder_attn.q_proj": {
 "mse": 2.7382375265005976e-06,
 "rmse": 0.001654762075496232,
 "relative_error": 0.1126409982320043,
 "max_error": 0.05305671691894531,
 "sparsity": 0.38878027598063153,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.encoder_attn.out_proj": {
 "mse": 5.209771188674495e-06,
 "rmse": 0.002282492319521469,
 "relative_error": 0.15552468907321956,
 "max_error": 0.14420032501220703,
 "sparsity": 0.3756395975748698,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.fc1": {
 "mse": 2.62208232015837e-06,
 "rmse": 0.0016192845087131444,
 "relative_error": 0.10946967739129748,
 "max_error": 0.18319380283355713,
 "sparsity": 0.39118019739786786,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.17.fc2": {
 "mse": 5.671533472195733e-06,
 "rmse": 0.0023814981570842615,
 "relative_error": 0.15605075908659866,
 "max_error": 0.10937494039535522,
 "sparsity": 0.37987271944681805,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.self_attn.k_proj": {
 "mse": 3.062216137550422e-06,
 "rmse": 0.001749918894563523,
 "relative_error": 0.11481806587196931,
 "max_error": 0.06973004341125488,
 "sparsity": 0.3889427185058594,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.self_attn.v_proj": {
 "mse": 2.6578031793178525e-06,
 "rmse": 0.0016302770253296991,
 "relative_error": 0.11580165808777118,
 "max_error": 0.07330894470214844,
 "sparsity": 0.3923803965250651,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.self_attn.q_proj": {
 "mse": 3.1825525184103753e-06,
 "rmse": 0.0017839709970765712,
 "relative_error": 0.11659504798291477,
 "max_error": 0.11487168073654175,
 "sparsity": 0.38934771219889325,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.self_attn.out_proj": {
 "mse": 3.1932124784361804e-06,
 "rmse": 0.0017869562049575194,
 "relative_error": 0.12925187774123928,
 "max_error": 0.061557769775390625,
 "sparsity": 0.3762219746907552,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.encoder_attn.k_proj": {
 "mse": 3.7404693102871533e-06,
 "rmse": 0.001934029294061275,
 "relative_error": 0.14225427023342369,
 "max_error": 0.06442117691040039,
 "sparsity": 0.3841543197631836,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.encoder_attn.v_proj": {
 "mse": 5.93807635596022e-06,
 "rmse": 0.002436816849080008,
 "relative_error": 0.15289312638788924,
 "max_error": 0.09826511144638062,
 "sparsity": 0.3840119043986003,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.encoder_attn.q_proj": {
 "mse": 2.600288553367136e-06,
 "rmse": 0.0016125410237780421,
 "relative_error": 0.11576435007585356,
 "max_error": 0.04055023193359375,
 "sparsity": 0.3882484436035156,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.encoder_attn.out_proj": {
 "mse": 6.008140189805999e-06,
 "rmse": 0.0024511507888757065,
 "relative_error": 0.16401463535585767,
 "max_error": 0.19881439208984375,
 "sparsity": 0.3687610626220703,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.fc1": {
 "mse": 2.5828803700278513e-06,
 "rmse": 0.0016071342103346103,
 "relative_error": 0.10860366900771316,
 "max_error": 0.18947696685791016,
 "sparsity": 0.39104294776916504,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.18.fc2": {
 "mse": 7.289684617717285e-06,
 "rmse": 0.002699941595241883,
 "relative_error": 0.17422221641452565,
 "max_error": 0.1270887851715088,
 "sparsity": 0.37224801381429035,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.self_attn.k_proj": {
 "mse": 3.049122142328997e-06,
 "rmse": 0.0017461735716500228,
 "relative_error": 0.11567203769595212,
 "max_error": 0.0702664852142334,
 "sparsity": 0.38868459065755206,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.self_attn.v_proj": {
 "mse": 2.7155556381330825e-06,
 "rmse": 0.0016478943042965719,
 "relative_error": 0.1192114629832996,
 "max_error": 0.07558798789978027,
 "sparsity": 0.39115556081136066,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.self_attn.q_proj": {
 "mse": 3.1372007924801437e-06,
 "rmse": 0.001771214496462849,
 "relative_error": 0.11738424348854617,
 "max_error": 0.06286239624023438,
 "sparsity": 0.38935597737630206,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.self_attn.out_proj": {
 "mse": 3.676388359963312e-06,
 "rmse": 0.0019173910294885891,
 "relative_error": 0.14067619140069135,
 "max_error": 0.07267498970031738,
 "sparsity": 0.3723777135213216,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.encoder_attn.k_proj": {
 "mse": 4.325136615079828e-06,
 "rmse": 0.002079696279527332,
 "relative_error": 0.1461241549187539,
 "max_error": 0.10210031270980835,
 "sparsity": 0.38515504201253253,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.encoder_attn.v_proj": {
 "mse": 7.3299588621011935e-06,
 "rmse": 0.0027073896768107088,
 "relative_error": 0.15419761759787312,
 "max_error": 0.11008405685424805,
 "sparsity": 0.38483206431070965,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.encoder_attn.q_proj": {
 "mse": 3.1062227208167315e-06,
 "rmse": 0.0017624479342144355,
 "relative_error": 0.12209272833975149,
 "max_error": 0.04779624938964844,
 "sparsity": 0.38499895731608075,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.encoder_attn.out_proj": {
 "mse": 8.42092777020298e-06,
 "rmse": 0.0029018834866691287,
 "relative_error": 0.1775200950024603,
 "max_error": 0.21614456176757812,
 "sparsity": 0.3764657974243164,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.fc1": {
 "mse": 2.607575652291416e-06,
 "rmse": 0.0016147989510435707,
 "relative_error": 0.10923039530021018,
 "max_error": 0.14707237482070923,
 "sparsity": 0.3912237485249837,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.19.fc2": {
 "mse": 8.792547305347398e-06,
 "rmse": 0.0029652229773403883,
 "relative_error": 0.18922121151644503,
 "max_error": 0.125,
 "sparsity": 0.3674290180206299,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.self_attn.k_proj": {
 "mse": 2.585725269454997e-06,
 "rmse": 0.0016080190513345907,
 "relative_error": 0.1106971274399791,
 "max_error": 0.04056340456008911,
 "sparsity": 0.38825352986653644,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.self_attn.v_proj": {
 "mse": 2.6116106255358318e-06,
 "rmse": 0.001616047841351187,
 "relative_error": 0.1165965263213501,
 "max_error": 0.0487060546875,
 "sparsity": 0.39172935485839844,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.self_attn.q_proj": {
 "mse": 2.7351811695552897e-06,
 "rmse": 0.0016538383142119092,
 "relative_error": 0.11324436363197928,
 "max_error": 0.05264711380004883,
 "sparsity": 0.3888041178385417,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.self_attn.out_proj": {
 "mse": 3.686214313347591e-06,
 "rmse": 0.0019199516434919893,
 "relative_error": 0.14039298497893754,
 "max_error": 0.08768653869628906,
 "sparsity": 0.37349573771158856,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.encoder_attn.k_proj": {
 "mse": 4.4726807573169935e-06,
 "rmse": 0.0021148713335134585,
 "relative_error": 0.14211945336391,
 "max_error": 0.08284521102905273,
 "sparsity": 0.38464196523030597,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.encoder_attn.v_proj": {
 "mse": 7.222415661090054e-06,
 "rmse": 0.0026874552388998136,
 "relative_error": 0.15828476986233206,
 "max_error": 0.11456680297851562,
 "sparsity": 0.38407135009765625,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.encoder_attn.q_proj": {
 "mse": 3.376102540642023e-06,
 "rmse": 0.001837417356139324,
 "relative_error": 0.12240313810067552,
 "max_error": 0.07167434692382812,
 "sparsity": 0.3857393264770508,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.encoder_attn.out_proj": {
 "mse": 5.347591013560304e-06,
 "rmse": 0.002312485894780832,
 "relative_error": 0.14779052883935034,
 "max_error": 0.14819598197937012,
 "sparsity": 0.370513916015625,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.fc1": {
 "mse": 2.671948550414527e-06,
 "rmse": 0.0016346096018360246,
 "relative_error": 0.11068499087775575,
 "max_error": 0.20884323120117188,
 "sparsity": 0.39017558097839355,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.20.fc2": {
 "mse": 8.031453035073355e-06,
 "rmse": 0.002833981833934959,
 "relative_error": 0.17959298852793198,
 "max_error": 0.14152908325195312,
 "sparsity": 0.37109629313151044,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.self_attn.k_proj": {
 "mse": 2.8241770451131742e-06,
 "rmse": 0.001680528799251347,
 "relative_error": 0.11666074817263411,
 "max_error": 0.08865010738372803,
 "sparsity": 0.38890234629313153,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.self_attn.v_proj": {
 "mse": 2.7611768018687144e-06,
 "rmse": 0.0016616789105807158,
 "relative_error": 0.11628176687983355,
 "max_error": 0.06029319763183594,
 "sparsity": 0.39069143931070965,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.self_attn.q_proj": {
 "mse": 2.902990217990009e-06,
 "rmse": 0.0017038163686236875,
 "relative_error": 0.11756411493004902,
 "max_error": 0.09850597381591797,
 "sparsity": 0.38947486877441406,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.self_attn.out_proj": {
 "mse": 3.309012527097366e-06,
 "rmse": 0.00181906913752539,
 "relative_error": 0.12909355415486923,
 "max_error": 0.10249519348144531,
 "sparsity": 0.3772468566894531,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.encoder_attn.k_proj": {
 "mse": 4.397637894726358e-06,
 "rmse": 0.0020970545760009103,
 "relative_error": 0.14199766950719964,
 "max_error": 0.056479454040527344,
 "sparsity": 0.3843771616617839,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.encoder_attn.v_proj": {
 "mse": 9.549195965519175e-06,
 "rmse": 0.0030901773356102357,
 "relative_error": 0.15587564523772446,
 "max_error": 0.1548752784729004,
 "sparsity": 0.38514550526936847,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.encoder_attn.q_proj": {
 "mse": 3.236666543671163e-06,
 "rmse": 0.0017990738016188115,
 "relative_error": 0.12158806361064572,
 "max_error": 0.044310808181762695,
 "sparsity": 0.3864482243855794,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.encoder_attn.out_proj": {
 "mse": 1.2453544513846282e-05,
 "rmse": 0.0035289579926440443,
 "relative_error": 0.19398958543544084,
 "max_error": 0.2276768684387207,
 "sparsity": 0.37346363067626953,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.fc1": {
 "mse": 2.6899620024778415e-06,
 "rmse": 0.0016401103628956928,
 "relative_error": 0.11088570231448468,
 "max_error": 0.19092488288879395,
 "sparsity": 0.39104406038920086,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.21.fc2": {
 "mse": 1.1716691915353294e-05,
 "rmse": 0.0034229653687049325,
 "relative_error": 0.21299683368336433,
 "max_error": 0.18568956851959229,
 "sparsity": 0.35869065920511883,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.self_attn.k_proj": {
 "mse": 2.4202918211813085e-06,
 "rmse": 0.0015557287106630477,
 "relative_error": 0.11283075734818315,
 "max_error": 0.08211708068847656,
 "sparsity": 0.3905862172444661,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.self_attn.v_proj": {
 "mse": 2.446655571475276e-06,
 "rmse": 0.0015641788809069364,
 "relative_error": 0.1136433297025706,
 "max_error": 0.04386502504348755,
 "sparsity": 0.3926156361897786,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.self_attn.q_proj": {
 "mse": 2.585672973509645e-06,
 "rmse": 0.001608002790267991,
 "relative_error": 0.11486764879772378,
 "max_error": 0.08422493934631348,
 "sparsity": 0.3920370737711589,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.self_attn.out_proj": {
 "mse": 3.3158557926071808e-06,
 "rmse": 0.0018209491460793683,
 "relative_error": 0.13301228810832066,
 "max_error": 0.07436180114746094,
 "sparsity": 0.37401390075683594,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.encoder_attn.k_proj": {
 "mse": 4.3133932194905356e-06,
 "rmse": 0.0020768710165753037,
 "relative_error": 0.14067432703895769,
 "max_error": 0.05864429473876953,
 "sparsity": 0.38448651631673175,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.encoder_attn.v_proj": {
 "mse": 8.616913873993326e-06,
 "rmse": 0.00293545803478662,
 "relative_error": 0.15301533169596557,
 "max_error": 0.17897844314575195,
 "sparsity": 0.3843116760253906,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.encoder_attn.q_proj": {
 "mse": 2.9848199574189493e-06,
 "rmse": 0.0017276631492912468,
 "relative_error": 0.11710600067795825,
 "max_error": 0.0589747428894043,
 "sparsity": 0.3881543477376302,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.encoder_attn.out_proj": {
 "mse": 8.628910109109711e-06,
 "rmse": 0.0029375006568696646,
 "relative_error": 0.17018883610312907,
 "max_error": 0.1784210205078125,
 "sparsity": 0.37581761678059894,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.fc1": {
 "mse": 2.6359286948718363e-06,
 "rmse": 0.0016235543399812143,
 "relative_error": 0.10956246786200693,
 "max_error": 0.15522384643554688,
 "sparsity": 0.3922089735666911,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.22.fc2": {
 "mse": 1.6377167412429117e-05,
 "rmse": 0.004046871311572573,
 "relative_error": 0.2533474753294872,
 "max_error": 0.22931241989135742,
 "sparsity": 0.334999958674113,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.self_attn.k_proj": {
 "mse": 2.916458697654889e-06,
 "rmse": 0.0017077642394823967,
 "relative_error": 0.11842748635264559,
 "max_error": 0.06891632080078125,
 "sparsity": 0.3846117655436198,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.self_attn.v_proj": {
 "mse": 2.9219972930150107e-06,
 "rmse": 0.001709385062826691,
 "relative_error": 0.12511812783253307,
 "max_error": 0.04297161102294922,
 "sparsity": 0.3864247004191081,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.self_attn.q_proj": {
 "mse": 2.794949296003324e-06,
 "rmse": 0.0016718101853988462,
 "relative_error": 0.11630976302364118,
 "max_error": 0.05386066436767578,
 "sparsity": 0.3846616744995117,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.self_attn.out_proj": {
 "mse": 2.7796131689683534e-06,
 "rmse": 0.0016672171931000333,
 "relative_error": 0.12332751993743979,
 "max_error": 0.09098243713378906,
 "sparsity": 0.3754889170328776,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.encoder_attn.k_proj": {
 "mse": 4.40665189671563e-06,
 "rmse": 0.0020992026811900823,
 "relative_error": 0.14626904581808187,
 "max_error": 0.09139752388000488,
 "sparsity": 0.38378461201985675,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.encoder_attn.v_proj": {
 "mse": 5.509216862265021e-06,
 "rmse": 0.0023471720989874222,
 "relative_error": 0.14926709862472565,
 "max_error": 0.11292362213134766,
 "sparsity": 0.3838071823120117,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.encoder_attn.q_proj": {
 "mse": 3.3597671063034795e-06,
 "rmse": 0.0018329667499175973,
 "relative_error": 0.12338244477979932,
 "max_error": 0.04654979705810547,
 "sparsity": 0.38701915740966797,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.encoder_attn.out_proj": {
 "mse": 3.527855142237968e-06,
 "rmse": 0.0018782585397750671,
 "relative_error": 0.1269762413270245,
 "max_error": 0.10678958892822266,
 "sparsity": 0.37538401285807294,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.fc1": {
 "mse": 3.3838575745903654e-06,
 "rmse": 0.0018395264538979496,
 "relative_error": 0.11950819790999201,
 "max_error": 0.0970156192779541,
 "sparsity": 0.38993239402770996,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 },
 "model.decoder.layers.23.fc2": {
 "mse": 1.7770662452676333e-05,
 "rmse": 0.00421552635535307,
 "relative_error": 0.2633558020199924,
 "max_error": 0.3130340576171875,
 "sparsity": 0.33670878410339355,
 "effective_bits": 9.0,
 "sparse_nnz": 0,
 "n_planes": 3,
 "rescued_rows": 0
 }
 },
 "summary": {
 "method_name": "Broad-tritplane3",
 "model_family": "seq2seq_lm",
 "selected_components": [
 "decoder"
 ],
 "quantized_modules": 240,
 "quantized_params": 402653184,
 "quantized_fraction": 0.5271309931564236,
 "avg_relative_error": 0.13568142455431076,
 "avg_effective_bits": 9.0,
 "full_model_effective_bits": 12.310083047905035,
 "compression_ratio": 1.299747527107295
 },
 "method_name": "Broad-tritplane3",
 "model_family": "seq2seq_lm",
 "format_family": "tritplane_small",
 "format_version": "1.0",
 "total_packed_bytes": 452984832,
 "total_fp16_bytes": 805306368,
 "compression_ratio": 1.7777777777777777
}