bart-base-cnn-r2-18.7-d23-hybrid / eval /sparsity_report.json
echarlaix's picture
echarlaix HF staff
Adding model, graphs and metadata.
07fe2fe
raw
history blame
1.8 kB
{"total": 139420416, "nnz": 63361439, "linear_total": 99090432, "linear_nnz": 23081472, "layers": {"0": {"total": 16539648, "nnz": 4338402, "linear_total": 16515072, "linear_nnz": 4321280, "linear_attention_total": 7077888, "linear_attention_nnz": 4170752, "linear_dense_total": 9437184, "linear_dense_nnz": 150528}, "1": {"total": 16539648, "nnz": 3697201, "linear_total": 16515072, "linear_nnz": 3680768, "linear_attention_total": 7077888, "linear_attention_nnz": 3261440, "linear_dense_total": 9437184, "linear_dense_nnz": 419328}, "2": {"total": 16539648, "nnz": 3582655, "linear_total": 16515072, "linear_nnz": 3566080, "linear_attention_total": 7077888, "linear_attention_nnz": 3076096, "linear_dense_total": 9437184, "linear_dense_nnz": 489984}, "3": {"total": 16539648, "nnz": 3230637, "linear_total": 16515072, "linear_nnz": 3214848, "linear_attention_total": 7077888, "linear_attention_nnz": 2703360, "linear_dense_total": 9437184, "linear_dense_nnz": 511488}, "4": {"total": 16539648, "nnz": 3432491, "linear_total": 16515072, "linear_nnz": 3416576, "linear_attention_total": 7077888, "linear_attention_nnz": 3104768, "linear_dense_total": 9437184, "linear_dense_nnz": 311808}, "5": {"total": 16539648, "nnz": 4899061, "linear_total": 16515072, "linear_nnz": 4881920, "linear_attention_total": 7077888, "linear_attention_nnz": 4456448, "linear_dense_total": 9437184, "linear_dense_nnz": 425472}}, "total_sparsity": 54.55368674269341, "linear_sparsity": 76.70665922619048, "pruned_heads": {"0": [8, 10], "1": [0, 5, 6], "2": [4, 6, 9], "3": [0, 1, 2, 4, 6, 7, 8, 9, 11], "4": [1, 2, 5, 6, 9, 10, 11], "5": [2, 3, 4, 5, 6, 10], "6": [0, 12, 13, 14], "7": [2, 11, 16, 17, 18, 19, 23], "8": [2, 14, 16, 21], "9": [3, 12, 13, 15, 16, 18, 21, 22], "10": [3, 12, 14, 16, 18, 20, 21], "11": [10]}}