Star217
339
代码
代码
Issues31
Pull Requests101
Actions
项目讨论
Wiki
分析
Star217
339
已合并
[pytorch][bugfix]testcase baseline fix #4015
guihaowen666创建于 1月6日
[pytorch][bugfix]testcase baseline fix #4015
已合并
guihaowen666创建于 1月6日
从guihaowen666/MindSpeed-LLM230br_220_ci_baseline_update合入到Ascend/MindSpeed-LLM2.2.0
共 4 个文件变更+181-181
Mtests/st/baseline_results/deepseek_v3_mcore_tp2_pp2_ep2_dualpipev_fb.json+47-47
@@ -1,54 +1,54 @@
1{1{
2 "lm loss": [2 "lm loss": [
3- 1.122216E+01,3+ 11.22216,
4- 9.771708E+00,4+ 9.771708,
5- 8.739755E+00,5+ 8.739496,
6- 8.649700E+00,6+ 8.649791,
7- 8.669522E+00,7+ 8.668625,
8- 8.266747E+00,8+ 8.265086,
9- 7.786237E+00,9+ 7.786827,
10- 7.639176E+00,10+ 7.639999,
11- 7.439544E+00,11+ 7.439589,
12- 7.343841E+00,12+ 7.344276,
13- 7.340831E+00,13+ 7.341005,
14- 7.366453E+00,14+ 7.366801,
15- 7.281910E+00,15+ 7.281403,
16- 7.234505E+00,16+ 7.23485,
17- 7.379526E+0017+ 7.379908
18 ],18 ],
19 "time info": [19 "time info": [
20- 11393.4,20+ 83526.6,
21- 3351.9,21+ 3752.8,
22- 3342.4,22+ 3757.1,
23- 3330.3,23+ 3733.0,
24- 3362.9,24+ 3764.8,
25- 3350.4,25+ 3769.3,
26- 3347.8,26+ 3730.0,
27- 3351.0,27+ 3732.3,
28- 3354.5,28+ 3737.0,
29- 3337.3,29+ 3754.9,
30- 3348.0,30+ 3748.3,
31- 3349.6,31+ 3772.4,
32- 3355.5,32+ 3766.5,
33- 3350.4,33+ 3730.4,
34- 3358.334+ 3726.0
35 ],35 ],
36 "grad norm": [36 "grad norm": [
37- 32.666,37+ 32.666,
38- 21.849,38+ 21.849,
39- 12.632,39+ 12.633,
40- 11.310,40+ 11.316,
41- 38.861,41+ 38.861,
42- 28.531,42+ 28.539,
43- 21.254,43+ 21.277,
44- 14.467,44+ 14.484,
45- 9.925,45+ 9.915,
46- 8.899,46+ 8.897,
47- 8.224,47+ 8.227,
48- 8.247,48+ 8.232,
49- 7.870,49+ 7.865,
50- 7.501,50+ 7.524,
51- 7.42951+ 7.434
52 ],52 ],
53 "memo info": [53 "memo info": [
54 {54 {
@@ -64,12 +64,12 @@
64 {64 {
65 "rank": 4,65 "rank": 4,
66 "allocated memory": 3951.3623046875,66 "allocated memory": 3951.3623046875,
67- "max allocated memory": 9733.5991210937567+ "max allocated memory": 9733.60009765625
68 },68 },
69 {69 {
70 "rank": 5,70 "rank": 5,
71 "allocated memory": 3951.3623046875,71 "allocated memory": 3951.3623046875,
72- "max allocated memory": 10129.289062572+ "max allocated memory": 10129.2900390625
73 }73 }
74 ]74 ]
75}75}
Mtests/st/baseline_results/llama2_tp2_cp4_general_double_ring.json+59-59
@@ -1,95 +1,95 @@
1{1{
2 "lm loss": [2 "lm loss": [
3- 2.17736,3+ 2.17818,
4- 2.121818,4+ 2.12259,
5- 2.116518,5+ 2.11765,
6- 2.087001,6+ 2.086977,
7- 2.07734,7+ 2.077899,
8- 1.925661,8+ 1.926156,
9- 1.912793,9+ 1.913502,
10- 1.869064,10+ 1.869504,
11- 1.917361,11+ 1.918084,
12- 1.585124,12+ 1.58575,
13- 1.553152,13+ 1.553498,
14- 1.595199,14+ 1.595088,
15- 1.570609,15+ 1.570792,
16- 1.550481,16+ 1.551176,
17- 1.51415317+ 1.514223
18 ],18 ],
19 "grad norm": [19 "grad norm": [
20 28.301,20 28.301,
21 27.05,21 27.05,
22- 26.351,22+ 26.355,
23- 26.57,23+ 26.574,
24- 26.895,24+ 26.893,
25- 23.497,25+ 23.482,
26- 22.75,26+ 22.766,
27- 22.254,27+ 22.25,
28- 23.012,28+ 23.01,
29- 17.275,29+ 17.282,
30- 16.61,30+ 16.612,
31- 17.031,31+ 17.034,
32- 17.574,32+ 17.597,
33- 16.649,33+ 16.657,
34- 16.23934+ 16.228
35 ],35 ],
36 "time info": [36 "time info": [
37- 43396.5,37+ 47282.8,
38- 5191.7,38+ 6383.5,
39- 5175.4,39+ 5787.4,
40- 5240.7,40+ 5864.6,
41- 5161.1,41+ 5766.5,
42- 5194.3,42+ 5783.8,
43- 5077.1,43+ 5808.4,
44- 5153.1,44+ 5822.4,
45- 5244.8,45+ 5866.3,
46- 5221.0,46+ 5756.9,
47- 5175.4,47+ 5707.2,
48- 5287.4,48+ 5734.6,
49- 5214.0,49+ 5763.9,
50- 5275.8,50+ 5676.4,
51- 5223.351+ 5685.7
52 ],52 ],
53 "memo info": [53 "memo info": [
54 {54 {
55 "rank": 0,55 "rank": 0,
56- "allocated memory": 27320.79052734375,56+ "allocated memory": 27320.7919921875,
57- "max allocated memory": 27511.29687557+ "max allocated memory": 27511.29833984375
58 },58 },
59 {59 {
60 "rank": 1,60 "rank": 1,
61- "allocated memory": 27320.79052734375,61+ "allocated memory": 27320.7919921875,
62- "max allocated memory": 27511.29687562+ "max allocated memory": 27511.29833984375
63 },63 },
64 {64 {
65 "rank": 2,65 "rank": 2,
66- "allocated memory": 27320.79248046875,66+ "allocated memory": 27320.7939453125,
67- "max allocated memory": 27511.29882812567+ "max allocated memory": 27511.30029296875
68 },68 },
69 {69 {
70 "rank": 3,70 "rank": 3,
71- "allocated memory": 27320.79248046875,71+ "allocated memory": 27320.7939453125,
72- "max allocated memory": 27511.29882812572+ "max allocated memory": 27511.30029296875
73 },73 },
74 {74 {
75 "rank": 4,75 "rank": 4,
76- "allocated memory": 27320.81884765625,76+ "allocated memory": 27320.8203125,
77- "max allocated memory": 27511.325195312577+ "max allocated memory": 27511.32666015625
78 },78 },
79 {79 {
80 "rank": 5,80 "rank": 5,
81- "allocated memory": 27320.81884765625,81+ "allocated memory": 27320.8203125,
82- "max allocated memory": 27511.325195312582+ "max allocated memory": 27511.32666015625
83 },83 },
84 {84 {
85 "rank": 6,85 "rank": 6,
86- "allocated memory": 27320.75830078125,86+ "allocated memory": 27320.759765625,
87- "max allocated memory": 27511.264648437587+ "max allocated memory": 27511.26611328125
88 },88 },
89 {89 {
90 "rank": 7,90 "rank": 7,
91- "allocated memory": 27320.75830078125,91+ "allocated memory": 27320.759765625,
92- "max allocated memory": 27511.264648437592+ "max allocated memory": 27511.26611328125
93 }93 }
94 ]94 ]
95}95}
Mtests/st/baseline_results/mixtral_mcore_tp4_cp2_ep2_ptd.json+46-46
@@ -1,78 +1,78 @@
1{1{
2 "lm loss": [2 "lm loss": [
3- 13.77022,3+ 13.77016,
4- 13.7421,4+ 13.74074,
5- 12.95115,5+ 12.95035,
6- 10.88439,6+ 10.88294,
7- 8.3833,7+ 8.384425,
8- 7.446447,8+ 7.442626,
9- 7.01274,9+ 7.010364,
10- 6.685776,10+ 6.682901,
11- 6.197952,11+ 6.19448,
12- 6.209463,12+ 6.211535,
13- 5.967876,13+ 5.972576,
14- 5.777855,14+ 5.772698,
15- 5.748918,15+ 5.746556,
16- 5.516537,16+ 5.517173,
17- 5.49463617+ 5.493339
18 ],18 ],
19 "time info": [19 "time info": [
20- 8920.6,20+ 16808.8,
21- 3040.8,21+ 3286.3,
22- 2956.6,22+ 3210.8,
23- 2977.2,23+ 3294.0,
24- 3009.9,24+ 3257.5,
25- 3015.0,25+ 3344.4,
26- 3075.9,26+ 3294.8,
27- 2987.4,27+ 3221.5,
28- 3050.9,28+ 3189.0,
29- 3012.0,29+ 3202.0,
30- 3020.7,30+ 3270.8,
31- 2998.3,31+ 3295.6,
32- 3007.2,32+ 3188.1,
33- 3142.3,33+ 3147.9,
34- 3095.934+ 3182.2
35 ],35 ],
36 "memo info": [36 "memo info": [
37 {37 {
38 "rank": 0,38 "rank": 0,
39- "allocated memory": 19643.443359375,39+ "allocated memory": 19660.0048828125,
40- "max allocated memory": 19755.4506835937540+ "max allocated memory": 19676.01318359375
41 },41 },
42 {42 {
43 "rank": 1,43 "rank": 1,
44- "allocated memory": 19643.443359375,44+ "allocated memory": 19660.0048828125,
45- "max allocated memory": 19755.4506835937545+ "max allocated memory": 19676.01318359375
46 },46 },
47 {47 {
48 "rank": 2,48 "rank": 2,
49- "allocated memory": 19643.443359375,49+ "allocated memory": 19660.0048828125,
50- "max allocated memory": 19755.4506835937550+ "max allocated memory": 19676.01318359375
51 },51 },
52 {52 {
53 "rank": 3,53 "rank": 3,
54- "allocated memory": 19643.443359375,54+ "allocated memory": 19660.0048828125,
55- "max allocated memory": 19755.4506835937555+ "max allocated memory": 19676.01318359375
56 },56 },
57 {57 {
58 "rank": 4,58 "rank": 4,
59- "allocated memory": 19643.43994140625,59+ "allocated memory": 19660.00341796875,
60- "max allocated memory": 19755.44726562560+ "max allocated memory": 19676.01171875
61 },61 },
62 {62 {
63 "rank": 5,63 "rank": 5,
64- "allocated memory": 19643.43994140625,64+ "allocated memory": 19660.00341796875,
65- "max allocated memory": 19755.44726562565+ "max allocated memory": 19676.01171875
66 },66 },
67 {67 {
68 "rank": 6,68 "rank": 6,
69- "allocated memory": 19643.43994140625,69+ "allocated memory": 19660.00341796875,
70- "max allocated memory": 19755.44726562570+ "max allocated memory": 19676.01171875
71 },71 },
72 {72 {
73 "rank": 7,73 "rank": 7,
74- "allocated memory": 19643.43994140625,74+ "allocated memory": 19660.00341796875,
75- "max allocated memory": 19755.44726562575+ "max allocated memory": 19676.01171875
76 }76 }
77 ]77 ]
78}78}
Mtests/st/baseline_results/tune_llama2_tp1_pp1_qlora_ptd.json+29-29
@@ -2,41 +2,41 @@
2 "lm loss": [2 "lm loss": [
3 1.305625,3 1.305625,
4 1.015415,4 1.015415,
5- 1.814439,5+ 1.816346,
6- 1.296094,6+ 1.297803,
7- 1.434726,7+ 1.433559,
8- 0.8110601,8+ 0.811111,
9- 1.242749,9+ 1.242873,
10- 0.7087235,10+ 0.7078964,
11- 1.05725,11+ 1.056498,
12- 1.639907,12+ 1.625611,
13- 1.450925,13+ 1.449872,
14- 1.281222,14+ 1.279048,
15- 1.407873,15+ 1.407222,
16- 1.470815,16+ 1.4698,
17- 1.14141717+ 1.146465
18 ],18 ],
19 "time info": [19 "time info": [
20- 2240.9,20+ 2338.9,
21- 497.9,21+ 575.3,
22- 495.9,22+ 562.3,
23- 483.1,23+ 544.6,
24- 491.1,24+ 553.5,
25- 508.9,25+ 580.6,
26- 509.9,26+ 575.1,
27- 495.9,27+ 570.3,
28- 495.6,28+ 559.9,
29- 506.3,29+ 573.2,
30- 499.8,30+ 569.8,
31- 504.5,31+ 561.2,
32- 492.3,32+ 549.7,
33- 507.5,33+ 559.6,
34- 490.834+ 557.6
35 ],35 ],
36 "memo info": [36 "memo info": [
37 {37 {
38 "rank": 0,38 "rank": 0,
39- "allocated memory": 4360.5517578125,39+ "allocated memory": 4360.55224609375,
40 "max allocated memory": 6036.45117187540 "max allocated memory": 6036.451171875
41 }41 }
42 ]42 ]