Hugging Face's logo

PhilipQuirke
/
VerifiedArithmetic

Model card Files Community
PhilipQuirke commited on
Commit
92810c3
·
verified ·
1 Parent(s): 57b77e7

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. add_d10_l2_h3_t40K_gf_s572091/behaviors.json +1 -0
  2. add_d10_l2_h3_t40K_gf_s572091/features.json +1 -0
  3. add_d10_l2_h3_t40K_gf_s572091/model.pth +3 -0
  4. add_d10_l2_h3_t40K_gf_s572091/training_loss.json +0 -0
  5. add_d10_l2_h3_t40K_s572091/behaviors.json +1 -0
  6. add_d10_l2_h3_t40K_s572091/features.json +1 -0
  7. add_d10_l2_h3_t40K_s572091/model.pth +3 -0
  8. add_d10_l2_h3_t40K_s572091/training_loss.json +0 -0
  9. add_d11_l2_h3_t50K_s572091/behaviors.json +1 -0
  10. add_d11_l2_h3_t50K_s572091/features.json +1 -0
  11. add_d11_l2_h3_t50K_s572091/model.pth +3 -0
  12. add_d11_l2_h3_t50K_s572091/training_loss.json +0 -0
  13. add_d12_l2_h3_t50K_s572091/behaviors.json +1 -0
  14. add_d12_l2_h3_t50K_s572091/features.json +1 -0
  15. add_d12_l2_h3_t50K_s572091/model.pth +3 -0
  16. add_d12_l2_h3_t50K_s572091/training_loss.json +0 -0
  17. add_d13_l2_h3_t50K_s572091/behaviors.json +1 -0
  18. add_d13_l2_h3_t50K_s572091/features.json +1 -0
  19. add_d13_l2_h3_t50K_s572091/model.pth +3 -0
  20. add_d13_l2_h3_t50K_s572091/training_loss.json +0 -0
  21. add_d14_l2_h3_t60K_s572091/behaviors.json +1 -0
  22. add_d14_l2_h3_t60K_s572091/features.json +1 -0
  23. add_d14_l2_h3_t60K_s572091/model.pth +3 -0
  24. add_d14_l2_h3_t60K_s572091/training_loss.json +0 -0
  25. add_d15_l2_h3_t80K_s572091/model.pth +3 -0
  26. add_d15_l2_h3_t80K_s572091/training_loss.json +0 -0
  27. add_d20_l2_h3_t80K_s572091/model.pth +3 -0
  28. add_d20_l2_h3_t80K_s572091/training_loss.json +0 -0
  29. add_d5_l1_h3_t15K_s372001/behaviors.json +1 -0
  30. add_d5_l1_h3_t15K_s372001/features.json +1 -0
  31. add_d5_l1_h3_t15K_s372001/model.pth +3 -0
  32. add_d5_l1_h3_t15K_s372001/training_loss.json +0 -0
  33. add_d5_l1_h3_t30K_s372001/model.pth +3 -0
  34. add_d5_l1_h3_t30K_s372001/training_loss.json +0 -0
  35. add_d5_l2_h3_t15K_s372001/behaviors.json +1 -0
  36. add_d5_l2_h3_t15K_s372001/features.json +1 -0
  37. add_d5_l2_h3_t15K_s372001/model.pth +3 -0
  38. add_d5_l2_h3_t15K_s372001/training_loss.json +0 -0
  39. add_d5_l2_h3_t40K_s372001/behaviors.json +1 -0
  40. add_d5_l2_h3_t40K_s372001/features.json +1 -0
  41. add_d5_l2_h3_t40K_s372001/model.pth +3 -0
  42. add_d5_l2_h3_t40K_s372001/training_loss.json +0 -0
  43. add_d6_l2_h3_t15K_s372001/behaviors.json +1 -0
  44. add_d6_l2_h3_t15K_s372001/features.json +1 -0
  45. add_d6_l2_h3_t15K_s372001/model.pth +3 -0
  46. add_d6_l2_h3_t15K_s372001/training_loss.json +0 -0
  47. add_d6_l2_h3_t20K_s173289/behaviors.json +1 -0
  48. add_d6_l2_h3_t20K_s173289/features.json +1 -0
  49. add_d6_l2_h3_t20K_s173289/model.pth +3 -0
  50. add_d6_l2_h3_t20K_s173289/training_loss.json +0 -0
add_d10_l2_h3_t40K_gf_s572091/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A9", "Math.Add:S23", "Attn:P2=50", "Attn:P13=47"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:1", "Impact:A9", "Math.Add:S2"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A987", "Math.Add:S2345", "Attn:P3=49", "Attn:P14=44", "Math.Add:A6.SP.Weak"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:3", "Impact:A8", "Math.Add:S245"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:8", "Impact:A876", "Math.Add:S2345", "Attn:P5=47", "Attn:P16=44", "Attn:P4=1", "Attn:P15=1"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:8", "Impact:A876", "Math.Add:S2345"]}, {"position": 18, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:4", "Impact:A876", "Math.Add:S2345", "Attn:P4=47", "Attn:P15=46", "Math.Add:A5.SP.Weak"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:4", "Impact:A87", "Math.Add:S2345"]}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A7", "Math.Add:S4", "Attn:P5=53", "Attn:P16=42"]}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:10", "Impact:A8765", "Math.Add:S12345", "Attn:P6=47", "Attn:P17=45", "Math.Add:A3.SP.Weak"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:12", "Impact:A8765", "Math.Add:S2345"]}, {"position": 20, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:10", "Impact:A6543", "Math.Add:S12345", "Attn:P8=58", "Attn:P19=41", "Math.Add:A1.SP.Weak"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:8", "Impact:A5432", "Math.Add:S345", "Attn:P9=38", "Attn:P20=29", "Attn:P10=17", "Attn:P11=2"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:17", "Impact:A76543", "Math.Add:S12345", "Attn:P7=48", "Attn:P18=46", "Math.Add:A2.SP.Weak"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:24", "Impact:A76543", "Math.Add:S2345"]}, {"position": 22, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:15", "Impact:A10", "Math.Add:S123", "Attn:P11=56", "Attn:P0=43", "Math.Add:A9.SP.Weak"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A10", "Math.Add:S2", "Attn:P1=46", "Attn:P12=46", "Math.Add:A8.SP.Weak"]}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:15", "Impact:A10", "Math.Add:S123"]}, {"position": 22, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:2", "Impact:A10", "Math.Add:S123"]}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:30", "Impact:A9", "Math.Add:S123", "Attn:P11=47", "Attn:P0=44", "Attn:P23=3", "Attn:P21=1"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:20", "Impact:A9", "Math.Add:S123", "Attn:P1=45", "Attn:P12=44", "Attn:P11=1", "Math.Add:A8.SP.Weak"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:31", "Impact:A9", "Math.Add:S123"]}, {"position": 23, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A9", "Math.Add:S2", "Attn:P1=22", "Attn:P21=17", "Attn:P3=11", "Attn:P23=11", "Math.Add:A8.SP.Weak"]}, {"position": 23, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:87", "Impact:A9", "Math.Add:S012345"]}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:30", "Impact:A8", "Math.Add:S123", "Attn:P1=48", "Attn:P12=47"]}, {"position": 24, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:17", "Impact:A8", "Math.Add:S12345", "Attn:P2=49", "Attn:P13=45", "Math.Add:A7.SP.Weak"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:32", "Impact:A8", "Math.Add:S12345"]}, {"position": 24, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A8", "Math.Add:S2345", "Attn:P1=20", "Attn:P3=16", "Attn:P21=14", "Attn:P4=11", "Math.Add:A7.SP.Weak"]}, {"position": 24, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:81", "Impact:A8", "Math.Add:S012345"]}, {"position": 25, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:35", "Impact:A7", "Math.Add:S012345", "Attn:P13=48", "Attn:P2=47"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:25", "Impact:A7", "Math.Add:S12345", "Attn:P3=47", "Attn:P14=46", "Math.Add:A6.SP.Weak"]}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:37", "Impact:A7", "Math.Add:S012345"]}, {"position": 25, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:4", "Impact:A7", "Math.Add:S2345", "Attn:P21=21", "Attn:P0=18", "Attn:P1=15", "Attn:P3=9", "Math.Add:A6.SP.Weak"]}, {"position": 25, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:87", "Impact:A7", "Math.Add:S012345"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:40", "Impact:A6", "Math.Add:S012345", "Attn:P3=49", "Attn:P14=46"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:30", "Impact:A6", "Math.Add:S12345", "Attn:P4=46", "Attn:P15=44", "Math.Add:A5.SP.Weak"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:41", "Impact:A6", "Math.Add:S012345"]}, {"position": 26, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:7", "Impact:A6", "Math.Add:S2345", "Attn:P0=33", "Attn:P21=18", "Attn:P1=15", "Attn:P17=8", "Math.Add:A5.SP.Weak"]}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:54", "Impact:A6", "Math.Add:S012345"]}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:72", "Impact:A5", "Math.Add:S012345", "Attn:P15=48", "Attn:P4=47"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:28", "Impact:A5", "Math.Add:S12345", "Attn:P5=43", "Attn:P16=41", "Attn:P27=1", "Attn:P4=1"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:46", "Impact:A5", "Math.Add:S012345"]}, {"position": 27, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A5", "Math.Add:S23", "Attn:P0=28", "Attn:P21=24", "Attn:P1=13", "Attn:P20=9", "Math.Add:A4.SP.Weak"]}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:67", "Impact:A5", "Math.Add:S012345"]}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:82", "Impact:A4", "Math.Add:S012345", "Attn:P5=47", "Attn:P16=47", "Math.Add:A0.SP.Weak"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:31", "Impact:A4", "Math.Add:S12345", "Attn:P6=45", "Attn:P17=43", "Math.Add:A2.SP.Weak", "Math.Add:A3.SP.Weak"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:62", "Impact:A4", "Math.Add:S01234"]}, {"position": 28, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A4", "Math.Add:S23", "Attn:P21=27", "Attn:P20=26", "Attn:P0=19", "Attn:P1=11", "Math.Add:A2.SP.Weak", "Math.Add:A3.SP.Weak"]}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:64", "Impact:A4", "Math.Add:S01234"]}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:66", "Impact:A3", "Math.Add:S01234", "Attn:P6=48", "Attn:P17=47"]}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:26", "Impact:A3", "Math.Add:S12345", "Attn:P18=40", "Attn:P7=38", "Attn:P8=2", "Attn:P19=1", "Math.Add:A1.SP.Weak", "Math.Add:A2.SP.Weak"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:69", "Impact:A3", "Math.Add:S012345"]}, {"position": 29, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A3", "Math.Add:S5", "Attn:P28=23", "Attn:P19=22", "Attn:P29=15", "Attn:P20=14", "Math.Add:A2.SP.Weak", "Math.Add:A3.SP.Weak"]}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:58", "Impact:A3", "Math.Add:S012345"]}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:13", "Impact:A2", "Math.Add:S2345", "Attn:P9=59", "Attn:P20=40"]}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:84", "Impact:A2", "Math.Add:S012345", "Attn:P18=49", "Attn:P7=48"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:33", "Impact:A2", "Math.Add:S12345", "Attn:P8=44", "Attn:P19=39", "Attn:P7=2", "Attn:P18=2", "Math.Add:A1.SP.Weak"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:77", "Impact:A2", "Math.Add:S012345"]}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:56", "Impact:A2", "Math.Add:S012345"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:83", "Impact:A1", "Math.Add:S012345", "Attn:P19=48", "Attn:P8=48"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:29", "Impact:A1", "Math.Add:S12345", "Attn:P9=45", "Attn:P20=38", "Attn:P23=1", "Attn:P10=1"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:78", "Impact:A1", "Math.Add:S012345"]}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:65", "Impact:A1", "Math.Add:S012345"]}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:93", "Impact:A0", "Math.Add:S012345", "Attn:P20=47", "Attn:P9=45", "Attn:P10=1"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:70", "Impact:A0", "Math.Add:S012345"]}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:59", "Impact:A0", "Math.Add:S012345"]}]
add_d10_l2_h3_t40K_gf_s572091/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A7.ST"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.ST"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 18, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.ST"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.ST"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 20, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.ST"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.ST"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.ST"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 22, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A9.ST"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.ST"]}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 22, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A9.SA"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.SC"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 23, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 23, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A8.SA"]}, {"position": 24, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A7.SC"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 24, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 24, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 25, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A7.SA"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.SC"]}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 25, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 25, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A6.SA"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.SC"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.SA"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.SC"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SC"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SC"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SS"]}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.SC"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.SC"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d10_l2_h3_t40K_gf_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94d99a3c448939000611c2912c6765b3e128d779428d13f63c5374d56fb11a3e
3
+ size 25176450
add_d10_l2_h3_t40K_gf_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d10_l2_h3_t40K_s572091/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A2", "Math.Add:S2", "Attn:P9=38", "Attn:P10=34", "Attn:P1=8", "Attn:P8=5"]}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:0", "Impact:A2", "Math.Add:S2"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A98", "Math.Add:S235", "Attn:P2=49", "Attn:P13=46"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:2", "Impact:A98", "Math.Add:S25"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A8", "Math.Add:S2", "Attn:P2=49", "Attn:P13=44"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A98", "Math.Add:S2345", "Attn:P3=49", "Attn:P14=46"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:2", "Impact:A8", "Math.Add:S245"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:8", "Impact:A876", "Math.Add:S2345", "Attn:P5=47", "Attn:P16=43", "Attn:P15=1"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:8", "Impact:A876", "Math.Add:S2345"]}, {"position": 18, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:4", "Impact:A876", "Math.Add:S2345", "Attn:P15=47", "Attn:P4=47"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:3", "Impact:A87", "Math.Add:S2345"]}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A87", "Math.Add:S5", "Attn:P5=50", "Attn:P16=46"]}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:4", "Impact:A8765", "Math.Add:S1235", "Attn:P6=50", "Attn:P17=42"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:11", "Impact:A87654", "Math.Add:S12345"]}, {"position": 20, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:11", "Impact:A6543", "Math.Add:S12345", "Attn:P8=56", "Attn:P19=43", "Math.Add:A1.SP"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:10", "Impact:A543", "Math.Add:S345", "Attn:P9=38", "Attn:P20=29", "Attn:P10=19", "Attn:P11=1"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:18", "Impact:A76543", "Math.Add:S12345", "Attn:P7=49", "Attn:P18=46"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:25", "Impact:A6543", "Math.Add:S2345"]}, {"position": 22, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:15", "Impact:A10", "Math.Add:S123", "Attn:P11=56", "Attn:P0=43"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A10", "Math.Add:S2", "Attn:P12=48", "Attn:P1=48"]}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:15", "Impact:A10", "Math.Add:S123"]}, {"position": 22, "layer": 1, "is_head": true, "num": 0, "tags": ["Math.Add:A8.SP"]}, {"position": 22, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:0", "Impact:A10", "Math.Add:S1"]}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:29", "Impact:A9", "Math.Add:S123", "Attn:P11=47", "Attn:P0=44", "Attn:P23=2", "Attn:P21=2"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:19", "Impact:A9", "Math.Add:S123", "Attn:P1=50", "Attn:P12=44"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:29", "Impact:A9", "Math.Add:S123"]}, {"position": 23, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A9", "Math.Add:S2", "Attn:P0=33", "Attn:P1=21", "Attn:P21=17", "Attn:P23=6", "Math.Add:A8.SP"]}, {"position": 23, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:88", "Impact:A9", "Math.Add:S012345"]}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:28", "Impact:A8", "Math.Add:S123", "Attn:P12=49", "Attn:P1=46"]}, {"position": 24, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:16", "Impact:A8", "Math.Add:S12345", "Attn:P2=48", "Attn:P13=46", "Attn:P1=1"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:30", "Impact:A8", "Math.Add:S12345"]}, {"position": 24, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A8", "Math.Add:S2345", "Attn:P0=48", "Attn:P21=14", "Attn:P1=12", "Attn:P3=6", "Math.Add:A7.SP"]}, {"position": 24, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:80", "Impact:A8", "Math.Add:S012345"]}, {"position": 25, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:34", "Impact:A7", "Math.Add:S012345", "Attn:P2=48", "Attn:P13=47"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:24", "Impact:A7", "Math.Add:S12345", "Attn:P3=47", "Attn:P14=46"]}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:35", "Impact:A7", "Math.Add:S01235"]}, {"position": 25, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:3", "Impact:A7", "Math.Add:S2345", "Attn:P0=60", "Attn:P21=20", "Attn:P1=3", "Attn:P7=3", "Math.Add:A6.SP"]}, {"position": 25, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:85", "Impact:A7", "Math.Add:S012345"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:38", "Impact:A6", "Math.Add:S01235", "Attn:P14=48", "Attn:P3=47"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:28", "Impact:A6", "Math.Add:S12345", "Attn:P15=49", "Attn:P4=47"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:38", "Impact:A6", "Math.Add:S012345"]}, {"position": 26, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:6", "Impact:A6", "Math.Add:S235", "Attn:P0=60", "Attn:P21=18", "Attn:P17=8", "Attn:P1=4", "Math.Add:A5.SP"]}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:75", "Impact:A6", "Math.Add:S012345"]}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:46", "Impact:A5", "Math.Add:S012345", "Attn:P15=48", "Attn:P4=47"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:21", "Impact:A5", "Math.Add:S12345", "Attn:P5=44", "Attn:P16=44", "Attn:P15=1", "Attn:P4=1"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:46", "Impact:A5", "Math.Add:S012345"]}, {"position": 27, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A5", "Math.Add:S23", "Attn:P0=44", "Attn:P21=23", "Attn:P20=10", "Attn:P1=8", "Math.Add:A4.SP"]}, {"position": 27, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A5", "Math.Add:S5", "Attn:P20=23", "Attn:P21=23", "Attn:P19=19", "Attn:P17=15", "Math.Add:A4.SP"]}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:40", "Impact:A5", "Math.Add:S012345"]}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:80", "Impact:A4", "Math.Add:S012345", "Attn:P16=48", "Attn:P5=47"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:24", "Impact:A4", "Math.Add:S1235", "Attn:P6=44", "Attn:P17=41"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:60", "Impact:A4", "Math.Add:S012345"]}, {"position": 28, "layer": 1, "is_head": true, "num": 0, "tags": ["Math.Add:A3.SP"]}, {"position": 28, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A4", "Math.Add:S2", "Attn:P27=20", "Attn:P17=17", "Attn:P19=11", "Attn:P28=10"]}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:55", "Impact:A4", "Math.Add:S01234"]}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:70", "Impact:A3", "Math.Add:S012345", "Attn:P6=48", "Attn:P17=47"]}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:43", "Impact:A3", "Math.Add:S12345", "Attn:P7=44", "Attn:P18=39", "Attn:P8=1", "Attn:P19=1"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:69", "Impact:A3", "Math.Add:S012345"]}, {"position": 29, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A3", "Math.Add:S1", "Attn:P0=31", "Attn:P21=24", "Attn:P20=23", "Attn:P1=9", "Math.Add:A2.SP"]}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:57", "Impact:A3", "Math.Add:S012345"]}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:15", "Impact:A2", "Math.Add:S2345", "Attn:P9=57", "Attn:P20=42"]}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:85", "Impact:A2", "Math.Add:S012345", "Attn:P7=49", "Attn:P18=48"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:41", "Impact:A2", "Math.Add:S12345", "Attn:P8=46", "Attn:P19=40", "Attn:P18=1"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:78", "Impact:A2", "Math.Add:S012345"]}, {"position": 30, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:3", "Impact:A2", "Math.Add:S1234", "Attn:P0=36", "Attn:P21=29", "Attn:P30=11", "Attn:P1=8", "Math.Add:A1.SP"]}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:55", "Impact:A2", "Math.Add:S012345"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:86", "Impact:A1", "Math.Add:S012345", "Attn:P8=47", "Attn:P19=46"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:31", "Impact:A1", "Math.Add:S12345", "Attn:P9=47", "Attn:P20=42"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:73", "Impact:A1", "Math.Add:S012345"]}, {"position": 31, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S1", "Attn:P0=47", "Attn:P21=31", "Attn:P1=10", "Attn:P3=3"]}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:62", "Impact:A1", "Math.Add:S012345"]}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:95", "Impact:A0", "Math.Add:S012345", "Attn:P20=46", "Attn:P9=44", "Attn:P10=2", "Attn:P22=1"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:65", "Impact:A0", "Math.Add:S012345"]}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:65", "Impact:A0", "Math.Add:S012345"]}]
add_d10_l2_h3_t40K_s572091/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A7.ST"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.ST"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 18, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.ST"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.ST"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 20, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.ST"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.ST"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.ST"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 22, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A9.ST"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.ST"]}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 22, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 22, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A9.SA"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.SC"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 23, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 23, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A8.SA"]}, {"position": 24, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A7.SC"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 24, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 24, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 25, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A7.SA"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.SC"]}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 25, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 25, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A6.SA"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.SC"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.SA"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.SC"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 27, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SC"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 28, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SC"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SS"]}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.SC"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.SC"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d10_l2_h3_t40K_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5c828b74f9b78ace014210a30c97e92d0c442b114317ee8d9182d2295706be6
3
+ size 25176450
add_d10_l2_h3_t40K_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d11_l2_h3_t50K_s572091/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 0, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S123", "Attn:P0=100"]}, {"position": 0, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A1", "Math.Add:S123", "Attn:P0=100"]}, {"position": 0, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:2", "Impact:A1", "Math.Add:S123"]}, {"position": 17, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A9", "Math.Add:S2", "Attn:P15=47", "Attn:P3=46", "Math.Add:A7.SP"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A910", "Math.Add:S23", "Attn:P2=52", "Attn:P14=46", "Math.Add:A8.SP"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:3", "Impact:A910", "Math.Add:S23"]}, {"position": 20, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:7", "Impact:A87654", "Math.Add:S2345", "Attn:P6=48", "Attn:P18=46", "Math.Add:A4.SP"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:5", "Impact:A8765", "Math.Add:S12345", "Attn:P7=53", "Attn:P19=45", "Math.Add:A2.SP", "Math.Add:A3.SP"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A8765", "Math.Add:S345", "Attn:P11=28", "Attn:P7=24", "Attn:P19=24", "Attn:P6=4"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:18", "Impact:A8765", "Math.Add:S2345"]}, {"position": 21, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:8", "Impact:A6543", "Math.Add:S2345", "Attn:P9=51", "Attn:P21=42", "Math.Add:A1.SP"]}, {"position": 21, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:9", "Impact:A76543", "Math.Add:S12345", "Attn:P8=50", "Attn:P20=48", "Math.Add:A2.SP"]}, {"position": 21, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:18", "Impact:A76543", "Math.Add:S2345"]}, {"position": 22, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:6", "Impact:A87", "Math.Add:S2345", "Attn:P5=48", "Attn:P17=44", "Attn:P16=1", "Math.Add:A5.SP"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:8", "Impact:A87", "Math.Add:S2345", "Attn:P4=52", "Attn:P16=46", "Math.Add:A6.SP"]}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:8", "Impact:A87", "Math.Add:S2345"]}, {"position": 23, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:10", "Impact:A54321", "Math.Add:S2345", "Attn:P10=53", "Attn:P22=43"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:7", "Impact:A42", "Math.Add:S2345", "Attn:P9=46", "Attn:P21=44", "Attn:P11=1"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:10", "Impact:A5432", "Math.Add:S2345"]}, {"position": 24, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S23", "Attn:P1=49", "Attn:P13=46", "Math.Add:A9.SP"]}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:13", "Impact:A1", "Math.Add:S123", "Attn:P12=90", "Attn:P0=7"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:5", "Impact:A1", "Math.Add:S123"]}, {"position": 24, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:3", "Impact:A1", "Math.Add:S123", "Attn:P24=24", "Attn:P0=17", "Attn:P23=6", "Attn:P17=6", "Math.Add:A9.SP", "Math.Add:A10.SP"]}, {"position": 24, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:5", "Impact:A1", "Math.Add:S12", "Attn:P0=79", "Attn:P24=9", "Attn:P11=7", "Math.Add:A10.SP"]}, {"position": 24, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S3", "Attn:P11=78", "Attn:P0=17", "Attn:P17=4", "Math.Add:A9.SP"]}, {"position": 25, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:19", "Impact:A10", "Math.Add:S123", "Attn:P1=50", "Attn:P13=44", "Math.Add:A9.SP"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:33", "Impact:A10", "Math.Add:S123", "Attn:P12=43", "Attn:P0=38", "Attn:P23=6", "Attn:P24=2"]}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:33", "Impact:A10", "Math.Add:S123"]}, {"position": 25, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A10", "Math.Add:S12", "Attn:P11=78", "Attn:P0=17", "Attn:P17=4", "Math.Add:A9.SP"]}, {"position": 25, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:22", "Impact:A10", "Math.Add:S123"]}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:16", "Impact:A9", "Math.Add:S123", "Attn:P14=46", "Attn:P2=40", "Attn:P15=2", "Attn:P3=1", "Math.Add:A8.SP"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:32", "Impact:A9", "Math.Add:S123", "Attn:P1=44", "Attn:P13=43"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:32", "Impact:A9", "Math.Add:S123"]}, {"position": 26, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A9", "Math.Add:S2", "Attn:P11=82", "Attn:P0=13", "Attn:P17=4", "Math.Add:A8.SP"]}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:24", "Impact:A9", "Math.Add:S123"]}, {"position": 27, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:17", "Impact:A8", "Math.Add:S12345", "Attn:P3=49", "Attn:P15=45", "Math.Add:A7.SP"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:32", "Impact:A8", "Math.Add:S123", "Attn:P14=45", "Attn:P2=45"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:34", "Impact:A8", "Math.Add:S12345"]}, {"position": 27, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A8", "Math.Add:S23", "Attn:P11=81", "Attn:P0=15", "Attn:P22=3", "Attn:P20=1", "Math.Add:A7.SP"]}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:84", "Impact:A8", "Math.Add:S012345"]}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:24", "Impact:A7", "Math.Add:S12345", "Attn:P4=46", "Attn:P16=42", "Attn:P3=1", "Attn:P15=1"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:40", "Impact:A7", "Math.Add:S012345", "Attn:P3=46", "Attn:P15=44"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:40", "Impact:A7", "Math.Add:S012345"]}, {"position": 28, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A7", "Math.Add:S12345", "Attn:P11=74", "Attn:P0=18", "Attn:P22=4", "Attn:P20=3", "Math.Add:A6.SP"]}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:26", "Impact:A7", "Math.Add:S012345"]}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:26", "Impact:A6", "Math.Add:S12345", "Attn:P5=47", "Attn:P17=44", "Math.Add:A5.SP"]}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:45", "Impact:A6", "Math.Add:S012345", "Attn:P4=44", "Attn:P16=44", "Attn:P5=2", "Attn:P17=1"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:45", "Impact:A6", "Math.Add:S012345"]}, {"position": 29, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:4", "Impact:A6", "Math.Add:S2345", "Attn:P11=72", "Attn:P0=16", "Attn:P20=9", "Attn:P21=3", "Math.Add:A5.SP"]}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:39", "Impact:A6", "Math.Add:S012345"]}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S12", "Attn:P18=38", "Attn:P6=36", "Attn:P19=2", "Attn:P7=2", "Math.Add:A4.SP"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:66", "Impact:A5", "Math.Add:S012345", "Attn:P5=44", "Attn:P17=44", "Attn:P18=1"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:54", "Impact:A5", "Math.Add:S012345"]}, {"position": 30, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:3", "Impact:A5", "Math.Add:S5", "Attn:P21=34", "Attn:P23=18", "Attn:P11=10", "Attn:P20=7"]}, {"position": 30, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:4", "Impact:A5", "Math.Add:S12345", "Attn:P11=57", "Attn:P0=24", "Attn:P20=10", "Attn:P21=8", "Math.Add:A4.SP"]}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:43", "Impact:A5", "Math.Add:S012345"]}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:36", "Impact:A4", "Math.Add:S12345", "Attn:P7=40", "Attn:P19=37", "Attn:P6=2", "Attn:P18=2", "Math.Add:A0.SP"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A4", "Math.Add:S1", "Attn:P8=11", "Attn:P24=10", "Attn:P20=7", "Attn:P23=6", "Math.Add:A5.SP"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:89", "Impact:A42", "Math.Add:S012345", "Attn:P18=44", "Attn:P6=43", "Attn:P23=1", "Attn:P7=1"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:67", "Impact:A42", "Math.Add:S012345"]}, {"position": 31, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:4", "Impact:A4", "Math.Add:S45", "Attn:P21=27", "Attn:P23=20", "Attn:P11=10", "Attn:P13=4"]}, {"position": 31, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A4", "Math.Add:S2", "Attn:P31=35", "Attn:P30=13", "Attn:P21=8", "Attn:P20=5", "Math.Add:A3.SP", "Math.Add:A4.SP"]}, {"position": 31, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A4", "Math.Add:S23", "Attn:P11=54", "Attn:P0=25", "Attn:P21=19", "Math.Add:A3.SP"]}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:57", "Impact:A4", "Math.Add:S012345"]}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:32", "Impact:A3", "Math.Add:S12345", "Attn:P8=38", "Attn:P20=36", "Attn:P21=2", "Attn:P9=2"]}, {"position": 32, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:84", "Impact:A32", "Math.Add:S012345", "Attn:P7=42", "Attn:P19=41", "Attn:P11=9"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:69", "Impact:A32", "Math.Add:S012345"]}, {"position": 32, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:5", "Impact:A3", "Math.Add:S345", "Attn:P23=25", "Attn:P21=23", "Attn:P11=13", "Attn:P32=4", "Math.Add:A2.SP"]}, {"position": 32, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A3", "Math.Add:S23", "Attn:P11=51", "Attn:P0=26", "Attn:P21=19", "Attn:P23=2", "Math.Add:A2.SP"]}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:55", "Impact:A3", "Math.Add:S012345"]}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:28", "Impact:A2", "Math.Add:S12345", "Attn:P9=35", "Attn:P21=35", "Attn:P1=2", "Attn:P13=2"]}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A2", "Math.Add:S25", "Attn:P22=22", "Attn:P10=18", "Attn:P24=12", "Attn:P23=9"]}, {"position": 33, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:79", "Impact:A2", "Math.Add:S012345", "Attn:P20=45", "Attn:P8=44"]}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:75", "Impact:A2", "Math.Add:S012345"]}, {"position": 33, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A2", "Math.Add:S25", "Attn:P23=31", "Attn:P11=12", "Attn:P21=10", "Attn:P33=8", "Math.Add:A0.SP", "Math.Add:A1.SP"]}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A2", "Math.Add:S25", "Attn:P33=29", "Attn:P21=13", "Attn:P32=11", "Attn:P31=5", "Math.Add:A1.SP", "Math.Add:A2.SP"]}, {"position": 33, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A2", "Math.Add:S012", "Attn:P11=63", "Attn:P0=24", "Attn:P23=11", "Math.Add:A0.SP", "Math.Add:A1.SP"]}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:52", "Impact:A2", "Math.Add:S012345"]}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:32", "Impact:A1", "Math.Add:S12345", "Attn:P10=40", "Attn:P22=34", "Attn:P25=2", "Attn:P9=1"]}, {"position": 34, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:82", "Impact:A1", "Math.Add:S012345", "Attn:P9=45", "Attn:P21=44"]}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:75", "Impact:A1", "Math.Add:S012345"]}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:58", "Impact:A1", "Math.Add:S012345"]}, {"position": 35, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:81", "Impact:A0", "Math.Add:S012345", "Attn:P10=44", "Attn:P22=43"]}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:84", "Impact:A0", "Math.Add:S012345"]}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:48", "Impact:A0", "Math.Add:S0123"]}]
add_d11_l2_h3_t50K_s572091/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 0, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 0, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 0, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A7.ST"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.ST"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 20, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A4.ST"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.ST"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.ST"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 21, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.ST"]}, {"position": 21, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.ST"]}, {"position": 21, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 22, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A5.ST"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.ST"]}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 23, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A0.ST"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.ST"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 24, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A9.ST"]}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A10.ST"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 24, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 24, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 24, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 25, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A9.SC"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 25, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 25, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A8.SC"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A7.SC"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A6.SC"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A5.SC"]}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A4.SC"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.SC"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 31, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.SC"]}, {"position": 32, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 32, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SC"]}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 33, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A0.SC"]}, {"position": 34, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d11_l2_h3_t50K_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e27674a13e0ae5fe6195ee34aa2fb296018cebdcac6131d4a2226d589f7cd489
3
+ size 25182914
add_d11_l2_h3_t50K_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d12_l2_h3_t50K_s572091/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:5", "Impact:A210", "Math.Add:S24", "Attn:P15=51", "Attn:P2=46"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:5", "Impact:A210", "Math.Add:S24"]}, {"position": 18, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:4", "Impact:A210", "Math.Add:S24", "Attn:P3=50", "Attn:P16=45", "Math.Add:A8.SP"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:4", "Impact:A210", "Math.Add:S24"]}, {"position": 19, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:6", "Impact:A98", "Math.Add:S2345", "Attn:P4=49", "Attn:P17=49"]}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:5", "Impact:A8", "Math.Add:S245", "Attn:P18=51", "Attn:P5=43", "Math.Add:A6.SP"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:6", "Impact:A98", "Math.Add:S2345"]}, {"position": 21, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A7", "Math.Add:S4", "Attn:P20=49", "Attn:P7=44", "Attn:P19=1", "Math.Add:A4.SP"]}, {"position": 21, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A765", "Math.Add:S24", "Attn:P21=46", "Attn:P8=45", "Attn:P9=1", "Math.Add:A3.SP"]}, {"position": 21, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:9", "Impact:A87654", "Math.Add:S245"]}, {"position": 22, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:12", "Impact:A876", "Math.Add:S2345", "Attn:P19=49", "Attn:P6=48", "Math.Add:A5.SP"]}, {"position": 22, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:4", "Impact:A876", "Math.Add:S24", "Attn:P20=43", "Attn:P7=42", "Attn:P18=1", "Attn:P19=1", "Math.Add:A4.SP"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A7", "Math.Add:S4", "Attn:P21=17", "Attn:P7=12", "Attn:P8=12", "Attn:P20=10", "Math.Add:A2.SP"]}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:10", "Impact:A8765", "Math.Add:S2345"]}, {"position": 23, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A764", "Math.Add:S245", "Attn:P8=47", "Attn:P21=47"]}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A7654", "Math.Add:S235", "Attn:P22=46", "Attn:P9=45", "Math.Add:A2.SP"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:15", "Impact:A7654", "Math.Add:S2345"]}, {"position": 24, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:4", "Impact:A5432", "Math.Add:S2345", "Attn:P11=24", "Attn:P24=23", "Attn:P12=8", "Attn:P23=7"]}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:9", "Impact:A65432", "Math.Add:S12345", "Attn:P10=53", "Attn:P23=44", "Math.Add:A1.SP"]}, {"position": 24, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:9", "Impact:A5432", "Math.Add:S2345", "Attn:P11=56", "Attn:P24=41"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:16", "Impact:A65432", "Math.Add:S2345"]}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:18", "Impact:A21", "Math.Add:S1234", "Attn:P13=47", "Attn:P0=44", "Attn:P26=2"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:2", "Impact:A21", "Math.Add:S24", "Attn:P14=52", "Attn:P1=45", "Math.Add:A10.SP"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:18", "Impact:A21", "Math.Add:S1234"]}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:10", "Impact:A21", "Math.Add:S124"]}, {"position": 27, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:33", "Impact:A1", "Math.Add:S1234", "Attn:P13=47", "Attn:P0=44", "Attn:P12=2", "Attn:P26=1"]}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:15", "Impact:A1", "Math.Add:S124", "Attn:P14=49", "Attn:P1=45", "Math.Add:A10.SP"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:33", "Impact:A1", "Math.Add:S123"]}, {"position": 27, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S2", "Attn:P25=83", "Attn:P0=14", "Attn:P15=2", "Math.Add:A10.SP"]}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:86", "Impact:A1", "Math.Add:S012345"]}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:34", "Impact:A10", "Math.Add:S1234", "Attn:P14=46", "Attn:P1=46"]}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:2", "Impact:A10", "Math.Add:S12", "Attn:P2=48", "Attn:P15=40", "Attn:P12=2", "Attn:P14=1"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:33", "Impact:A10", "Math.Add:S123"]}, {"position": 28, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:2", "Impact:A10", "Math.Add:S12", "Attn:P25=71", "Attn:P0=11", "Attn:P15=11", "Attn:P18=6", "Math.Add:A9.SP"]}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:84", "Impact:A10", "Math.Add:S012345"]}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:33", "Impact:A9", "Math.Add:S1234", "Attn:P15=46", "Attn:P2=46"]}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:23", "Impact:A9", "Math.Add:S1234", "Attn:P3=51", "Attn:P16=42", "Math.Add:A8.SP"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:33", "Impact:A9", "Math.Add:S1234"]}, {"position": 29, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A9", "Math.Add:S2", "Attn:P25=75", "Attn:P0=23", "Attn:P19=2", "Math.Add:A8.SP"]}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:80", "Impact:A9", "Math.Add:S012345"]}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:32", "Impact:A8", "Math.Add:S1234", "Attn:P3=48", "Attn:P16=47"]}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:24", "Impact:A8", "Math.Add:S12345", "Attn:P4=45", "Attn:P17=40", "Attn:P3=1", "Attn:P18=1", "Math.Add:A7.SP"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:35", "Impact:A8", "Math.Add:S12345"]}, {"position": 30, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:6", "Impact:A8", "Math.Add:S2345", "Attn:P25=74", "Attn:P0=18", "Attn:P19=6", "Attn:P22=2", "Math.Add:A7.SP", "Math.Add:A9.SP"]}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:77", "Impact:A8", "Math.Add:S012345"]}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:41", "Impact:A7", "Math.Add:S012345", "Attn:P4=45", "Attn:P17=45", "Attn:P25=2"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:25", "Impact:A7", "Math.Add:S12345", "Attn:P18=48", "Attn:P5=46"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:41", "Impact:A7", "Math.Add:S012345"]}, {"position": 31, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:5", "Impact:A7", "Math.Add:S2345", "Attn:P25=72", "Attn:P0=20", "Attn:P22=6", "Attn:P21=1", "Math.Add:A6.SP"]}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:76", "Impact:A7", "Math.Add:S012345"]}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:86", "Impact:A6", "Math.Add:S012345", "Attn:P5=46", "Attn:P18=46"]}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:31", "Impact:A6", "Math.Add:S12345", "Attn:P19=44", "Attn:P6=42", "Math.Add:A5.SP"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:45", "Impact:A6", "Math.Add:S012345"]}, {"position": 32, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:6", "Impact:A6", "Math.Add:S2345", "Attn:P25=66", "Attn:P0=21", "Attn:P22=8", "Attn:P21=2", "Math.Add:A5.SP"]}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:45", "Impact:A6", "Math.Add:S01235"]}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:80", "Impact:A5", "Math.Add:S012345", "Attn:P19=46", "Attn:P6=45"]}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:29", "Impact:A5", "Math.Add:S1245", "Attn:P7=42", "Attn:P20=41", "Attn:P21=2", "Attn:P8=2", "Math.Add:A4.SP"]}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:52", "Impact:A5", "Math.Add:S012345"]}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:11", "Impact:A5", "Math.Add:S2345", "Attn:P25=57", "Attn:P0=28", "Attn:P24=6", "Attn:P21=6", "Math.Add:A4.SP"]}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:57", "Impact:A5", "Math.Add:S012345"]}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:87", "Impact:A4", "Math.Add:S012345", "Attn:P20=46", "Attn:P7=44"]}, {"position": 34, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:16", "Impact:A4", "Math.Add:S123", "Attn:P8=40", "Attn:P21=38", "Attn:P22=1", "Attn:P9=1"]}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:69", "Impact:A4", "Math.Add:S012345"]}, {"position": 34, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:19", "Impact:A4", "Math.Add:S12345", "Attn:P25=52", "Attn:P0=27", "Attn:P24=11", "Attn:P23=9", "Math.Add:A3.SP"]}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:65", "Impact:A4", "Math.Add:S012345"]}, {"position": 35, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:86", "Impact:A3", "Math.Add:S012345", "Attn:P8=46", "Attn:P21=45"]}, {"position": 35, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:30", "Impact:A3", "Math.Add:S12345", "Attn:P22=40", "Attn:P9=37", "Attn:P23=1", "Attn:P10=1", "Math.Add:A2.SP"]}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:86", "Impact:A3", "Math.Add:S012345"]}, {"position": 35, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:13", "Impact:A3", "Math.Add:S2345", "Attn:P25=58", "Attn:P0=24", "Attn:P24=17", "Math.Add:A2.SP"]}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:74", "Impact:A3", "Math.Add:S012345"]}, {"position": 36, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:83", "Impact:A2", "Math.Add:S012345", "Attn:P22=47", "Attn:P9=46"]}, {"position": 36, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:21", "Impact:A2", "Math.Add:S12345", "Attn:P10=37", "Attn:P23=36", "Attn:P24=2", "Attn:P11=2", "Math.Add:A1.SP"]}, {"position": 36, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:90", "Impact:A2", "Math.Add:S012345"]}, {"position": 36, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:10", "Impact:A2", "Math.Add:S12345", "Attn:P25=59", "Attn:P0=27", "Attn:P24=13", "Math.Add:A1.SP"]}, {"position": 36, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:69", "Impact:A2", "Math.Add:S012345"]}, {"position": 37, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:93", "Impact:A1", "Math.Add:S012345", "Attn:P10=43", "Attn:P23=42", "Attn:P24=1"]}, {"position": 37, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:28", "Impact:A1", "Math.Add:S12345", "Attn:P11=44", "Attn:P24=39"]}, {"position": 37, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:82", "Impact:A1", "Math.Add:S012345"]}, {"position": 37, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S1", "Attn:P25=72", "Attn:P0=28", "Math.Add:A0.SP"]}, {"position": 37, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:62", "Impact:A1", "Math.Add:S012345"]}, {"position": 38, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:93", "Impact:A0", "Math.Add:S012345", "Attn:P11=43", "Attn:P24=41", "Attn:P12=2"]}, {"position": 38, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:73", "Impact:A0", "Math.Add:S012345"]}, {"position": 38, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:66", "Impact:A0", "Math.Add:S012345"]}]
add_d12_l2_h3_t50K_s572091/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A9.ST"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 18, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A8.ST"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 19, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A7.ST"]}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A6.ST"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 21, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A4.ST"]}, {"position": 21, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.ST"]}, {"position": 21, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 22, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A5.ST"]}, {"position": 22, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.ST"]}, {"position": 22, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 22, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 23, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.ST"]}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.ST"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 24, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 24, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.ST"]}, {"position": 24, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.ST"]}, {"position": 24, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A11.ST"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A10.ST"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 27, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A9.SC"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 1, "is_head": true, "num": 1, "tags": ["Algo:A9.SC"]}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A8.SC"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A7.SC"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A6.SC"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.SC"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SC"]}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 34, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SC"]}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 35, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SC"]}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 36, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 36, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SC"]}, {"position": 36, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 36, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 36, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 37, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 37, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SC"]}, {"position": 37, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 37, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 37, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 38, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 38, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 38, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d12_l2_h3_t50K_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db98e4bccb819d17489e6a1fea543b7b0a280084586d33a5a1982230b5f0117c
3
+ size 25189570
add_d12_l2_h3_t50K_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d13_l2_h3_t50K_s572091/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 23, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A76", "Math.Add:S34", "Attn:P21=31", "Attn:P7=29", "Attn:P22=9", "Attn:P8=8"]}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:4", "Impact:A76", "Math.Add:S345", "Attn:P9=32", "Attn:P23=30", "Attn:P13=28", "Attn:P7=1"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:5", "Impact:A76", "Math.Add:S345", "Attn:P8=38", "Attn:P22=38", "Attn:P9=8", "Attn:P23=7"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:5", "Impact:A76", "Math.Add:S345"]}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A98", "Math.Add:S3", "Attn:P5=45", "Attn:P19=43", "Attn:P21=2", "Attn:P7=2"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A8", "Math.Add:S3", "Attn:P21=44", "Attn:P7=43", "Attn:P22=2", "Attn:P8=2", "Math.Add:A4.SP", "Math.Add:A5.SP"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A98", "Math.Add:S3", "Attn:P6=45", "Attn:P20=42", "Attn:P21=3", "Attn:P7=3", "Math.Add:A6.SP"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:1", "Impact:A98", "Math.Add:S3"]}, {"position": 27, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:12", "Impact:A76543", "Math.Add:S345", "Attn:P24=41", "Attn:P10=40", "Attn:P11=5", "Attn:P25=4", "Math.Add:A1.SP"]}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A6543", "Math.Add:S345", "Attn:P27=34", "Attn:P12=25", "Attn:P26=24", "Attn:P13=1"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:13", "Impact:A6543", "Math.Add:S345", "Attn:P11=34", "Attn:P25=32", "Attn:P12=11", "Attn:P26=10"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:15", "Impact:A76543", "Math.Add:S345"]}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A31", "Math.Add:S3", "Attn:P2=31", "Attn:P16=29", "Attn:P13=23", "Attn:P15=1"]}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:22", "Impact:A31", "Math.Add:S123", "Attn:P14=49", "Attn:P0=48"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A31", "Math.Add:S23", "Attn:P1=39", "Attn:P15=35", "Attn:P14=2", "Attn:P0=2"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:22", "Impact:A31", "Math.Add:S123"]}, {"position": 28, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:4", "Impact:A31", "Math.Add:S12", "Attn:P0=88", "Attn:P2=2", "Attn:P1=2", "Attn:P4=2", "Math.Add:A12.SP"]}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:8", "Impact:A31", "Math.Add:S123"]}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A21", "Math.Add:S23", "Attn:P2=33", "Attn:P16=32", "Attn:P29=11", "Attn:P1=2", "Math.Add:A10.SP"]}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:34", "Impact:A21", "Math.Add:S123", "Attn:P14=44", "Attn:P0=43", "Attn:P27=5"]}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:25", "Impact:A21", "Math.Add:S123", "Attn:P1=42", "Attn:P15=35", "Attn:P2=2", "Attn:P16=2", "Math.Add:A11.SP"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:34", "Impact:A21", "Math.Add:S123"]}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:30", "Impact:A21", "Math.Add:S123"]}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S2", "Attn:P3=24", "Attn:P17=23", "Attn:P18=5", "Attn:P2=5"]}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:34", "Impact:A1", "Math.Add:S123", "Attn:P1=46", "Attn:P15=45"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:21", "Impact:A1", "Math.Add:S123", "Attn:P2=43", "Attn:P16=40", "Attn:P17=2", "Attn:P3=2", "Math.Add:A10.SP"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:35", "Impact:A1", "Math.Add:S123"]}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:29", "Impact:A1", "Math.Add:S123"]}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A10", "Math.Add:S23", "Attn:P18=31", "Attn:P4=28", "Attn:P27=3", "Attn:P3=3", "Math.Add:A8.SP"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:35", "Impact:A10", "Math.Add:S123", "Attn:P2=46", "Attn:P16=44"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:19", "Impact:A10", "Math.Add:S123", "Attn:P17=43", "Attn:P3=42", "Attn:P2=1", "Attn:P16=1", "Math.Add:A9.SP"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:35", "Impact:A10", "Math.Add:S123"]}, {"position": 31, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A10", "Math.Add:S1", "Attn:P1=21", "Attn:P2=19", "Attn:P13=19", "Attn:P0=8", "Math.Add:A9.SP"]}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:21", "Impact:A10", "Math.Add:S123"]}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:3", "Impact:A9", "Math.Add:S23", "Attn:P19=27", "Attn:P5=25", "Attn:P6=5", "Attn:P20=5"]}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:35", "Impact:A9", "Math.Add:S123", "Attn:P3=47", "Attn:P17=44", "Attn:P28=1"]}, {"position": 32, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:19", "Impact:A9", "Math.Add:S123", "Attn:P18=41", "Attn:P4=38", "Attn:P5=2", "Attn:P19=2", "Math.Add:A8.SP"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:36", "Impact:A9", "Math.Add:S123"]}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:29", "Impact:A9", "Math.Add:S123"]}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A8", "Math.Add:S23", "Attn:P20=30", "Attn:P6=29", "Attn:P21=6", "Attn:P7=5"]}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:30", "Impact:A8", "Math.Add:S123", "Attn:P4=45", "Attn:P18=45", "Attn:P28=1"]}, {"position": 33, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:22", "Impact:A8", "Math.Add:S123", "Attn:P5=42", "Attn:P19=41", "Attn:P6=3", "Attn:P20=2", "Math.Add:A7.SP"]}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:31", "Impact:A8", "Math.Add:S123"]}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A8", "Math.Add:S3", "Attn:P13=77", "Attn:P23=3", "Attn:P32=3", "Attn:P26=3", "Math.Add:A7.SP", "Math.Add:A8.SP"]}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:25", "Impact:A8", "Math.Add:S123"]}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:5", "Impact:A7", "Math.Add:S2345", "Attn:P7=35", "Attn:P21=34", "Attn:P22=5", "Attn:P8=5", "Math.Add:A4.SP"]}, {"position": 34, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:39", "Impact:A76", "Math.Add:S012345", "Attn:P5=43", "Attn:P19=42", "Attn:P6=1", "Attn:P28=1"]}, {"position": 34, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:25", "Impact:A7", "Math.Add:S12345", "Attn:P6=40", "Attn:P20=40", "Attn:P21=2", "Attn:P7=2", "Math.Add:A6.SP"]}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:39", "Impact:A76", "Math.Add:S012345"]}, {"position": 34, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:2", "Impact:A7", "Math.Add:S345", "Attn:P13=76", "Attn:P23=5", "Attn:P33=4", "Attn:P26=2", "Math.Add:A6.SP", "Math.Add:A7.SP", "Math.Add:A12.SP"]}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:34", "Impact:A7", "Math.Add:S012345"]}, {"position": 35, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:11", "Impact:A65", "Math.Add:S2345", "Attn:P22=31", "Attn:P8=30", "Attn:P9=7", "Attn:P23=6", "Math.Add:A4.SP"]}, {"position": 35, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:58", "Impact:A65", "Math.Add:S012345", "Attn:P6=43", "Attn:P20=42", "Attn:P27=2", "Attn:P28=2"]}, {"position": 35, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:29", "Impact:A65", "Math.Add:S12345", "Attn:P7=40", "Attn:P21=40", "Attn:P9=2", "Attn:P23=2", "Math.Add:A5.SP"]}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:46", "Impact:A65", "Math.Add:S012345"]}, {"position": 35, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:4", "Impact:A6", "Math.Add:S345", "Attn:P13=71", "Attn:P23=9", "Attn:P34=5", "Attn:P35=1", "Math.Add:A5.SP", "Math.Add:A6.SP"]}, {"position": 35, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A6", "Math.Add:S45", "Attn:P23=21", "Attn:P27=16", "Attn:P13=11", "Attn:P26=8", "Math.Add:A6.SP"]}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:40", "Impact:A6", "Math.Add:S012345"]}, {"position": 36, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:10", "Impact:A54", "Math.Add:S2345", "Attn:P9=30", "Attn:P23=30", "Attn:P8=7", "Attn:P22=7", "Math.Add:A2.SP"]}, {"position": 36, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:92", "Impact:A54", "Math.Add:S012345", "Attn:P21=43", "Attn:P7=43", "Attn:P28=2", "Attn:P22=1"]}, {"position": 36, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:30", "Impact:A54", "Math.Add:S12345", "Attn:P22=37", "Attn:P8=37", "Attn:P23=2", "Attn:P9=2", "Math.Add:A4.SP"]}, {"position": 36, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:56", "Impact:A54", "Math.Add:S012345"]}, {"position": 36, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S2", "Attn:P1=27", "Attn:P13=25", "Attn:P2=21", "Attn:P36=11", "Math.Add:A3.SP", "Math.Add:A4.SP", "Math.Add:A5.SP"]}, {"position": 36, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A5", "Math.Add:S345", "Attn:P13=65", "Attn:P35=9", "Attn:P27=7", "Attn:P36=1", "Math.Add:A4.SP", "Math.Add:A5.SP"]}, {"position": 36, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S45", "Attn:P27=28", "Attn:P36=13", "Attn:P23=13", "Attn:P13=11", "Math.Add:A2.SP", "Math.Add:A5.SP"]}, {"position": 36, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:46", "Impact:A5", "Math.Add:S012345"]}, {"position": 37, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:16", "Impact:A43", "Math.Add:S2345", "Attn:P24=30", "Attn:P10=30", "Attn:P27=4", "Attn:P11=4", "Math.Add:A2.SP"]}, {"position": 37, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:92", "Impact:A43", "Math.Add:S012345", "Attn:P8=44", "Attn:P22=43", "Attn:P28=2"]}, {"position": 37, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:36", "Impact:A43", "Math.Add:S12345", "Attn:P23=37", "Attn:P9=36", "Attn:P24=2", "Attn:P10=2", "Math.Add:A3.SP"]}, {"position": 37, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:79", "Impact:A432", "Math.Add:S012345"]}, {"position": 37, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A4", "Math.Add:S12", "Attn:P13=28", "Attn:P1=21", "Attn:P37=18", "Attn:P2=16", "Math.Add:A0.SP", "Math.Add:A2.SP", "Math.Add:A3.SP", "Math.Add:A4.SP"]}, {"position": 37, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:11", "Impact:A4", "Math.Add:S345", "Attn:P13=55", "Attn:P36=9", "Attn:P27=8", "Attn:P37=3", "Math.Add:A0.SP", "Math.Add:A3.SP", "Math.Add:A4.SP", "Math.Add:A10.SP"]}, {"position": 37, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A4", "Math.Add:S45", "Attn:P27=25", "Attn:P37=18", "Attn:P13=12", "Attn:P23=10", "Math.Add:A2.SP", "Math.Add:A4.SP"]}, {"position": 37, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:53", "Impact:A4", "Math.Add:S012345"]}, {"position": 38, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:15", "Impact:A31", "Math.Add:S2345", "Attn:P11=29", "Attn:P25=27", "Attn:P24=5", "Attn:P27=5"]}, {"position": 38, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:94", "Impact:A321", "Math.Add:S012345", "Attn:P23=42", "Attn:P9=42", "Attn:P28=3", "Attn:P13=2"]}, {"position": 38, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:30", "Impact:A3", "Math.Add:S12345", "Attn:P24=39", "Attn:P10=36", "Attn:P11=2", "Attn:P25=2", "Math.Add:A2.SP"]}, {"position": 38, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:71", "Impact:A321", "Math.Add:S012345"]}, {"position": 38, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A3", "Math.Add:S2", "Attn:P13=28", "Attn:P1=22", "Attn:P2=18", "Attn:P38=11", "Math.Add:A1.SP", "Math.Add:A2.SP", "Math.Add:A3.SP"]}, {"position": 38, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A3", "Math.Add:S345", "Attn:P13=62", "Attn:P37=12", "Attn:P27=5", "Attn:P36=2", "Math.Add:A0.SP", "Math.Add:A2.SP", "Math.Add:A3.SP", "Math.Add:A5.SP"]}, {"position": 38, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A3", "Math.Add:S4", "Attn:P38=20", "Attn:P27=15", "Attn:P13=14", "Attn:P37=12", "Math.Add:A1.SP", "Math.Add:A2.SP", "Math.Add:A3.SP"]}, {"position": 38, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:57", "Impact:A3", "Math.Add:S012345"]}, {"position": 39, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:8", "Impact:A2", "Math.Add:S2345", "Attn:P12=38", "Attn:P26=31", "Attn:P11=4", "Attn:P25=4"]}, {"position": 39, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:94", "Impact:A2", "Math.Add:S012345", "Attn:P24=44", "Attn:P10=44", "Attn:P11=1"]}, {"position": 39, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:30", "Impact:A2", "Math.Add:S12345", "Attn:P11=37", "Attn:P25=34", "Attn:P24=2", "Attn:P12=2", "Math.Add:A1.SP"]}, {"position": 39, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:80", "Impact:A2", "Math.Add:S012345"]}, {"position": 39, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:7", "Impact:A2", "Math.Add:S1234", "Attn:P13=29", "Attn:P1=22", "Attn:P2=19", "Attn:P0=5", "Math.Add:A1.SP", "Math.Add:A2.SP", "Math.Add:A4.SP", "Math.Add:A5.SP", "Math.Add:A7.SP", "Math.Add:A12.SP"]}, {"position": 39, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:66", "Impact:A2", "Math.Add:S012345"]}, {"position": 40, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:4", "Impact:A1", "Math.Add:S145", "Attn:P12=32", "Attn:P26=27", "Attn:P29=5", "Attn:P11=2"]}, {"position": 40, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:93", "Impact:A1", "Math.Add:S012345", "Attn:P11=46", "Attn:P25=44", "Attn:P26=1", "Attn:P12=1"]}, {"position": 40, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:23", "Impact:A1", "Math.Add:S12345", "Attn:P12=37", "Attn:P26=31", "Attn:P11=3", "Attn:P25=3"]}, {"position": 40, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:78", "Impact:A1", "Math.Add:S012345"]}, {"position": 40, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A1", "Math.Add:S1234", "Attn:P1=25", "Attn:P2=22", "Attn:P13=22", "Attn:P0=7", "Math.Add:A1.SP", "Math.Add:A2.SP", "Math.Add:A5.SP", "Math.Add:A11.SP"]}, {"position": 40, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:67", "Impact:A1", "Math.Add:S012345"]}, {"position": 41, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:94", "Impact:A0", "Math.Add:S012345", "Attn:P12=44", "Attn:P26=42", "Attn:P27=2"]}, {"position": 41, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:74", "Impact:A0", "Math.Add:S012345"]}, {"position": 41, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:81", "Impact:A0", "Math.Add:S012345"]}]
add_d13_l2_h3_t50K_s572091/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 23, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 23, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.ST"]}, {"position": 23, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 23, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A7.ST"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.ST"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.ST"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.ST"]}, {"position": 27, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.ST"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.ST"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A10.ST"]}, {"position": 28, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A12.ST"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A11.ST"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 28, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A9.SS"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A10.SA"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A9.SC"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A8.SS"]}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A9.SA"]}, {"position": 32, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.SC"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A7.SS"]}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A8.SA"]}, {"position": 33, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A7.SC"]}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A6.SS"]}, {"position": 34, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A7.SA"]}, {"position": 34, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.SC"]}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A5.SS"]}, {"position": 35, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A6.SA"]}, {"position": 35, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.SC"]}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 35, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 36, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A4.SS"]}, {"position": 36, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.SA"]}, {"position": 36, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.SC"]}, {"position": 36, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 36, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 36, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 36, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 36, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 37, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.SS"]}, {"position": 37, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA"]}, {"position": 37, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SC"]}, {"position": 37, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 37, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 37, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 37, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 37, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 38, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.SS"]}, {"position": 38, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 38, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SC"]}, {"position": 38, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 38, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 38, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 38, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 38, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 39, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SS"]}, {"position": 39, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 39, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.SC"]}, {"position": 39, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 39, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 39, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 40, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 40, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 40, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.SC"]}, {"position": 40, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 40, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 40, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 41, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 41, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 41, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d13_l2_h3_t50K_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05b44b9336ea285a8eff8ebcdb8876532bb2be0a4259c92b3dcadf4322c4b828
3
+ size 25196162
add_d13_l2_h3_t50K_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d14_l2_h3_t60K_s572091/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 25, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A31", "Math.Add:S2", "Attn:P18=37", "Attn:P3=37", "Attn:P19=4", "Attn:P4=4"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A31", "Math.Add:S2", "Attn:P2=30", "Attn:P17=29", "Attn:P3=11", "Attn:P18=11", "Math.Add:A10.SP"]}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:0", "Impact:A31", "Math.Add:S2"]}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A8", "Math.Add:S345", "Attn:P8=43", "Attn:P23=43", "Attn:P7=1", "Attn:P22=1", "Math.Add:A2.SP", "Math.Add:A5.SP"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A8", "Math.Add:S45", "Attn:P24=18", "Attn:P9=18", "Attn:P22=13", "Attn:P7=13"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A8", "Math.Add:S345", "Attn:P21=27", "Attn:P6=27", "Attn:P22=18", "Attn:P7=18"]}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:2", "Impact:A8", "Math.Add:S345"]}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A8", "Math.Add:S5", "Attn:P25=18", "Attn:P10=15", "Attn:P24=9", "Attn:P9=8"]}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:0", "Impact:A8", "Math.Add:S5"]}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S4", "Attn:P6=46", "Attn:P21=44", "Attn:P20=1", "Attn:P23=1", "Math.Add:A7.SP"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S4", "Attn:P19=30", "Attn:P4=29", "Attn:P20=12", "Attn:P5=12"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S4"]}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A7654", "Math.Add:S245", "Attn:P11=54", "Attn:P26=45", "Math.Add:A2.SP"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:2", "Impact:A7654", "Math.Add:S245"]}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A41", "Math.Add:S2", "Attn:P1=49", "Attn:P16=46", "Math.Add:A2.SP", "Math.Add:A12.SP"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:19", "Impact:A41", "Math.Add:S1234", "Attn:P15=42", "Attn:P0=41", "Attn:P1=2", "Attn:P16=2"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:18", "Impact:A41", "Math.Add:S1234"]}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:40", "Impact:A41", "Math.Add:S1234"]}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:42", "Impact:A31", "Math.Add:S1234", "Attn:P15=39", "Attn:P0=36", "Attn:P29=7", "Attn:P1=2"]}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:10", "Impact:A31", "Math.Add:S12", "Attn:P17=24", "Attn:P2=23", "Attn:P3=9", "Attn:P18=8"]}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:22", "Impact:A31", "Math.Add:S1234", "Attn:P1=29", "Attn:P16=27", "Attn:P17=11", "Attn:P2=10"]}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:41", "Impact:A31", "Math.Add:S1234"]}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:95", "Impact:A31", "Math.Add:S012345"]}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:40", "Impact:A21", "Math.Add:S1234", "Attn:P16=40", "Attn:P1=39", "Attn:P29=3", "Attn:P30=3"]}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:3", "Impact:A21", "Math.Add:S12", "Attn:P18=21", "Attn:P3=20", "Attn:P4=9", "Attn:P19=9"]}, {"position": 32, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:18", "Impact:A21", "Math.Add:S12", "Attn:P2=26", "Attn:P17=26", "Attn:P3=10", "Attn:P18=10"]}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:42", "Impact:A21", "Math.Add:S1234"]}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:96", "Impact:A21", "Math.Add:S012345"]}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:61", "Impact:A1", "Math.Add:S01234", "Attn:P2=39", "Attn:P17=39", "Attn:P29=3", "Attn:P30=2"]}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A1", "Math.Add:S1234", "Attn:P19=28", "Attn:P4=27", "Attn:P14=5", "Attn:P20=4"]}, {"position": 33, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:25", "Impact:A10", "Math.Add:S1234", "Attn:P18=28", "Attn:P3=27", "Attn:P4=10", "Attn:P19=10"]}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:44", "Impact:A10", "Math.Add:S1234"]}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S4", "Attn:P14=74", "Attn:P0=20", "Attn:P28=4", "Attn:P33=1", "Math.Add:A10.SP"]}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:92", "Impact:A1", "Math.Add:S012345"]}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:86", "Impact:A10", "Math.Add:S01234", "Attn:P18=33", "Attn:P3=31", "Attn:P29=7", "Attn:P30=4"]}, {"position": 34, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A10", "Math.Add:S124", "Attn:P5=23", "Attn:P20=22", "Attn:P21=12", "Attn:P6=12"]}, {"position": 34, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:25", "Impact:A10", "Math.Add:S1234", "Attn:P4=29", "Attn:P19=29", "Attn:P5=12", "Attn:P20=12"]}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:40", "Impact:A10", "Math.Add:S1234"]}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:96", "Impact:A10", "Math.Add:S012345"]}, {"position": 35, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:40", "Impact:A9", "Math.Add:S1234", "Attn:P19=35", "Attn:P4=32", "Attn:P29=6", "Attn:P30=5"]}, {"position": 35, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:8", "Impact:A9", "Math.Add:S124", "Attn:P21=28", "Attn:P6=27", "Attn:P22=10", "Attn:P7=9"]}, {"position": 35, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:20", "Impact:A9", "Math.Add:S1234", "Attn:P5=29", "Attn:P20=28", "Attn:P21=11", "Attn:P6=11"]}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:37", "Impact:A9", "Math.Add:S1234"]}, {"position": 35, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A9", "Math.Add:S3", "Attn:P14=79", "Attn:P0=19", "Math.Add:A8.SP", "Math.Add:A9.SP", "Math.Add:A12.SP"]}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:93", "Impact:A9", "Math.Add:S012345"]}, {"position": 36, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:60", "Impact:A8", "Math.Add:S012345", "Attn:P5=39", "Attn:P20=39", "Attn:P29=3", "Attn:P30=2"]}, {"position": 36, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A8", "Math.Add:S12345", "Attn:P22=28", "Attn:P7=28", "Attn:P23=6", "Attn:P8=6"]}, {"position": 36, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:22", "Impact:A8", "Math.Add:S12345", "Attn:P21=29", "Attn:P6=29", "Attn:P22=8", "Attn:P7=8"]}, {"position": 36, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:41", "Impact:A8", "Math.Add:S012345"]}, {"position": 36, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A8", "Math.Add:S35", "Attn:P14=23", "Attn:P27=11", "Attn:P35=10", "Attn:P36=4"]}, {"position": 36, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A8", "Math.Add:S4", "Attn:P14=77", "Attn:P0=20", "Attn:P26=2", "Math.Add:A7.SP", "Math.Add:A8.SP", "Math.Add:A10.SP", "Math.Add:A11.SP", "Math.Add:A12.SP"]}, {"position": 36, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A8", "Math.Add:S5", "Attn:P0=13", "Attn:P29=9", "Attn:P27=6", "Attn:P14=6", "Math.Add:A2.SP", "Math.Add:A8.SP"]}, {"position": 36, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:86", "Impact:A8", "Math.Add:S012345"]}, {"position": 37, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:60", "Impact:A7654", "Math.Add:S012345", "Attn:P21=26", "Attn:P6=25", "Attn:P14=11", "Attn:P29=7"]}, {"position": 37, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:9", "Impact:A7", "Math.Add:S12345", "Attn:P8=21", "Attn:P23=21", "Attn:P24=16", "Attn:P9=15"]}, {"position": 37, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:32", "Impact:A7", "Math.Add:S12345", "Attn:P7=28", "Attn:P22=28", "Attn:P8=13", "Attn:P23=13"]}, {"position": 37, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:50", "Impact:A7", "Math.Add:S012345"]}, {"position": 37, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A7", "Math.Add:S45", "Attn:P14=25", "Attn:P36=11", "Attn:P27=7", "Attn:P37=5"]}, {"position": 37, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A7", "Math.Add:S1", "Attn:P14=67", "Attn:P0=30", "Attn:P37=2", "Math.Add:A6.SP", "Math.Add:A7.SP", "Math.Add:A9.SP"]}, {"position": 37, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A7", "Math.Add:S5", "Attn:P0=17", "Attn:P14=8", "Attn:P27=6", "Attn:P29=4", "Math.Add:A7.SP"]}, {"position": 37, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:87", "Impact:A7", "Math.Add:S012345"]}, {"position": 38, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:69", "Impact:A654", "Math.Add:S012345", "Attn:P7=34", "Attn:P22=33", "Attn:P29=4", "Attn:P14=4"]}, {"position": 38, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A654", "Math.Add:S12345", "Attn:P24=23", "Attn:P9=23", "Attn:P10=13", "Attn:P25=12"]}, {"position": 38, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:37", "Impact:A654", "Math.Add:S12345", "Attn:P8=28", "Attn:P23=27", "Attn:P9=12", "Attn:P24=12"]}, {"position": 38, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:56", "Impact:A654", "Math.Add:S012345"]}, {"position": 38, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A6", "Math.Add:S5", "Attn:P14=20", "Attn:P37=11", "Attn:P27=8", "Attn:P38=5"]}, {"position": 38, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A6", "Math.Add:S4", "Attn:P14=62", "Attn:P0=33", "Attn:P38=2", "Attn:P29=1", "Math.Add:A5.SP", "Math.Add:A7.SP", "Math.Add:A10.SP", "Math.Add:A11.SP", "Math.Add:A12.SP"]}, {"position": 38, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:52", "Impact:A6", "Math.Add:S012345"]}, {"position": 39, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:78", "Impact:A54", "Math.Add:S012345", "Attn:P8=42", "Attn:P23=42", "Attn:P24=2", "Attn:P14=1"]}, {"position": 39, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A54", "Math.Add:S2345", "Attn:P10=21", "Attn:P25=21", "Attn:P24=9", "Attn:P9=9"]}, {"position": 39, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:39", "Impact:A5", "Math.Add:S12345", "Attn:P24=27", "Attn:P9=27", "Attn:P25=10", "Attn:P10=10"]}, {"position": 39, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:52", "Impact:A5", "Math.Add:S0123"]}, {"position": 39, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:2", "Impact:A5", "Math.Add:S235", "Attn:P38=12", "Attn:P14=10", "Attn:P39=9", "Attn:P37=8"]}, {"position": 39, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S0", "Attn:P14=59", "Attn:P0=34", "Attn:P39=3", "Attn:P29=2", "Math.Add:A4.SP", "Math.Add:A5.SP", "Math.Add:A6.SP", "Math.Add:A9.SP", "Math.Add:A10.SP", "Math.Add:A11.SP", "Math.Add:A12.SP"]}, {"position": 39, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A5", "Math.Add:S5", "Attn:P0=15", "Attn:P39=7", "Attn:P14=7", "Attn:P27=7", "Math.Add:A5.SP"]}, {"position": 39, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:55", "Impact:A5", "Math.Add:S012345"]}, {"position": 40, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:86", "Impact:A4", "Math.Add:S012345", "Attn:P9=41", "Attn:P24=41", "Attn:P10=1", "Attn:P30=1"]}, {"position": 40, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:10", "Impact:A4", "Math.Add:S1245", "Attn:P11=26", "Attn:P26=24", "Attn:P30=4", "Attn:P14=4"]}, {"position": 40, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:37", "Impact:A4", "Math.Add:S12345", "Attn:P25=25", "Attn:P10=25", "Attn:P26=9", "Attn:P11=9"]}, {"position": 40, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:80", "Impact:A4", "Math.Add:S01234"]}, {"position": 40, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A4", "Math.Add:S5", "Attn:P39=12", "Attn:P27=8", "Attn:P40=7", "Attn:P14=7", "Math.Add:A3.SP"]}, {"position": 40, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A4", "Math.Add:S5", "Attn:P0=21", "Attn:P14=9", "Attn:P29=9", "Attn:P27=7", "Math.Add:A4.SP"]}, {"position": 40, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:62", "Impact:A4", "Math.Add:S012345"]}, {"position": 41, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:82", "Impact:A3", "Math.Add:S01234", "Attn:P25=41", "Attn:P10=41", "Attn:P26=1", "Attn:P11=1"]}, {"position": 41, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:3", "Impact:A3", "Math.Add:S12", "Attn:P12=26", "Attn:P27=24", "Attn:P14=5", "Attn:P30=5"]}, {"position": 41, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:27", "Impact:A3", "Math.Add:S1245", "Attn:P11=25", "Attn:P26=25", "Attn:P12=9", "Attn:P27=9"]}, {"position": 41, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:82", "Impact:A3", "Math.Add:S012345"]}, {"position": 41, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:62", "Impact:A3", "Math.Add:S012345"]}, {"position": 42, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:90", "Impact:A2", "Math.Add:S012345", "Attn:P11=40", "Attn:P26=38", "Attn:P12=2", "Attn:P27=1"]}, {"position": 42, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A2", "Math.Add:S124", "Attn:P13=24", "Attn:P28=24", "Attn:P12=4", "Attn:P27=4"]}, {"position": 42, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:31", "Impact:A2", "Math.Add:S123", "Attn:P12=25", "Attn:P27=24", "Attn:P13=11", "Attn:P28=10"]}, {"position": 42, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:82", "Impact:A2", "Math.Add:S012345"]}, {"position": 42, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:70", "Impact:A2", "Math.Add:S012345"]}, {"position": 43, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:87", "Impact:A1", "Math.Add:S01234", "Attn:P12=42", "Attn:P27=38", "Attn:P13=2", "Attn:P28=2"]}, {"position": 43, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S12", "Attn:P31=14", "Attn:P13=10", "Attn:P28=10", "Attn:P14=5"]}, {"position": 43, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:25", "Impact:A1", "Math.Add:S1234", "Attn:P13=28", "Attn:P28=26", "Attn:P12=5", "Attn:P27=5"]}, {"position": 43, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:75", "Impact:A1", "Math.Add:S01234"]}, {"position": 43, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S1", "Attn:P14=11", "Attn:P27=6", "Attn:P42=6", "Attn:P40=5", "Math.Add:A2.SP"]}, {"position": 43, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:70", "Impact:A1", "Math.Add:S012345"]}, {"position": 44, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:77", "Impact:A0", "Math.Add:S01234", "Attn:P13=40", "Attn:P28=39", "Attn:P12=1", "Attn:P31=1"]}, {"position": 44, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:3", "Impact:A0", "Math.Add:S12", "Attn:P13=8", "Attn:P28=6", "Attn:P26=4", "Attn:P12=4"]}, {"position": 44, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:67", "Impact:A0", "Math.Add:S01234"]}, {"position": 44, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:81", "Impact:A0", "Math.Add:S012345"]}]
add_d14_l2_h3_t60K_s572091/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 25, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A10.ST"]}, {"position": 25, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A10.ST", "Algo:A11.ST"]}, {"position": 25, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 26, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A5.ST"]}, {"position": 26, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.ST", "Algo:A6.ST"]}, {"position": 26, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 26, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 27, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 27, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 28, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A7.ST"]}, {"position": 28, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.ST", "Algo:A9.ST"]}, {"position": 28, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 29, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.ST"]}, {"position": 29, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A12.ST"]}, {"position": 30, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A13.ST"]}, {"position": 30, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 30, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 31, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 31, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 31, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 32, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 32, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 32, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 33, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 33, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 33, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 33, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 34, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 34, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A9.SS", "Algo:A9.SC"]}, {"position": 34, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 34, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 35, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 35, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A8.SS", "Algo:A8.SC"]}, {"position": 35, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 35, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 35, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 36, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 36, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 36, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A7.SS", "Algo:A7.SC"]}, {"position": 36, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 36, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 36, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 36, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 36, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 37, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 37, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 37, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A6.SS", "Algo:A6.SC"]}, {"position": 37, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 37, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 37, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 37, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 37, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 38, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 38, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 38, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.SS", "Algo:A5.SC"]}, {"position": 38, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 38, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 38, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 38, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 39, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 39, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 39, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.SS", "Algo:A4.SC"]}, {"position": 39, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 39, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 39, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 39, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 39, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 40, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 40, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 40, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SS", "Algo:A3.SC"]}, {"position": 40, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 40, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 40, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 40, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 41, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 41, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 41, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SS", "Algo:A2.SC"]}, {"position": 41, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 41, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 42, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 42, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 42, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.SS", "Algo:A1.SC"]}, {"position": 42, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 42, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 43, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 43, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 43, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.SC"]}, {"position": 43, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 43, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 43, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 44, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 44, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 44, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 44, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d14_l2_h3_t60K_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:badcfd775dde32939a39e4759091a090275ee4375dca07d8e09bff1225ae2a99
3
+ size 25202818
add_d14_l2_h3_t60K_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d15_l2_h3_t80K_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b5cbea1bbab77ffefa625c72e1b8ddf61e7663254bd6cc14e630d3d825d6ae3
3
+ size 25209474
add_d15_l2_h3_t80K_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d20_l2_h3_t80K_s572091/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af6304aab42d0d441e259b4b3cd337548ff882f4d1d242992556bb202cc513a9
3
+ size 25243522
add_d20_l2_h3_t80K_s572091/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d5_l1_h3_t15K_s372001/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 12, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:23", "Impact:A5", "Math.Add:S12345", "Attn:P0=51", "Attn:P6=48", "Math.Add:A4.SP"]}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:9", "Impact:A5", "Math.Add:S345", "Attn:P8=29", "Attn:P2=28", "Attn:P3=8", "Attn:P9=8"]}, {"position": 12, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:10", "Impact:A5", "Math.Add:S1245", "Attn:P1=52", "Attn:P7=45", "Math.Add:A3.SP"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:31", "Impact:A5", "Math.Add:S12345"]}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:23", "Impact:A4", "Math.Add:S123", "Attn:P1=52", "Attn:P7=45", "Math.Add:A3.SP"]}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:89", "Impact:A4", "Math.Add:S012345", "Attn:P6=40", "Attn:P0=40", "Attn:P5=5", "Attn:P12=5"]}, {"position": 13, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:17", "Impact:A4", "Math.Add:S2345", "Attn:P8=41", "Attn:P2=37", "Attn:P3=6", "Attn:P9=5"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:65", "Impact:A4", "Math.Add:S0123"]}, {"position": 14, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:26", "Impact:A3", "Math.Add:S123", "Attn:P8=52", "Attn:P2=47", "Math.Add:A2.SP"]}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:89", "Impact:A3", "Math.Add:S01235", "Attn:P1=40", "Attn:P7=40", "Attn:P10=4", "Attn:P4=4"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:14", "Impact:A3", "Math.Add:S2345", "Attn:P3=43", "Attn:P9=37", "Attn:P10=8", "Attn:P4=7"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:67", "Impact:A3", "Math.Add:S0123"]}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:31", "Impact:A2", "Math.Add:S1234", "Attn:P3=56", "Attn:P9=43", "Math.Add:A1.SP"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:74", "Impact:A2", "Math.Add:S01235", "Attn:P2=49", "Attn:P8=49"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:13", "Impact:A2", "Math.Add:S25", "Attn:P4=49", "Attn:P10=45", "Attn:P11=1"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:84", "Impact:A2", "Math.Add:S01234"]}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:22", "Impact:A1", "Math.Add:S1235", "Attn:P4=51", "Attn:P10=47"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:76", "Impact:A1", "Math.Add:S012345", "Attn:P9=49", "Attn:P3=49"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:84", "Impact:A1", "Math.Add:S01234"]}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:83", "Impact:A0", "Math.Add:S012345", "Attn:P4=48", "Attn:P10=48"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:78", "Impact:A0", "Math.Add:S012345"]}]
add_d5_l1_h3_t15K_s372001/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 12, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A4.ST"]}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.ST", "Algo:A2.ST"]}, {"position": 12, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.ST"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.SC"]}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA.A4"]}, {"position": 13, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SS"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.SC"]}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SS"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SC"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.SS"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A0.SC"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}]
add_d5_l1_h3_t15K_s372001/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a50417c0dce442db594e1e316c3bfcb91eebfed99e8da0ff5bfc790ba8d8866b
3
+ size 12625286
add_d5_l1_h3_t15K_s372001/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d5_l1_h3_t30K_s372001/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8558485a5695cb1e584160b91250b7938eaeb11a6e60b0e9c5b888546fb52ffc
3
+ size 12625286
add_d5_l1_h3_t30K_s372001/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d5_l2_h3_t15K_s372001/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 0, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S1", "Attn:P0=100"]}, {"position": 0, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S1"]}, {"position": 6, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:0", "Impact:A5", "Math.Add:S1", "Attn:P6=52", "Attn:P5=21", "Attn:P0=17", "Attn:P1=5"]}, {"position": 6, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:12", "Impact:A5", "Math.Add:S012345", "Attn:P0=83", "Attn:P6=15"]}, {"position": 6, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:8", "Impact:A5", "Math.Add:S1345", "Attn:P6=95", "Attn:P0=2"]}, {"position": 6, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:8", "Impact:A5", "Math.Add:S12345"]}, {"position": 9, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:14", "Impact:A543", "Math.Add:S2345", "Attn:P8=51", "Attn:P2=46"]}, {"position": 9, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:24", "Impact:A543", "Math.Add:S2345"]}, {"position": 10, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:7", "Impact:A54", "Math.Add:S23", "Attn:P8=51", "Attn:P2=48", "Math.Add:A2.SP"]}, {"position": 10, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:6", "Impact:A5432", "Math.Add:S345", "Attn:P9=44", "Attn:P3=43", "Attn:P5=4", "Attn:P4=1"]}, {"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:12", "Impact:A543", "Math.Add:S234", "Attn:P3=50", "Attn:P9=45", "Math.Add:A1.SP"]}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:30", "Impact:A543", "Math.Add:S2345"]}, {"position": 11, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:23", "Impact:A54321", "Math.Add:S12345", "Attn:P4=56", "Attn:P10=44"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:24", "Impact:A54321", "Math.Add:S012345"]}, {"position": 12, "layer": 0, "is_head": true, "num": 0, "tags": ["Math.Add:A3.SP"]}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:3", "Impact:A5", "Math.Add:S15", "Attn:P1=44", "Attn:P7=33", "Attn:P12=7", "Attn:P5=6"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:17", "Impact:A5", "Math.Add:S2345"]}, {"position": 12, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:11", "Impact:A5", "Math.Add:S345", "Attn:P6=29", "Attn:P10=17", "Attn:P12=17", "Attn:P9=13", "Math.Add:A3.SP"]}, {"position": 12, "layer": 1, "is_head": true, "num": 1, "tags": ["Math.Add:A4.SP"]}, {"position": 12, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:11", "Impact:A5", "Math.Add:S12345", "Attn:P6=63", "Attn:P0=36", "Math.Add:A4.SP"]}, {"position": 12, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:13", "Impact:A5", "Math.Add:S012"]}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:22", "Impact:A4", "Math.Add:S123", "Attn:P1=51", "Attn:P7=47", "Math.Add:A3.SP"]}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:56", "Impact:A4", "Math.Add:S01234", "Attn:P0=48", "Attn:P6=47", "Attn:P13=1"]}, {"position": 13, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A4", "Math.Add:S1", "Attn:P6=35", "Attn:P0=32", "Attn:P13=20", "Attn:P5=5"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:53", "Impact:A4", "Math.Add:S01234"]}, {"position": 13, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:5", "Impact:A4", "Math.Add:S234", "Attn:P13=35", "Attn:P10=20", "Attn:P9=16", "Attn:P5=15", "Math.Add:A2.SP"]}, {"position": 13, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:35", "Impact:A4", "Math.Add:S01234"]}, {"position": 14, "layer": 0, "is_head": true, "num": 0, "tags": ["Math.Add:A2.SP"]}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:33", "Impact:A3", "Math.Add:S0123", "Attn:P7=51", "Attn:P1=46"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A3", "Math.Add:S1", "Attn:P1=46", "Attn:P7=39", "Attn:P12=4", "Attn:P13=3"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:59", "Impact:A3", "Math.Add:S0123"]}, {"position": 14, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:2", "Impact:A3", "Math.Add:S34", "Attn:P10=30", "Attn:P9=22", "Attn:P11=20", "Attn:P5=17", "Math.Add:A2.SP"]}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:50", "Impact:A3", "Math.Add:S012345"]}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:17", "Impact:A2", "Math.Add:S1234", "Attn:P3=53", "Attn:P9=46", "Math.Add:A1.SP"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:31", "Impact:A2", "Math.Add:S01234", "Attn:P8=49", "Attn:P2=46"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A2", "Math.Add:S12", "Attn:P2=44", "Attn:P8=36", "Attn:P12=9", "Attn:P15=1"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:61", "Impact:A2", "Math.Add:S0123"]}, {"position": 15, "layer": 1, "is_head": true, "num": 1, "tags": ["Math.Add:A1.SP"]}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:58", "Impact:A2", "Math.Add:S012345"]}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S1", "Attn:P4=51", "Attn:P10=42", "Attn:P13=2", "Attn:P16=1"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:32", "Impact:A1", "Math.Add:S01234", "Attn:P9=50", "Attn:P3=47"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S13", "Attn:P3=44", "Attn:P9=40", "Attn:P12=6", "Attn:P13=1"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:67", "Impact:A1", "Math.Add:S012345"]}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:49", "Impact:A1", "Math.Add:S012345"]}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:44", "Impact:A0", "Math.Add:S012345", "Attn:P10=50", "Attn:P4=47"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A0", "Math.Add:S5", "Attn:P4=41", "Attn:P10=32", "Attn:P12=6", "Attn:P5=4"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:59", "Impact:A0", "Math.Add:S012345"]}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:45", "Impact:A0", "Math.Add:S012345"]}]
add_d5_l2_h3_t15K_s372001/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 0, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 0, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 6, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 6, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 6, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 6, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 9, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.ST"]}, {"position": 9, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 10, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.ST"]}, {"position": 10, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.ST"]}, {"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.ST"]}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 11, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.ST"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 12, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.ST"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 12, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 12, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 12, "layer": 1, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 12, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.SC"]}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA"]}, {"position": 13, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.SA"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 13, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 13, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SA"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SC"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SA"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A0.SC"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.SA"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.SA"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d5_l2_h3_t15K_s372001/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2b399c29209ae9e56294df30fd0b2266653dbe13e32b3f57e15f43c3f09ccd3
3
+ size 25144598
add_d5_l2_h3_t15K_s372001/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d5_l2_h3_t40K_s372001/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 6, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:5", "Impact:A5", "Math.Add:S02345", "Attn:P0=71", "Attn:P6=24", "Attn:P1=1", "Attn:P5=1"]}, {"position": 6, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:6", "Impact:A5", "Math.Add:S12345", "Attn:P6=88", "Attn:P0=6", "Attn:P5=3", "Attn:P1=2", "Math.Add:A4.SP.Weak"]}, {"position": 6, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:3", "Impact:A5", "Math.Add:S12"]}, {"position": 9, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A5", "Math.Add:S5", "Attn:P6=62", "Attn:P0=31", "Attn:P5=1"]}, {"position": 9, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:5", "Impact:A5", "Math.Add:S12345", "Attn:P0=80", "Attn:P6=9", "Attn:P5=3", "Attn:P4=1"]}, {"position": 9, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S14"]}, {"position": 10, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:10", "Impact:A543", "Math.Add:S123", "Attn:P8=50", "Attn:P2=49", "Math.Add:A2.SP.Weak"]}, {"position": 10, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:6", "Impact:A543", "Math.Add:S2345", "Attn:P9=42", "Attn:P3=36", "Attn:P5=6", "Attn:P10=3"]}, {"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:15", "Impact:A543", "Math.Add:S234", "Attn:P3=50", "Attn:P9=45"]}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:32", "Impact:A5432", "Math.Add:S2345"]}, {"position": 11, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:22", "Impact:A54321", "Math.Add:S2345", "Attn:P4=54", "Attn:P10=46"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:24", "Impact:A5432", "Math.Add:S012345"]}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A5", "Math.Add:S0345", "Attn:P1=43", "Attn:P7=40", "Attn:P12=5", "Attn:P5=5"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:10", "Impact:A5", "Math.Add:S2345"]}, {"position": 12, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:15", "Impact:A5", "Math.Add:S345", "Attn:P10=19", "Attn:P0=12", "Attn:P6=12", "Attn:P1=9", "Math.Add:A3.SP.Weak"]}, {"position": 12, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A5", "Math.Add:S1", "Attn:P6=39", "Attn:P9=33", "Attn:P0=26", "Math.Add:A4.SP.Weak"]}, {"position": 12, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:44", "Impact:A5", "Math.Add:S01234"]}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:24", "Impact:A4", "Math.Add:S123", "Attn:P1=52", "Attn:P7=46", "Math.Add:A3.SP.Weak"]}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:88", "Impact:A4", "Math.Add:S012345", "Attn:P0=48", "Attn:P6=47"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:57", "Impact:A4", "Math.Add:S01234"]}, {"position": 13, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A4", "Math.Add:S4", "Attn:P10=38", "Attn:P13=31", "Attn:P5=17", "Attn:P11=12", "Math.Add:A1.SP.Weak", "Math.Add:A2.SP.Weak", "Math.Add:A3.SP.Weak", "Math.Add:A4.SP.Weak"]}, {"position": 13, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:45", "Impact:A4", "Math.Add:S012345"]}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:69", "Impact:A3", "Math.Add:S012345", "Attn:P7=48", "Attn:P1=48"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:60", "Impact:A3", "Math.Add:S012345"]}, {"position": 14, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A3", "Math.Add:S3", "Attn:P10=48", "Attn:P11=20", "Attn:P5=17", "Attn:P14=12", "Math.Add:A1.SP.Weak", "Math.Add:A2.SP.Weak"]}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:54", "Impact:A3", "Math.Add:S012345"]}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:8", "Impact:A2", "Math.Add:S13", "Attn:P3=53", "Attn:P9=45", "Math.Add:A1.SP.Weak"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:71", "Impact:A2", "Math.Add:S012345", "Attn:P2=49", "Attn:P8=47"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:56", "Impact:A2", "Math.Add:S0123"]}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:45", "Impact:A2", "Math.Add:S01234"]}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:3", "Impact:A1", "Math.Add:S123", "Attn:P4=52", "Attn:P10=41", "Attn:P12=2", "Attn:P13=1"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:43", "Impact:A1", "Math.Add:S012345", "Attn:P9=49", "Attn:P3=47"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A1", "Math.Add:S1", "Attn:P3=37", "Attn:P9=35", "Attn:P12=14", "Attn:P5=3", "Math.Add:A1.SP.Weak"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:57", "Impact:A1", "Math.Add:S012345"]}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:42", "Impact:A1", "Math.Add:S01234"]}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:66", "Impact:A0", "Math.Add:S012345", "Attn:P10=48", "Attn:P4=46"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:64", "Impact:A0", "Math.Add:S012345"]}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:66", "Impact:A0", "Math.Add:S012345"]}]
add_d5_l2_h3_t40K_s372001/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 6, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 6, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 6, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 9, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 9, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 9, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 10, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.ST"]}, {"position": 10, "layer": 0, "is_head": true, "num": 1, "tags": []}, {"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.ST"]}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 11, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.ST"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.ST"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 12, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 12, "layer": 1, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 12, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.SC"]}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 13, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 13, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SC"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A0.SC"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": []}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d5_l2_h3_t40K_s372001/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e013f8665d4ce1c418ce25dc4f5da3c875b9f1b53aa2e19edfa0ad94d7b38401
3
+ size 25144084
add_d5_l2_h3_t40K_s372001/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d6_l2_h3_t15K_s372001/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 11, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:12", "Impact:A654", "Math.Add:S1234", "Attn:P10=50", "Attn:P3=49", "Math.Add:A2.SP"]}, {"position": 11, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:14", "Impact:A65432", "Math.Add:S2345", "Attn:P4=56", "Attn:P11=40", "Math.Add:A1.SP"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:16", "Impact:A6543", "Math.Add:S2345"]}, {"position": 12, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:8", "Impact:A654", "Math.Add:S123", "Attn:P2=52", "Attn:P9=43", "Math.Add:A3.SP"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:33", "Impact:A654", "Math.Add:S12345"]}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:3", "Impact:A6543", "Math.Add:S1", "Attn:P4=54", "Attn:P11=45", "Math.Add:A1.SP"]}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:18", "Impact:A54321", "Math.Add:S2345", "Attn:P5=58", "Attn:P12=40", "Math.Add:A0.SP"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:18", "Impact:A54321", "Math.Add:S2345"]}, {"position": 14, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:6", "Impact:A6", "Math.Add:S123", "Attn:P1=52", "Attn:P8=47", "Math.Add:A4.SP"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:28", "Impact:A6", "Math.Add:S12345", "Attn:P0=51", "Attn:P7=48"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:24", "Impact:A6", "Math.Add:S12345"]}, {"position": 14, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:4", "Impact:A6", "Math.Add:S345", "Attn:P11=16", "Attn:P14=15", "Attn:P12=11", "Attn:P6=10", "Math.Add:A4.SP"]}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:6", "Impact:A6", "Math.Add:S2345"]}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:14", "Impact:A5", "Math.Add:S123", "Attn:P1=53", "Attn:P8=45", "Math.Add:A4.SP"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:10", "Impact:A5", "Math.Add:S1234", "Attn:P7=37", "Attn:P0=34", "Attn:P15=20", "Attn:P13=2"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:25", "Impact:A5", "Math.Add:S01234", "Attn:P0=46", "Attn:P7=45", "Attn:P14=1", "Attn:P15=1"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:45", "Impact:A5", "Math.Add:S012345"]}, {"position": 15, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:8", "Impact:A5", "Math.Add:S2345", "Attn:P15=28", "Attn:P11=16", "Attn:P14=13", "Attn:P12=12", "Math.Add:A4.SP"]}, {"position": 15, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:0", "Impact:A5", "Math.Add:S1", "Attn:P15=36", "Attn:P13=23", "Attn:P14=17", "Attn:P12=6", "Math.Add:A4.SP"]}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:27", "Impact:A5", "Math.Add:S012345"]}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:5", "Impact:A4", "Math.Add:S12", "Attn:P2=52", "Attn:P9=44", "Math.Add:A3.SP"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:8", "Impact:A4", "Math.Add:S0123", "Attn:P1=45", "Attn:P8=42", "Attn:P6=3", "Attn:P14=3"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:35", "Impact:A4", "Math.Add:S0123", "Attn:P8=46", "Attn:P1=45", "Attn:P14=1"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:48", "Impact:A4", "Math.Add:S01234"]}, {"position": 16, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:14", "Impact:A4", "Math.Add:S1234", "Attn:P11=32", "Attn:P12=18", "Attn:P16=17", "Attn:P13=9", "Math.Add:A3.SP"]}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:43", "Impact:A4", "Math.Add:S012345"]}, {"position": 17, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:11", "Impact:A3", "Math.Add:S124", "Attn:P10=51", "Attn:P3=48", "Math.Add:A2.SP"]}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:5", "Impact:A3", "Math.Add:S123", "Attn:P9=43", "Attn:P2=42", "Attn:P14=5", "Attn:P6=3"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:30", "Impact:A3", "Math.Add:S0123", "Attn:P2=48", "Attn:P9=47"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:56", "Impact:A3", "Math.Add:S0123"]}, {"position": 17, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A3", "Math.Add:S4", "Attn:P11=31", "Attn:P13=17", "Attn:P17=9", "Attn:P4=8", "Math.Add:A2.SP"]}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:27", "Impact:A3", "Math.Add:S0123"]}, {"position": 18, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:5", "Impact:A2", "Math.Add:S1", "Attn:P4=55", "Attn:P11=43", "Math.Add:A1.SP"]}, {"position": 18, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:8", "Impact:A2", "Math.Add:S01234", "Attn:P3=45", "Attn:P10=44", "Attn:P14=3", "Attn:P6=1"]}, {"position": 18, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:21", "Impact:A2", "Math.Add:S01234", "Attn:P10=48", "Attn:P3=47"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:54", "Impact:A2", "Math.Add:S01234"]}, {"position": 18, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:26", "Impact:A2", "Math.Add:S0123"]}, {"position": 19, "layer": 0, "is_head": true, "num": 0, "tags": ["Fail%:29", "Impact:A1", "Math.Add:S12345", "Attn:P5=53", "Attn:P12=43"]}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:7", "Impact:A1", "Math.Add:S012345", "Attn:P4=45", "Attn:P11=44", "Attn:P14=4", "Attn:P6=2"]}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:27", "Impact:A1", "Math.Add:S012345", "Attn:P11=49", "Attn:P4=48"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:73", "Impact:A1", "Math.Add:S012345"]}, {"position": 19, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:32", "Impact:A1", "Math.Add:S01234"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:8", "Impact:A0", "Math.Add:S12345", "Attn:P5=33", "Attn:P12=30", "Attn:P14=24", "Attn:P6=6"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:43", "Impact:A0", "Math.Add:S012345", "Attn:P12=45", "Attn:P5=45", "Attn:P14=2", "Attn:P6=1"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:81", "Impact:A0", "Math.Add:S012345"]}, {"position": 20, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:55", "Impact:A0", "Math.Add:S012345"]}]
add_d6_l2_h3_t15K_s372001/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 11, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.ST"]}, {"position": 11, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.ST"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 12, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.ST"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 13, "layer": 0, "is_head": true, "num": 0, "tags": []}, {"position": 13, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.ST"]}, {"position": 13, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A4.ST"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.ST"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A4.SC"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.SA"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A5.SA"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 15, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A3.SC"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.SA"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A2.SC"]}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SA"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 18, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A1.SC"]}, {"position": 18, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 18, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SA"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 18, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 19, "layer": 0, "is_head": true, "num": 0, "tags": ["Algo:A0.SC"]}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A1.SA"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 19, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 20, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.SA"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 20, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d6_l2_h3_t15K_s372001/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72fc74eb60b99c58b759b5cd577b0ba6668eef88387b419e7c350b5b92eba7c5
3
+ size 25150806
add_d6_l2_h3_t15K_s372001/training_loss.json ADDED
The diff for this file is too large to render. See raw diff
 
add_d6_l2_h3_t20K_s173289/behaviors.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A65", "Math.Add:S23", "Attn:P2=51", "Attn:P9=44"]}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:2", "Impact:A65", "Math.Add:S35"]}, {"position": 11, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:9", "Impact:A654", "Math.Add:S234", "Attn:P3=49", "Attn:P10=48"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:13", "Impact:A654", "Math.Add:S2345"]}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:12", "Impact:A65432", "Math.Add:S12345", "Attn:P4=52", "Attn:P11=46"]}, {"position": 12, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:17", "Impact:A5432", "Math.Add:S2345", "Attn:P5=48", "Attn:P12=43", "Attn:P6=2", "Attn:P4=1"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:20", "Impact:A5432", "Math.Add:S2345"]}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:26", "Impact:A6", "Math.Add:S12345", "Attn:P0=52", "Attn:P7=47"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A6", "Math.Add:S2", "Attn:P1=48", "Attn:P8=44", "Attn:P13=3"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:20", "Impact:A6", "Math.Add:S12345"]}, {"position": 14, "layer": 1, "is_head": true, "num": 0, "tags": ["Fail%:5", "Impact:A6", "Math.Add:S12", "Attn:P14=69", "Attn:P13=19", "Attn:P0=7", "Attn:P6=1", "Math.Add:A5.SP"]}, {"position": 14, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:0", "Impact:A6", "Math.Add:S5", "Attn:P13=26", "Attn:P10=22", "Attn:P11=14", "Attn:P0=12", "Math.Add:A4.SP"]}, {"position": 14, "layer": 1, "is_head": true, "num": 2, "tags": ["Math.Add:A4.SP"]}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:3", "Impact:A6", "Math.Add:S2345"]}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:36", "Impact:A5", "Math.Add:S0123", "Attn:P0=43", "Attn:P7=43", "Attn:P13=5", "Attn:P6=2"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:33", "Impact:A5", "Math.Add:S1234", "Attn:P1=52", "Attn:P8=42"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:50", "Impact:A5", "Math.Add:S012345"]}, {"position": 15, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S0", "Attn:P15=42", "Attn:P13=26", "Attn:P12=9", "Attn:P10=8", "Math.Add:A4.SP"]}, {"position": 15, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:1", "Impact:A5", "Math.Add:S23", "Attn:P15=50", "Attn:P13=26", "Attn:P11=10", "Attn:P10=10", "Math.Add:A4.SP"]}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:28", "Impact:A5", "Math.Add:S012345"]}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:79", "Impact:A4", "Math.Add:S012345", "Attn:P1=49", "Attn:P8=45"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:7", "Impact:A4", "Math.Add:S12", "Attn:P2=52", "Attn:P9=43"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:62", "Impact:A4", "Math.Add:S012345"]}, {"position": 16, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A4", "Math.Add:S01", "Attn:P16=32", "Attn:P13=24", "Attn:P12=17", "Attn:P11=13", "Math.Add:A3.SP"]}, {"position": 16, "layer": 1, "is_head": true, "num": 2, "tags": ["Fail%:3", "Impact:A4", "Math.Add:S123", "Attn:P16=32", "Attn:P11=26", "Attn:P13=24", "Attn:P10=10", "Math.Add:A3.SP"]}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:49", "Impact:A4", "Math.Add:S012345"]}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:75", "Impact:A3", "Math.Add:S012345", "Attn:P9=48", "Attn:P2=47"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:27", "Impact:A3", "Math.Add:S123", "Attn:P3=48", "Attn:P10=47"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:55", "Impact:A3", "Math.Add:S0123"]}, {"position": 17, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A3", "Math.Add:S2", "Attn:P17=30", "Attn:P13=28", "Attn:P12=25", "Attn:P11=8", "Math.Add:A2.SP"]}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:59", "Impact:A3", "Math.Add:S012345"]}, {"position": 18, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:67", "Impact:A2", "Math.Add:S012345", "Attn:P10=48", "Attn:P3=48"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:62", "Impact:A2", "Math.Add:S01234"]}, {"position": 18, "layer": 1, "is_head": true, "num": 1, "tags": ["Math.Add:A1.SP"]}, {"position": 18, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:51", "Impact:A2", "Math.Add:S012345"]}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:81", "Impact:A1", "Math.Add:S012345", "Attn:P4=49", "Attn:P11=48"]}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Fail%:37", "Impact:A1", "Math.Add:S12345", "Attn:P5=52", "Attn:P12=44"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:63", "Impact:A1", "Math.Add:S01234"]}, {"position": 19, "layer": 1, "is_head": true, "num": 1, "tags": ["Fail%:1", "Impact:A1", "Math.Add:S2", "Attn:P19=58", "Attn:P13=37", "Attn:P12=1"]}, {"position": 19, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:40", "Impact:A1", "Math.Add:S012345"]}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Fail%:76", "Impact:A0", "Math.Add:S012345", "Attn:P5=49", "Attn:P12=48"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": ["Fail%:78", "Impact:A0", "Math.Add:S012345"]}, {"position": 20, "layer": 1, "is_head": false, "num": 0, "tags": ["Fail%:24", "Impact:A0", "Math.Add:S012345"]}]
add_d6_l2_h3_t20K_s173289/features.json ADDED
@@ -0,0 +1 @@
 
 
1
+ [{"position": 10, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.ST"]}, {"position": 10, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 11, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.ST"]}, {"position": 11, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 12, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.ST"]}, {"position": 12, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.ST"]}, {"position": 12, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.ST"]}, {"position": 14, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.ST"]}, {"position": 14, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 14, "layer": 1, "is_head": true, "num": 0, "tags": []}, {"position": 14, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 14, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 14, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A5.SA"]}, {"position": 15, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A4.SC"]}, {"position": 15, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 15, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 15, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 15, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A4.SA"]}, {"position": 16, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A3.SC"]}, {"position": 16, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 16, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 16, "layer": 1, "is_head": true, "num": 2, "tags": []}, {"position": 16, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A3.SA"]}, {"position": 17, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A2.SC"]}, {"position": 17, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 17, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 17, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 18, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A2.SA"]}, {"position": 18, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 18, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 18, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 19, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A1.SA"]}, {"position": 19, "layer": 0, "is_head": true, "num": 2, "tags": ["Algo:A0.SC"]}, {"position": 19, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 19, "layer": 1, "is_head": true, "num": 1, "tags": []}, {"position": 19, "layer": 1, "is_head": false, "num": 0, "tags": []}, {"position": 20, "layer": 0, "is_head": true, "num": 1, "tags": ["Algo:A0.SA"]}, {"position": 20, "layer": 0, "is_head": false, "num": 0, "tags": []}, {"position": 20, "layer": 1, "is_head": false, "num": 0, "tags": []}]
add_d6_l2_h3_t20K_s173289/model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac4685afc7f6e5992c199608935b5a9ee37eb350db149b3ab1d6f9afaa2316cc
3
+ size 25150484
add_d6_l2_h3_t20K_s173289/training_loss.json ADDED
The diff for this file is too large to render. See raw diff