[nn.Graph] Fix eager tensor not in job bug

[BBuf]

If eager free tensor not in nn.Graph's job, such as num_batches_tracked parameter in bn, the bug happened in compile!

with the help of xuxiaoyu, I have fixed this by add a map find check for variable in job.

[github-actions]

CI failed when running job: Build cu102_xla. PR label automerge has been removed

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8114/

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8114/

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 129.2ms (= 12924.0ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 142.5ms (= 14248.4ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.10 (= 142.5ms / 129.2ms)

OneFlow resnet50 time: 82.5ms (= 8254.3ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 86.1ms (= 8612.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.04 (= 86.1ms / 82.5ms)

OneFlow resnet50 time: 55.3ms (= 11050.2ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 57.8ms (= 11553.6ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.05 (= 57.8ms / 55.3ms)

OneFlow resnet50 time: 43.1ms (= 8611.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.6ms (= 8712.2ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.01 (= 43.6ms / 43.1ms)

OneFlow resnet50 time: 37.7ms (= 7546.4ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.4ms (= 7675.6ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.02 (= 38.4ms / 37.7ms)

OneFlow swin dataloader time: 0.262s (= 52.386s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.814s / 200, num_workers=1)
Relative speed: 0.569 (= 0.149s / 0.262s)

OneFlow swin dataloader time: 0.068s (= 13.614s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 7.978s / 200, num_workers=4)
Relative speed: 0.586 (= 0.040s / 0.068s)

OneFlow swin dataloader time: 0.060s (= 11.943s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.378s / 200, num_workers=8)
Relative speed: 0.367 (= 0.022s / 0.060s)

❌ OneFlow resnet50 time: 144.7ms (= 14473.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 169.5ms (= 16951.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 169.5ms / 144.7ms)

OneFlow resnet50 time: 100.8ms (= 10083.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 122.5ms (= 12253.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 122.5ms / 100.8ms)

OneFlow resnet50 time: 74.9ms (= 14987.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 88.4ms (= 17682.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 88.4ms / 74.9ms)

OneFlow resnet50 time: 65.0ms (= 12996.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 74.6ms (= 14928.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.15 (= 74.6ms / 65.0ms)

OneFlow resnet50 time: 57.4ms (= 11478.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.2ms (= 15045.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 75.2ms / 57.4ms)