michal-sokolski-tcl
commited on
Commit
•
d829efc
1
Parent(s):
1b9fd7c
Upload 2 files
Browse files- config.dict.json +1 -0
- modules.pt +3 -0
config.dict.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"0": {"name": "x", "type": "input"}, "1": {"name": "blocks_0_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 0}, "2": {"name": "blocks_0_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 1}, "3": {"name": "blocks_1_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 2}, "4": {"name": "blocks_1_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 3}, "5": {"name": "blocks_2_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 4}, "6": {"name": "blocks_2_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 5}, "7": {"name": "blocks_3_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 6}, "8": {"name": "blocks_3_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 7}, "9": {"name": "blocks_4_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 8}, "10": {"name": "blocks_4_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 9}, "11": {"name": "blocks_5_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 10}, "12": {"name": "blocks_5_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 11}, "13": {"name": "blocks_6_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 12}, "14": {"name": "blocks_6_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 13}, "15": {"name": "blocks_7_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 14}, "16": {"name": "blocks_7_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 15}, "17": {"name": "blocks_8_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 16}, "18": {"name": "blocks_8_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 17}, "19": {"name": "blocks_9_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 18}, "20": {"name": "blocks_9_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 19}, "21": {"name": "blocks_10_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 20}, "22": {"name": "blocks_10_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 21}, "23": {"name": "blocks_11_gamma_1", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 22}, "24": {"name": "blocks_11_gamma_2", "predecessor_indices": [], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 23}, "25": {"name": "patch_embed", "predecessor_indices": [0], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 24}, "26": {"name": "cls_token", "predecessor_indices": [25], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 25}, "27": {"name": "pos_drop", "predecessor_indices": [26], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 26}, "28": {"name": "blocks_0_norm1", "predecessor_indices": [27], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 27}, "29": {"name": "blocks_0_attn", "predecessor_indices": [28], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 28}, "30": {"name": "mul", "predecessor_indices": [1, 29], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 29}, "31": {"name": "add", "predecessor_indices": [27, 30], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 30}, "32": {"name": "blocks_0_norm2", "predecessor_indices": [31], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 31}, "33": {"name": "blocks_0_mlp_fc1", "predecessor_indices": [32], "is_atomic": true, "type": "inner", "orbit": "orbit_0", "module_index": 32}, "34": {"name": "blocks_0_mlp_act", "predecessor_indices": [33], "is_atomic": true, "type": "inner", "orbit": "orbit_0", "module_index": 33}, "35": {"name": "blocks_0_mlp_drop1", "predecessor_indices": [34], "is_atomic": true, "type": "inner", "orbit": "orbit_0", "module_index": 34}, "36": {"name": "orbit_0_mask_0", "predecessor_indices": [35], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 35}, "37": {"name": "blocks_0_mlp_fc2", "predecessor_indices": [36], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 36}, "38": {"name": "blocks_0_mlp_drop2", "predecessor_indices": [37], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 37}, "39": {"name": "mul_1", "predecessor_indices": [2, 38], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 38}, "40": {"name": "add_1", "predecessor_indices": [31, 39], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 39}, "41": {"name": "blocks_1_norm1", "predecessor_indices": [40], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 40}, "42": {"name": "blocks_1_attn", "predecessor_indices": [41], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 41}, "43": {"name": "mul_2", "predecessor_indices": [3, 42], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 42}, "44": {"name": "add_2", "predecessor_indices": [40, 43], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 43}, "45": {"name": "blocks_1_norm2", "predecessor_indices": [44], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 44}, "46": {"name": "blocks_1_mlp_fc1", "predecessor_indices": [45], "is_atomic": true, "type": "inner", "orbit": "orbit_1", "module_index": 45}, "47": {"name": "blocks_1_mlp_act", "predecessor_indices": [46], "is_atomic": true, "type": "inner", "orbit": "orbit_1", "module_index": 46}, "48": {"name": "blocks_1_mlp_drop1", "predecessor_indices": [47], "is_atomic": true, "type": "inner", "orbit": "orbit_1", "module_index": 47}, "49": {"name": "orbit_1_mask_0", "predecessor_indices": [48], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 48}, "50": {"name": "blocks_1_mlp_fc2", "predecessor_indices": [49], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 49}, "51": {"name": "blocks_1_mlp_drop2", "predecessor_indices": [50], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 50}, "52": {"name": "mul_3", "predecessor_indices": [4, 51], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 51}, "53": {"name": "add_3", "predecessor_indices": [44, 52], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 52}, "54": {"name": "blocks_2_norm1", "predecessor_indices": [53], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 53}, "55": {"name": "blocks_2_attn", "predecessor_indices": [54], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 54}, "56": {"name": "mul_4", "predecessor_indices": [5, 55], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 55}, "57": {"name": "add_4", "predecessor_indices": [53, 56], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 56}, "58": {"name": "blocks_2_norm2", "predecessor_indices": [57], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 57}, "59": {"name": "blocks_2_mlp_fc1", "predecessor_indices": [58], "is_atomic": true, "type": "inner", "orbit": "orbit_2", "module_index": 58}, "60": {"name": "blocks_2_mlp_act", "predecessor_indices": [59], "is_atomic": true, "type": "inner", "orbit": "orbit_2", "module_index": 59}, "61": {"name": "blocks_2_mlp_drop1", "predecessor_indices": [60], "is_atomic": true, "type": "inner", "orbit": "orbit_2", "module_index": 60}, "62": {"name": "orbit_2_mask_0", "predecessor_indices": [61], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 61}, "63": {"name": "blocks_2_mlp_fc2", "predecessor_indices": [62], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 62}, "64": {"name": "blocks_2_mlp_drop2", "predecessor_indices": [63], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 63}, "65": {"name": "mul_5", "predecessor_indices": [6, 64], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 64}, "66": {"name": "add_5", "predecessor_indices": [57, 65], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 65}, "67": {"name": "blocks_3_norm1", "predecessor_indices": [66], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 66}, "68": {"name": "blocks_3_attn", "predecessor_indices": [67], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 67}, "69": {"name": "mul_6", "predecessor_indices": [7, 68], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 68}, "70": {"name": "add_6", "predecessor_indices": [66, 69], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 69}, "71": {"name": "blocks_3_norm2", "predecessor_indices": [70], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 70}, "72": {"name": "blocks_3_mlp_fc1", "predecessor_indices": [71], "is_atomic": true, "type": "inner", "orbit": "orbit_3", "module_index": 71}, "73": {"name": "blocks_3_mlp_act", "predecessor_indices": [72], "is_atomic": true, "type": "inner", "orbit": "orbit_3", "module_index": 72}, "74": {"name": "blocks_3_mlp_drop1", "predecessor_indices": [73], "is_atomic": true, "type": "inner", "orbit": "orbit_3", "module_index": 73}, "75": {"name": "orbit_3_mask_0", "predecessor_indices": [74], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 74}, "76": {"name": "blocks_3_mlp_fc2", "predecessor_indices": [75], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 75}, "77": {"name": "blocks_3_mlp_drop2", "predecessor_indices": [76], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 76}, "78": {"name": "mul_7", "predecessor_indices": [8, 77], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 77}, "79": {"name": "add_7", "predecessor_indices": [70, 78], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 78}, "80": {"name": "blocks_4_norm1", "predecessor_indices": [79], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 79}, "81": {"name": "blocks_4_attn", "predecessor_indices": [80], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 80}, "82": {"name": "mul_8", "predecessor_indices": [9, 81], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 81}, "83": {"name": "add_8", "predecessor_indices": [79, 82], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 82}, "84": {"name": "blocks_4_norm2", "predecessor_indices": [83], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 83}, "85": {"name": "blocks_4_mlp_fc1", "predecessor_indices": [84], "is_atomic": true, "type": "inner", "orbit": "orbit_4", "module_index": 84}, "86": {"name": "blocks_4_mlp_act", "predecessor_indices": [85], "is_atomic": true, "type": "inner", "orbit": "orbit_4", "module_index": 85}, "87": {"name": "blocks_4_mlp_drop1", "predecessor_indices": [86], "is_atomic": true, "type": "inner", "orbit": "orbit_4", "module_index": 86}, "88": {"name": "orbit_4_mask_0", "predecessor_indices": [87], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 87}, "89": {"name": "blocks_4_mlp_fc2", "predecessor_indices": [88], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 88}, "90": {"name": "blocks_4_mlp_drop2", "predecessor_indices": [89], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 89}, "91": {"name": "mul_9", "predecessor_indices": [10, 90], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 90}, "92": {"name": "add_9", "predecessor_indices": [83, 91], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 91}, "93": {"name": "blocks_5_norm1", "predecessor_indices": [92], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 92}, "94": {"name": "blocks_5_attn", "predecessor_indices": [93], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 93}, "95": {"name": "mul_10", "predecessor_indices": [11, 94], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 94}, "96": {"name": "add_10", "predecessor_indices": [92, 95], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 95}, "97": {"name": "blocks_5_norm2", "predecessor_indices": [96], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 96}, "98": {"name": "blocks_5_mlp_fc1", "predecessor_indices": [97], "is_atomic": true, "type": "inner", "orbit": "orbit_5", "module_index": 97}, "99": {"name": "blocks_5_mlp_act", "predecessor_indices": [98], "is_atomic": true, "type": "inner", "orbit": "orbit_5", "module_index": 98}, "100": {"name": "blocks_5_mlp_drop1", "predecessor_indices": [99], "is_atomic": true, "type": "inner", "orbit": "orbit_5", "module_index": 99}, "101": {"name": "orbit_5_mask_0", "predecessor_indices": [100], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 100}, "102": {"name": "blocks_5_mlp_fc2", "predecessor_indices": [101], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 101}, "103": {"name": "blocks_5_mlp_drop2", "predecessor_indices": [102], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 102}, "104": {"name": "mul_11", "predecessor_indices": [12, 103], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 103}, "105": {"name": "add_11", "predecessor_indices": [96, 104], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 104}, "106": {"name": "blocks_6_norm1", "predecessor_indices": [105], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 105}, "107": {"name": "blocks_6_attn", "predecessor_indices": [106], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 106}, "108": {"name": "mul_12", "predecessor_indices": [13, 107], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 107}, "109": {"name": "add_12", "predecessor_indices": [105, 108], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 108}, "110": {"name": "blocks_6_norm2", "predecessor_indices": [109], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 109}, "111": {"name": "blocks_6_mlp_fc1", "predecessor_indices": [110], "is_atomic": true, "type": "inner", "orbit": "orbit_6", "module_index": 110}, "112": {"name": "blocks_6_mlp_act", "predecessor_indices": [111], "is_atomic": true, "type": "inner", "orbit": "orbit_6", "module_index": 111}, "113": {"name": "blocks_6_mlp_drop1", "predecessor_indices": [112], "is_atomic": true, "type": "inner", "orbit": "orbit_6", "module_index": 112}, "114": {"name": "orbit_6_mask_0", "predecessor_indices": [113], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 113}, "115": {"name": "blocks_6_mlp_fc2", "predecessor_indices": [114], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 114}, "116": {"name": "blocks_6_mlp_drop2", "predecessor_indices": [115], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 115}, "117": {"name": "mul_13", "predecessor_indices": [14, 116], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 116}, "118": {"name": "add_13", "predecessor_indices": [109, 117], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 117}, "119": {"name": "blocks_7_norm1", "predecessor_indices": [118], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 118}, "120": {"name": "blocks_7_attn", "predecessor_indices": [119], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 119}, "121": {"name": "mul_14", "predecessor_indices": [15, 120], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 120}, "122": {"name": "add_14", "predecessor_indices": [118, 121], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 121}, "123": {"name": "blocks_7_norm2", "predecessor_indices": [122], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 122}, "124": {"name": "blocks_7_mlp_fc1", "predecessor_indices": [123], "is_atomic": true, "type": "inner", "orbit": "orbit_7", "module_index": 123}, "125": {"name": "blocks_7_mlp_act", "predecessor_indices": [124], "is_atomic": true, "type": "inner", "orbit": "orbit_7", "module_index": 124}, "126": {"name": "blocks_7_mlp_drop1", "predecessor_indices": [125], "is_atomic": true, "type": "inner", "orbit": "orbit_7", "module_index": 125}, "127": {"name": "orbit_7_mask_0", "predecessor_indices": [126], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 126}, "128": {"name": "blocks_7_mlp_fc2", "predecessor_indices": [127], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 127}, "129": {"name": "blocks_7_mlp_drop2", "predecessor_indices": [128], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 128}, "130": {"name": "mul_15", "predecessor_indices": [16, 129], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 129}, "131": {"name": "add_15", "predecessor_indices": [122, 130], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 130}, "132": {"name": "blocks_8_norm1", "predecessor_indices": [131], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 131}, "133": {"name": "blocks_8_attn", "predecessor_indices": [132], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 132}, "134": {"name": "mul_16", "predecessor_indices": [17, 133], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 133}, "135": {"name": "add_16", "predecessor_indices": [131, 134], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 134}, "136": {"name": "blocks_8_norm2", "predecessor_indices": [135], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 135}, "137": {"name": "blocks_8_mlp_fc1", "predecessor_indices": [136], "is_atomic": true, "type": "inner", "orbit": "orbit_8", "module_index": 136}, "138": {"name": "blocks_8_mlp_act", "predecessor_indices": [137], "is_atomic": true, "type": "inner", "orbit": "orbit_8", "module_index": 137}, "139": {"name": "blocks_8_mlp_drop1", "predecessor_indices": [138], "is_atomic": true, "type": "inner", "orbit": "orbit_8", "module_index": 138}, "140": {"name": "orbit_8_mask_0", "predecessor_indices": [139], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 139}, "141": {"name": "blocks_8_mlp_fc2", "predecessor_indices": [140], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 140}, "142": {"name": "blocks_8_mlp_drop2", "predecessor_indices": [141], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 141}, "143": {"name": "mul_17", "predecessor_indices": [18, 142], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 142}, "144": {"name": "add_17", "predecessor_indices": [135, 143], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 143}, "145": {"name": "blocks_9_norm1", "predecessor_indices": [144], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 144}, "146": {"name": "blocks_9_attn", "predecessor_indices": [145], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 145}, "147": {"name": "mul_18", "predecessor_indices": [19, 146], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 146}, "148": {"name": "add_18", "predecessor_indices": [144, 147], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 147}, "149": {"name": "blocks_9_norm2", "predecessor_indices": [148], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 148}, "150": {"name": "blocks_9_mlp_fc1", "predecessor_indices": [149], "is_atomic": true, "type": "inner", "orbit": "orbit_9", "module_index": 149}, "151": {"name": "blocks_9_mlp_act", "predecessor_indices": [150], "is_atomic": true, "type": "inner", "orbit": "orbit_9", "module_index": 150}, "152": {"name": "blocks_9_mlp_drop1", "predecessor_indices": [151], "is_atomic": true, "type": "inner", "orbit": "orbit_9", "module_index": 151}, "153": {"name": "orbit_9_mask_0", "predecessor_indices": [152], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 152}, "154": {"name": "blocks_9_mlp_fc2", "predecessor_indices": [153], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 153}, "155": {"name": "blocks_9_mlp_drop2", "predecessor_indices": [154], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 154}, "156": {"name": "mul_19", "predecessor_indices": [20, 155], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 155}, "157": {"name": "add_19", "predecessor_indices": [148, 156], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 156}, "158": {"name": "blocks_10_norm1", "predecessor_indices": [157], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 157}, "159": {"name": "blocks_10_attn", "predecessor_indices": [158], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 158}, "160": {"name": "mul_20", "predecessor_indices": [21, 159], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 159}, "161": {"name": "add_20", "predecessor_indices": [157, 160], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 160}, "162": {"name": "blocks_10_norm2", "predecessor_indices": [161], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 161}, "163": {"name": "blocks_10_mlp_fc1", "predecessor_indices": [162], "is_atomic": true, "type": "inner", "orbit": "orbit_10", "module_index": 162}, "164": {"name": "blocks_10_mlp_act", "predecessor_indices": [163], "is_atomic": true, "type": "inner", "orbit": "orbit_10", "module_index": 163}, "165": {"name": "blocks_10_mlp_drop1", "predecessor_indices": [164], "is_atomic": true, "type": "inner", "orbit": "orbit_10", "module_index": 164}, "166": {"name": "orbit_10_mask_0", "predecessor_indices": [165], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 165}, "167": {"name": "blocks_10_mlp_fc2", "predecessor_indices": [166], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 166}, "168": {"name": "blocks_10_mlp_drop2", "predecessor_indices": [167], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 167}, "169": {"name": "mul_21", "predecessor_indices": [22, 168], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 168}, "170": {"name": "add_21", "predecessor_indices": [161, 169], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 169}, "171": {"name": "blocks_11_norm1", "predecessor_indices": [170], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 170}, "172": {"name": "blocks_11_attn", "predecessor_indices": [171], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 171}, "173": {"name": "mul_22", "predecessor_indices": [23, 172], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 172}, "174": {"name": "add_22", "predecessor_indices": [170, 173], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 173}, "175": {"name": "blocks_11_norm2", "predecessor_indices": [174], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 174}, "176": {"name": "blocks_11_mlp_fc1", "predecessor_indices": [175], "is_atomic": true, "type": "inner", "orbit": "orbit_11", "module_index": 175}, "177": {"name": "blocks_11_mlp_act", "predecessor_indices": [176], "is_atomic": true, "type": "inner", "orbit": "orbit_11", "module_index": 176}, "178": {"name": "blocks_11_mlp_drop1", "predecessor_indices": [177], "is_atomic": true, "type": "inner", "orbit": "orbit_11", "module_index": 177}, "179": {"name": "orbit_11_mask_0", "predecessor_indices": [178], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 178}, "180": {"name": "blocks_11_mlp_fc2", "predecessor_indices": [179], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 179}, "181": {"name": "blocks_11_mlp_drop2", "predecessor_indices": [180], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 180}, "182": {"name": "mul_23", "predecessor_indices": [24, 181], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 181}, "183": {"name": "add_23", "predecessor_indices": [174, 182], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 182}, "184": {"name": "getitem", "predecessor_indices": [183], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 183}, "185": {"name": "mean", "predecessor_indices": [184], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 184}, "186": {"name": "fc_norm", "predecessor_indices": [185], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 185}, "187": {"name": "head_drop", "predecessor_indices": [186], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 186}, "188": {"name": "head", "predecessor_indices": [187], "is_atomic": true, "type": "inner", "orbit": null, "module_index": 187}, "name": "ModifiedBeit", "class": "DagModule", "output_index": 188}
|
modules.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:68969dc918130b0ea956b6e66443bb8f41863ffe8acaab70725303c05035f75f
|
3 |
+
size 230590595
|