Show code cell content
Hide code cell content
from functools import partial
import numpy as onp
import jax.numpy as jnp
import jax
import optax
import matplotlib.pyplot as plt
from chemtrain import quantity, trainers, ensemble
/home/docs/checkouts/readthedocs.org/user_builds/chemtrain/envs/latest/lib/python3.11/site-packages/chemtrain/__init__.py:35: RuntimeWarning: JAX default matmul precision is not set. For float32 model training, evaluation, and deployment export, consider setting JAX_DEFAULT_MATMUL_PRECISION=highest or jax.config.update('jax_default_matmul_precision', 'highest') for more reproducible matmul/contraction behavior.
_warn_if_default_matmul_precision_unset()
Differentiable Trajectory Reweighting (DiffTRe)#
Concepts#
Molecular dynamics is an efficient method to sample positions $\mathbf r$ from a molecular systems. However, back-propagating gradients through the simulation is costly and prone to exploding gradients [1]. Instead, DiffTRe [2] estimates gradients of ensemble averages by employing a probabilistic perspective like in the Umbrella Sampling method [3].
In the umbrella sampling method, biasing a potential enables an efficient computation of ensemble averages of an unbiased potential. A postprocessing step then corrects the effect of the bias on the ensemble average. This correction re-weights the collected samples by accounting for the change in relative probability between the biased reference and unbiased target distribution.
For a trainable target potential $U_\theta$ and samples $x^{(i)}$ from the biased reference potential $\tilde U$, the discretized reweighting reads
\[\langle a \rangle_{U_\theta} \approx \sum_{i=1}^N w^{(i)}a\left(x^{(i)}\right), \quad w^{(i)} = \frac{e^{-\beta\left(U_\theta(x^{(i)}) - \tilde U(x^{(i)})\right)}}{\sum_{j=1}^Ne^{-\beta\left(U_\theta(x^{(j)}) - \tilde U(x^{(j)})\right)}}\]
Since the reference and target potential are independent, DiffTRe assumes that the reference potential $\tilde U$ no longer depend on the learnable parameters. Therefore, also the samples $x^{(i)}$ are independent of the target potential. Hence, the only contribution to the gradients arises through the weights $w^{(i)}$ and the instantaneous states $a(x)$ of the observables. Thus, by employing this umbrella sampling procedure, DiffTRe can compute gradients of the loss function without differentiating through the costly molecular dynamics simulation.
Unfortunately, the statistical error of the approximation grows exponentially fast with the difference between the target and the reference potential [4]. If the number of effective samples[5]
\[N_{eff} = e^{-\sum_{i=1}^N w^{(i)}\log w^{(i)}}\]
decreases below a threshold, DiffTRe replaces the reference potential by the current potential $\tilde U \leftarrow U_\theta$ and resamples the conformations $x^{(i)}$.
To accelerate the resampling and use the full computational capabilities of a GPU, DiffTRe enables to sample from multiple simulations in parallel, using the vectorization capabilities of JAX. Additionally, DiffTRe provides the option to re-seed these parallel simulations by choosing initial states from the simulated trajectory, with probability corresponding to their weight. In principle, this resampling of initial states enables a faster convergence of the simulated trajectories to the new equilibrium distribution.
Toy Example#
For a canonical system of two-particles connected by a spring, the Boltzmann factor is
\[ \rho(\mathbf{r}) \propto e^{-\frac{1}{2}b(||\mathbf{r}_1 - \mathbf{r}_2|| - r_0)^2}, \quad b = \beta\cdot b_S,\]
with temperature dependent effective spring constant $b$.
Hence, the probability of finding the two identical particles in a distance of $r$ is
\[ p(r) = \sqrt{\frac{2b}{\pi}}\left(\frac{b}{1 + br_0^2}\right)r^{2}e^{-\frac{1}{2}b(r - r_0)^2}.\]
The term $r^2$ in front of the exponential factor emerges in the transformation from cartesian to spherical coordinates. With this probability distribution, we can directly compute the radial distribution function.
\[ g(r) = \frac{V}{4 \pi r^2 N^2}p(r).\]
box = 1.0
def radial_distribution(r, r_0=0.35, b=250.0, kbt=2.56):
b = b / kbt
norm = onp.sqrt(onp.pi / (2 * b)) * (1 + b * r_0 ** 2) / b
g_r = box ** 3 / (16 * onp.pi) * onp.exp(-0.5 * b * (r - r_0) ** 2) / norm
return g_r
We now want to learn the parameters of this harmonic bond based on a reference radial distribution function.
r = onp.linspace(0.0, box, 100)
target = onp.vstack((r, radial_distribution(r))).T
Although we could find an analytic relation to the potential parameters, this is not possible for more complex systems. Thus, we need to simulate this relation and set up a model of the system.
Thus, we first need to define an appropriate potential model.
from jax_md import energy, space, simulate, partition
def energy_fn_template(params):
energy_fn = energy.simple_spring_bond(
displacement_fn,
jnp.asarray([[0, 1]]),
length=params["r_0"],
epsilon=100 * params["scaled_b"],
alpha=2.0
)
return energy_fn
init_params = {"r_0": 0.3, "scaled_b": 1.5}
Secondly, we need a routine to simulate the positions of the particles.
r_init = jnp.asarray([[0.0, 0.0, 0.0], [0.11, 0.09, 0.12]])
displacement_fn, shift_fn = space.periodic_general(box)
dt = 0.01
timings = ensemble.sampling.process_printouts(dt, 1100, 100, 1.0)
simulator_template = partial(
simulate.nvt_langevin, shift_fn=shift_fn,
dt=dt, kT=2.56, gamma=0.5, mass=10.0)
neighbor_fn = partition.neighbor_list(displacement_fn, box, 0.5)
simulator_init, _ = simulator_template(energy_fn_template(init_params))
simulator_init_state = simulator_init(jax.random.PRNGKey(0), r_init)
nbrs_init = neighbor_fn.allocate(r_init)
reference_state = ensemble.sampling.SimulatorState(
sim_state=simulator_init_state, nbrs=nbrs_init)
system = {
'displacement_fn': displacement_fn,
'reference_box': box
}
/home/docs/checkouts/readthedocs.org/user_builds/chemtrain/envs/latest/lib/python3.11/site-packages/jax/_src/numpy/reductions.py:230: UserWarning: Explicitly requested dtype float64 requested in sum is not available, and will be truncated to dtype float32. To enable more dtypes, set the jax_enable_x64 configuration option or the JAX_ENABLE_X64 shell environment variable. See https://github.com/jax-ml/jax#current-gotchas for more.
return _reduction(a, "sum", lax.add, 0, preproc=_cast_to_numeric,
/home/docs/checkouts/readthedocs.org/user_builds/chemtrain/envs/latest/lib/python3.11/site-packages/jax/_src/numpy/reductions.py:161: UserWarning: Explicitly requested dtype float64 requested in convert_element_type is not available, and will be truncated to dtype float32. To enable more dtypes, set the jax_enable_x64 configuration option or the JAX_ENABLE_X64 shell environment variable. See https://github.com/jax-ml/jax#current-gotchas for more.
return lax.convert_element_type(result, dtype or result_dtype)
There are multiple classical approaches that enable the inversion of a radial distribution function into a pair-potential. However, they are not applicable to general models, e.g., neural networks. Thus, DiffTRe enables gradient based training, which we are going to set up in the next step.
import optax
lr_schedule = optax.exponential_decay(-0.05, 300, 0.1)
optimizer = optax.chain(
optax.scale_by_rms(0.9),
optax.scale_by_schedule(lr_schedule)
)
Finally, we have to specify the training targets, which is in our case the radial distribution function. Since we only have two particles in a box, we approximate the distribution with slightly coarser bins.
target_builder = quantity.targets.TargetBuilder()
target_builder['rdf'] = quantity.targets.init_radial_distribution_target(
target, rdf_start=0.00, rdf_cut=1.0, nbins=50)
targets, compute_fns = target_builder.build(system)
We now created a numerical representation of the system and can run the trainer.
trainer = trainers.Difftre(
init_params, optimizer, reweight_ratio=0.99
)
trainer.add_statepoint(
energy_fn_template, simulator_template, neighbor_fn, timings,
{'kT': 2.56}, compute_fns, reference_state, targets=targets)
/home/docs/checkouts/readthedocs.org/user_builds/chemtrain/envs/latest/lib/python3.11/site-packages/chemtrain/ensemble/reweighting.py:842: UserWarning: Propagation function is not safe by default. Do not forget to use the wrapper around the compute function to ensure that the neighborlist does not overflow.
warnings.warn(
[Propagation] Time for trajectory compilation 0: 0.031613147258758544 mins
[Propagation] Time for trajectory simulation 0: 9.202957153320313e-06 mins
trainer.train(300)
Show code cell output
Hide code cell output
[Propagate] Effective sample size: 1000.0020141601562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.691 ref_kT = 2.560
Predicted entropy: 8.992245170702517e-08
Predicted free_energy: 0.0
[DiffTRe] Epoch 0
Epoch loss = 0.07805
Gradient norm: 8.621106147766113
Elapsed time = 0.056 min
[Propagate] Effective sample size: 640.3593139648438 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.491 ref_kT = 2.560
Predicted entropy: 0.0020977535750716925
Predicted free_energy: -1.201110601425171
[DiffTRe] Epoch 1
Epoch loss = 0.07289
Gradient norm: 0.9898247718811035
Elapsed time = 0.048 min
[Propagate] Effective sample size: 947.9994506835938 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.536 ref_kT = 2.560
Predicted entropy: 0.0019176447531208396
Predicted free_energy: -1.005387783050537
[DiffTRe] Epoch 2
Epoch loss = 0.03520
Gradient norm: 0.5269869565963745
Elapsed time = 0.003 min
[Propagate] Effective sample size: 970.5472412109375 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.633 ref_kT = 2.560
Predicted entropy: 0.0013965637190267444
Predicted free_energy: -0.6735400557518005
[DiffTRe] Epoch 3
Epoch loss = 0.01555
Gradient norm: 0.16415472328662872
Elapsed time = 0.003 min
[Propagate] Effective sample size: 987.6515502929688 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.601 ref_kT = 2.560
Predicted entropy: 0.000888149777892977
Predicted free_energy: -0.4043285846710205
[DiffTRe] Epoch 4
Epoch loss = 0.01190
Gradient norm: 0.1767328530550003
Elapsed time = 0.003 min
[Propagate] Effective sample size: 989.0111694335938 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.511 ref_kT = 2.560
Predicted entropy: 0.0012246861588209867
Predicted free_energy: -0.5941545963287354
[DiffTRe] Epoch 5
Epoch loss = 0.00985
Gradient norm: 0.07393622398376465
Elapsed time = 0.003 min
[Propagate] Effective sample size: 992.9175415039062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.511 ref_kT = 2.560
Predicted entropy: -0.10878604650497437
Predicted free_energy: -0.40845024585723877
[DiffTRe] Epoch 6
Epoch loss = 0.00789
Gradient norm: 0.022706778720021248
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.7485961914062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.511 ref_kT = 2.560
Predicted entropy: -0.07591664791107178
Predicted free_energy: -0.470764696598053
[DiffTRe] Epoch 7
Epoch loss = 0.00706
Gradient norm: 0.014603657647967339
Elapsed time = 0.000 min
[Propagate] Effective sample size: 991.6792602539062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.511 ref_kT = 2.560
Predicted entropy: -0.1412149965763092
Predicted free_energy: -0.36615410447120667
[DiffTRe] Epoch 8
Epoch loss = 0.00636
Gradient norm: 0.011066979728639126
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.6892700195312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.511 ref_kT = 2.560
Predicted entropy: -0.11646749824285507
Predicted free_energy: -0.4125664532184601
[DiffTRe] Epoch 9
Epoch loss = 0.00585
Gradient norm: 0.011576979421079159
Elapsed time = 0.000 min
[Propagate] Effective sample size: 988.5787963867188 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: 0.0006274547195062041
Predicted free_energy: -0.3118324279785156
[DiffTRe] Epoch 10
Epoch loss = 0.00286
Gradient norm: 0.00019381963647902012
Elapsed time = 0.003 min
[Propagate] Effective sample size: 999.9266357421875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: -0.016180144622921944
Predicted free_energy: -0.2882252633571625
[DiffTRe] Epoch 11
Epoch loss = 0.00269
Gradient norm: 0.0003098873421549797
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9209594726562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: -0.017969690263271332
Predicted free_energy: -0.28795668482780457
[DiffTRe] Epoch 12
Epoch loss = 0.00252
Gradient norm: 0.0002821967354975641
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6835327148438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: -0.037052493542432785
Predicted free_energy: -0.2609376609325409
[DiffTRe] Epoch 13
Epoch loss = 0.00237
Gradient norm: 0.0006022539455443621
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7383422851562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: -0.03277299553155899
Predicted free_energy: -0.2698744237422943
[DiffTRe] Epoch 14
Epoch loss = 0.00224
Gradient norm: 0.0009683975367806852
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.1292724609375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: -0.062017399817705154
Predicted free_energy: -0.22744034230709076
[DiffTRe] Epoch 15
Epoch loss = 0.00214
Gradient norm: 0.0027484458405524492
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5138549804688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: -0.03674651309847832
Predicted free_energy: -0.268230140209198
[DiffTRe] Epoch 16
Epoch loss = 0.00210
Gradient norm: 0.007267693057656288
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.2159423828125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: -0.1064683198928833
Predicted free_energy: -0.16483169794082642
[DiffTRe] Epoch 17
Epoch loss = 0.00230
Gradient norm: 0.02840416319668293
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.4641723632812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.617 ref_kT = 2.560
Predicted entropy: 0.005788491107523441
Predicted free_energy: -0.3379957675933838
[DiffTRe] Epoch 18
Epoch loss = 0.00361
Gradient norm: 0.09553550183773041
Elapsed time = 0.000 min
[Propagate] Effective sample size: 982.2827758789062 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.574 ref_kT = 2.560
Predicted entropy: -0.0002031462936429307
Predicted free_energy: 0.03062543272972107
[DiffTRe] Epoch 19
Epoch loss = 0.01751
Gradient norm: 1.5513288974761963
Elapsed time = 0.003 min
[Propagate] Effective sample size: 753.0906372070312 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.602 ref_kT = 2.560
Predicted entropy: 0.0016253252979367971
Predicted free_energy: -0.8161637783050537
[DiffTRe] Epoch 20
Epoch loss = 0.05807
Gradient norm: 1.0967507362365723
Elapsed time = 0.003 min
[Propagate] Effective sample size: 868.0507202148438 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.721 ref_kT = 2.560
Predicted entropy: 0.0004624005814548582
Predicted free_energy: -0.179671049118042
[DiffTRe] Epoch 21
Epoch loss = 0.00415
Gradient norm: 0.07377838343381882
Elapsed time = 0.003 min
[Propagate] Effective sample size: 987.45751953125 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.629 ref_kT = 2.560
Predicted entropy: -3.5683591704582796e-05
Predicted free_energy: 0.06777669489383698
[DiffTRe] Epoch 22
Epoch loss = 0.01521
Gradient norm: 1.064697027206421
Elapsed time = 0.003 min
[Propagate] Effective sample size: 875.4130859375 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.665 ref_kT = 2.560
Predicted entropy: 0.0010542700765654445
Predicted free_energy: -0.5359545946121216
[DiffTRe] Epoch 23
Epoch loss = 0.03150
Gradient norm: 0.8854995369911194
Elapsed time = 0.003 min
[Propagate] Effective sample size: 908.1356201171875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: 1.066225922841113e-05
Predicted free_energy: 0.019085168838500977
[DiffTRe] Epoch 24
Epoch loss = 0.00293
Gradient norm: 0.0027430204208940268
Elapsed time = 0.003 min
[Propagate] Effective sample size: 999.6277465820312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.030828965827822685
Predicted free_energy: 0.0659528449177742
[DiffTRe] Epoch 25
Epoch loss = 0.00290
Gradient norm: 0.0018395337974652648
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9762573242188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.009884165599942207
Predicted free_energy: 0.03234932944178581
[DiffTRe] Epoch 26
Epoch loss = 0.00289
Gradient norm: 0.0013686713064089417
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6158447265625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.0343148335814476
Predicted free_energy: 0.06924141943454742
[DiffTRe] Epoch 27
Epoch loss = 0.00288
Gradient norm: 0.001394977211020887
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9600219726562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.014260992407798767
Predicted free_energy: 0.037143029272556305
[DiffTRe] Epoch 28
Epoch loss = 0.00287
Gradient norm: 0.0015060622245073318
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4852294921875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.04175495356321335
Predicted free_energy: 0.07879464328289032
[DiffTRe] Epoch 29
Epoch loss = 0.00288
Gradient norm: 0.0021828787866979837
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9657592773438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.01302536390721798
Predicted free_energy: 0.033383265137672424
[DiffTRe] Epoch 30
Epoch loss = 0.00289
Gradient norm: 0.0033526478800922632
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.077392578125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.05613088980317116
Predicted free_energy: 0.09919624775648117
[DiffTRe] Epoch 31
Epoch loss = 0.00294
Gradient norm: 0.006707399617880583
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.86328125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: 0.001396532286889851
Predicted free_energy: 0.009352503344416618
[DiffTRe] Epoch 32
Epoch loss = 0.00308
Gradient norm: 0.013904303312301636
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.3585815429688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: -0.09149900078773499
Predicted free_energy: 0.15199288725852966
[DiffTRe] Epoch 33
Epoch loss = 0.00343
Gradient norm: 0.038520704954862595
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.3267211914062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.579 ref_kT = 2.560
Predicted entropy: 0.059604063630104065
Predicted free_energy: -0.08269461244344711
[DiffTRe] Epoch 34
Epoch loss = 0.00471
Gradient norm: 0.09265776723623276
Elapsed time = 0.000 min
[Propagate] Effective sample size: 987.662841796875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.620 ref_kT = 2.560
Predicted entropy: -0.0005918648093938828
Predicted free_energy: 0.2796662449836731
[DiffTRe] Epoch 35
Epoch loss = 0.02231
Gradient norm: 2.20186448097229
Elapsed time = 0.003 min
[Propagate] Effective sample size: 742.880859375 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.627 ref_kT = 2.560
Predicted entropy: 0.00119204877410084
Predicted free_energy: -0.5685192942619324
[DiffTRe] Epoch 36
Epoch loss = 0.04553
Gradient norm: 1.2412208318710327
Elapsed time = 0.003 min
[Propagate] Effective sample size: 876.4418334960938 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.650 ref_kT = 2.560
Predicted entropy: 0.0001087093842215836
Predicted free_energy: -0.012865245342254639
[DiffTRe] Epoch 37
Epoch loss = 0.00313
Gradient norm: 0.13507471978664398
Elapsed time = 0.003 min
[Propagate] Effective sample size: 985.4039306640625 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.434 ref_kT = 2.560
Predicted entropy: -0.0004939059144817293
Predicted free_energy: 0.24780702590942383
[DiffTRe] Epoch 38
Epoch loss = 0.00691
Gradient norm: 0.40985122323036194
Elapsed time = 0.003 min
[Propagate] Effective sample size: 954.9569091796875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.568 ref_kT = 2.560
Predicted entropy: 0.00043080656905658543
Predicted free_energy: -0.1657293736934662
[DiffTRe] Epoch 39
Epoch loss = 0.00814
Gradient norm: 0.275419682264328
Elapsed time = 0.003 min
[Propagate] Effective sample size: 972.8927001953125 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.578 ref_kT = 2.560
Predicted entropy: -0.00019904102373402566
Predicted free_energy: 0.1501023769378662
[DiffTRe] Epoch 40
Epoch loss = 0.01315
Gradient norm: 0.7315949201583862
Elapsed time = 0.003 min
[Propagate] Effective sample size: 930.8846435546875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.642 ref_kT = 2.560
Predicted entropy: 0.0005503186257556081
Predicted free_energy: -0.24011442065238953
[DiffTRe] Epoch 41
Epoch loss = 0.02646
Gradient norm: 0.19472871720790863
Elapsed time = 0.003 min
[Propagate] Effective sample size: 980.189453125 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.597 ref_kT = 2.560
Predicted entropy: 0.0007235967786982656
Predicted free_energy: -0.35251444578170776
[DiffTRe] Epoch 42
Epoch loss = 0.03731
Gradient norm: 1.020520567893982
Elapsed time = 0.003 min
[Propagate] Effective sample size: 905.8434448242188 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.00011519400140969083
Predicted free_energy: 0.06595158576965332
[DiffTRe] Epoch 43
Epoch loss = 0.00241
Gradient norm: 0.0027337155770510435
Elapsed time = 0.003 min
[Propagate] Effective sample size: 999.68115234375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.022205200046300888
Predicted free_energy: 0.09788517653942108
[DiffTRe] Epoch 44
Epoch loss = 0.00238
Gradient norm: 0.002339037600904703
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9910278320312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.006385511718690395
Predicted free_energy: 0.07364323735237122
[DiffTRe] Epoch 45
Epoch loss = 0.00236
Gradient norm: 0.0025129092391580343
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.569580078125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.029556825757026672
Predicted free_energy: 0.10708073526620865
[DiffTRe] Epoch 46
Epoch loss = 0.00234
Gradient norm: 0.003133052261546254
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9805297851562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.008685192093253136
Predicted free_energy: 0.07542912662029266
[DiffTRe] Epoch 47
Epoch loss = 0.00234
Gradient norm: 0.004663305357098579
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.2098388671875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.042015742510557175
Predicted free_energy: 0.12374821305274963
[DiffTRe] Epoch 48
Epoch loss = 0.00237
Gradient norm: 0.00829609576612711
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8689575195312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.0025867889635264874
Predicted free_energy: 0.06479313969612122
[DiffTRe] Epoch 49
Epoch loss = 0.00248
Gradient norm: 0.016324812546372414
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.7999877929688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.06805668771266937
Predicted free_energy: 0.16017766296863556
[DiffTRe] Epoch 50
Epoch loss = 0.00275
Gradient norm: 0.041078001260757446
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.0393676757812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: 0.03013334609568119
Predicted free_energy: 0.01457951683551073
[DiffTRe] Epoch 51
Epoch loss = 0.00373
Gradient norm: 0.09475696086883545
Elapsed time = 0.000 min
[Propagate] Effective sample size: 990.4146728515625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.589 ref_kT = 2.560
Predicted entropy: -0.13411207497119904
Predicted free_energy: 0.25410953164100647
[DiffTRe] Epoch 52
Epoch loss = 0.00584
Gradient norm: 0.3241754472255707
Elapsed time = 0.000 min
[Propagate] Effective sample size: 977.3443603515625 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.647 ref_kT = 2.560
Predicted entropy: 0.00031357500120066106
Predicted free_energy: -0.14945155382156372
[DiffTRe] Epoch 53
Epoch loss = 0.01810
Gradient norm: 0.6942555904388428
Elapsed time = 0.003 min
[Propagate] Effective sample size: 901.228271484375 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.552 ref_kT = 2.560
Predicted entropy: -0.0009643760859034956
Predicted free_energy: 0.4235854744911194
[DiffTRe] Epoch 54
Epoch loss = 0.00931
Gradient norm: 0.6122599244117737
Elapsed time = 0.003 min
[Propagate] Effective sample size: 920.9889526367188 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.628 ref_kT = 2.560
Predicted entropy: 0.0002931655035354197
Predicted free_energy: -0.11858117580413818
[DiffTRe] Epoch 55
Epoch loss = 0.01418
Gradient norm: 0.5813651084899902
Elapsed time = 0.003 min
[Propagate] Effective sample size: 931.8968505859375 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.442 ref_kT = 2.560
Predicted entropy: -0.0007153092883527279
Predicted free_energy: 0.36797788739204407
[DiffTRe] Epoch 56
Epoch loss = 0.00571
Gradient norm: 0.4162130653858185
Elapsed time = 0.003 min
[Propagate] Effective sample size: 955.8037719726562 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.603 ref_kT = 2.560
Predicted entropy: 0.00012678872735705227
Predicted free_energy: -0.05611446499824524
[DiffTRe] Epoch 57
Epoch loss = 0.01040
Gradient norm: 0.492768794298172
Elapsed time = 0.003 min
[Propagate] Effective sample size: 946.8885498046875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.666 ref_kT = 2.560
Predicted entropy: -0.0008381620282307267
Predicted free_energy: 0.4066760838031769
[DiffTRe] Epoch 58
Epoch loss = 0.00553
Gradient norm: 0.3607192039489746
Elapsed time = 0.003 min
[Propagate] Effective sample size: 963.4417114257812 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.576 ref_kT = 2.560
Predicted entropy: 5.379340291256085e-05
Predicted free_energy: -0.005834996700286865
[DiffTRe] Epoch 59
Epoch loss = 0.00954
Gradient norm: 0.4227699339389801
Elapsed time = 0.003 min
[Propagate] Effective sample size: 955.0051879882812 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.567 ref_kT = 2.560
Predicted entropy: -0.0008921152912080288
Predicted free_energy: 0.4186151623725891
[DiffTRe] Epoch 60
Epoch loss = 0.01431
Gradient norm: 1.4039450883865356
Elapsed time = 0.003 min
[Propagate] Effective sample size: 893.4404296875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.710 ref_kT = 2.560
Predicted entropy: 0.00041972449980676174
Predicted free_energy: -0.13859587907791138
[DiffTRe] Epoch 61
Epoch loss = 0.01006
Gradient norm: 0.5048559308052063
Elapsed time = 0.003 min
[Propagate] Effective sample size: 958.7086791992188 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.0004100153164472431
Predicted free_energy: 0.1918398141860962
[DiffTRe] Epoch 62
Epoch loss = 0.00119
Gradient norm: 0.0011762890499085188
Elapsed time = 0.003 min
[Propagate] Effective sample size: 999.9066162109375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: 0.01156073808670044
Predicted free_energy: 0.1766066551208496
[DiffTRe] Epoch 63
Epoch loss = 0.00117
Gradient norm: 0.0006121537298895419
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.993408203125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.0039056988898664713
Predicted free_energy: 0.19517599046230316
[DiffTRe] Epoch 64
Epoch loss = 0.00116
Gradient norm: 0.0005250716349110007
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9299926757812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: 0.003954899497330189
Predicted free_energy: 0.18504172563552856
[DiffTRe] Epoch 65
Epoch loss = 0.00114
Gradient norm: 0.00039614035631529987
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9776611328125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.01000569574534893
Predicted free_energy: 0.20173971354961395
[DiffTRe] Epoch 66
Epoch loss = 0.00113
Gradient norm: 0.0005020336830057204
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9166259765625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.0011880038073286414
Predicted free_energy: 0.1904677450656891
[DiffTRe] Epoch 67
Epoch loss = 0.00112
Gradient norm: 0.0005618344875983894
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9443359375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.018223091959953308
Predicted free_energy: 0.21092306077480316
[DiffTRe] Epoch 68
Epoch loss = 0.00112
Gradient norm: 0.0009679766371846199
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8580322265625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.0028159685898572206
Predicted free_energy: 0.19161398708820343
[DiffTRe] Epoch 69
Epoch loss = 0.00112
Gradient norm: 0.0016033949796110392
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8436889648438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.03134654462337494
Predicted free_energy: 0.22611837089061737
[DiffTRe] Epoch 70
Epoch loss = 0.00113
Gradient norm: 0.0036505665630102158
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6082153320312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: 0.005302092991769314
Predicted free_energy: 0.18084006011486053
[DiffTRe] Epoch 71
Epoch loss = 0.00120
Gradient norm: 0.008508962579071522
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.2083740234375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.061589471995830536
Predicted free_energy: 0.2620888352394104
[DiffTRe] Epoch 72
Epoch loss = 0.00140
Gradient norm: 0.025910068303346634
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.3671264648438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: 0.04965490847826004
Predicted free_energy: 0.12528586387634277
[DiffTRe] Epoch 73
Epoch loss = 0.00216
Gradient norm: 0.07484621554613113
Elapsed time = 0.000 min
[Propagate] Effective sample size: 992.9649047851562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.588 ref_kT = 2.560
Predicted entropy: -0.15502247214317322
Predicted free_energy: 0.37352803349494934
[DiffTRe] Epoch 74
Epoch loss = 0.00446
Gradient norm: 0.29706159234046936
Elapsed time = 0.000 min
[Propagate] Effective sample size: 972.8523559570312 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.553 ref_kT = 2.560
Predicted entropy: 0.0002262990892631933
Predicted free_energy: -0.07311373949050903
[DiffTRe] Epoch 75
Epoch loss = 0.01609
Gradient norm: 0.6855965852737427
Elapsed time = 0.003 min
[Propagate] Effective sample size: 904.6710815429688 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.532 ref_kT = 2.560
Predicted entropy: -0.0011741453781723976
Predicted free_energy: 0.5132046341896057
[DiffTRe] Epoch 76
Epoch loss = 0.01064
Gradient norm: 0.8555970788002014
Elapsed time = 0.003 min
[Propagate] Effective sample size: 908.4249267578125 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.603 ref_kT = 2.560
Predicted entropy: 0.00027577151195146143
Predicted free_energy: -0.07845312356948853
[DiffTRe] Epoch 77
Epoch loss = 0.01163
Gradient norm: 0.46661272644996643
Elapsed time = 0.003 min
[Propagate] Effective sample size: 949.3755493164062 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.493 ref_kT = 2.560
Predicted entropy: -0.0005935025401413441
Predicted free_energy: 0.31518253684043884
[DiffTRe] Epoch 78
Epoch loss = 0.00386
Gradient norm: 0.19213931262493134
Elapsed time = 0.003 min
[Propagate] Effective sample size: 981.5813598632812 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.742 ref_kT = 2.560
Predicted entropy: -8.458567026536912e-05
Predicted free_energy: 0.06943610310554504
[DiffTRe] Epoch 79
Epoch loss = 0.00456
Gradient norm: 0.1352195292711258
Elapsed time = 0.003 min
[Propagate] Effective sample size: 983.7227172851562 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.722 ref_kT = 2.560
Predicted entropy: -0.0006073091644793749
Predicted free_energy: 0.3060610592365265
[DiffTRe] Epoch 80
Epoch loss = 0.10339
Gradient norm: 56.60963439941406
Elapsed time = 0.003 min
[Propagate] Effective sample size: 766.240234375 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.514 ref_kT = 2.560
Predicted entropy: 0.0008445967687293887
Predicted free_energy: -0.49233147501945496
[DiffTRe] Epoch 81
Epoch loss = 0.08773
Gradient norm: 1.543127179145813
Elapsed time = 0.003 min
[Propagate] Effective sample size: 993.2798461914062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.514 ref_kT = 2.560
Predicted entropy: -0.05271802097558975
Predicted free_energy: -0.3754882216453552
[DiffTRe] Epoch 82
Epoch loss = 0.07001
Gradient norm: 1.5144460201263428
Elapsed time = 0.000 min
[Propagate] Effective sample size: 972.3649291992188 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.613 ref_kT = 2.560
Predicted entropy: 0.0004783071344718337
Predicted free_energy: -0.2620996832847595
[DiffTRe] Epoch 83
Epoch loss = 0.03963
Gradient norm: 1.3076379299163818
Elapsed time = 0.003 min
[Propagate] Effective sample size: 992.4915161132812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.613 ref_kT = 2.560
Predicted entropy: -0.06714846938848495
Predicted free_energy: -0.1378040313720703
[DiffTRe] Epoch 84
Epoch loss = 0.02472
Gradient norm: 0.9482665061950684
Elapsed time = 0.000 min
[Propagate] Effective sample size: 973.2398071289062 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.552 ref_kT = 2.560
Predicted entropy: -8.37863717606524e-06
Predicted free_energy: -0.008399367332458496
[DiffTRe] Epoch 85
Epoch loss = 0.01183
Gradient norm: 0.5340383648872375
Elapsed time = 0.003 min
[Propagate] Effective sample size: 995.932373046875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.552 ref_kT = 2.560
Predicted entropy: -0.06892790645360947
Predicted free_energy: 0.09839139878749847
[DiffTRe] Epoch 86
Epoch loss = 0.00622
Gradient norm: 0.17785875499248505
Elapsed time = 0.000 min
[Propagate] Effective sample size: 989.4833374023438 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.0004073564778082073
Predicted free_energy: 0.16431325674057007
[DiffTRe] Epoch 87
Epoch loss = 0.00204
Gradient norm: 0.09006495773792267
Elapsed time = 0.003 min
[Propagate] Effective sample size: 999.2202758789062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.03927798941731453
Predicted free_energy: 0.21374928951263428
[DiffTRe] Epoch 88
Epoch loss = 0.00105
Gradient norm: 0.024478930979967117
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.1412353515625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.06079554185271263
Predicted free_energy: 0.2410845011472702
[DiffTRe] Epoch 89
Epoch loss = 0.00078
Gradient norm: 0.005243360996246338
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.4532470703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07128418982028961
Predicted free_energy: 0.25440114736557007
[DiffTRe] Epoch 90
Epoch loss = 0.00073
Gradient norm: 0.0009130621911026537
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.119384765625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07586341351270676
Predicted free_energy: 0.2602165639400482
[DiffTRe] Epoch 91
Epoch loss = 0.00072
Gradient norm: 0.00013197239604778588
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9824829101562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07766830176115036
Predicted free_energy: 0.2625127136707306
[DiffTRe] Epoch 92
Epoch loss = 0.00071
Gradient norm: 1.5950445231283084e-05
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9344482421875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07830256223678589
Predicted free_energy: 0.263324499130249
[DiffTRe] Epoch 93
Epoch loss = 0.00071
Gradient norm: 1.6276856058539124e-06
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9177856445312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07849058508872986
Predicted free_energy: 0.26356983184814453
[DiffTRe] Epoch 94
Epoch loss = 0.00071
Gradient norm: 1.6161952487436793e-07
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.912109375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07852284610271454
Predicted free_energy: 0.26361867785453796
[DiffTRe] Epoch 95
Epoch loss = 0.00071
Gradient norm: 3.67551997726423e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9102172851562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07850473374128342
Predicted free_energy: 0.2636040151119232
[DiffTRe] Epoch 96
Epoch loss = 0.00071
Gradient norm: 2.7047478923236667e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.91259765625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07847312092781067
Predicted free_energy: 0.2635722756385803
[DiffTRe] Epoch 97
Epoch loss = 0.00071
Gradient norm: 2.595615100631221e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9107055664062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07843513041734695
Predicted free_energy: 0.26353323459625244
[DiffTRe] Epoch 98
Epoch loss = 0.00071
Gradient norm: 2.5570619399672978e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9130859375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07839702069759369
Predicted free_energy: 0.2634941637516022
[DiffTRe] Epoch 99
Epoch loss = 0.00071
Gradient norm: 2.5194951902562934e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9135131835938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07835682481527328
Predicted free_energy: 0.2634526491165161
[DiffTRe] Epoch 100
Epoch loss = 0.00071
Gradient norm: 2.484206085284768e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9144897460938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07831565290689468
Predicted free_energy: 0.26340994238853455
[DiffTRe] Epoch 101
Epoch loss = 0.00071
Gradient norm: 2.4499373196817942e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9154052734375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07827290147542953
Predicted free_energy: 0.2633659839630127
[DiffTRe] Epoch 102
Epoch loss = 0.00071
Gradient norm: 2.411466226703851e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9154052734375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07822684198617935
Predicted free_energy: 0.26331838965415955
[DiffTRe] Epoch 103
Epoch loss = 0.00071
Gradient norm: 2.3734592957680434e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9168701171875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07818078249692917
Predicted free_energy: 0.263270765542984
[DiffTRe] Epoch 104
Epoch loss = 0.00071
Gradient norm: 2.3338035504139043e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9182739257812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.0781310647726059
Predicted free_energy: 0.2632195055484772
[DiffTRe] Epoch 105
Epoch loss = 0.00071
Gradient norm: 2.2906139207634624e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.921142578125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07808220386505127
Predicted free_energy: 0.26316946744918823
[DiffTRe] Epoch 106
Epoch loss = 0.00071
Gradient norm: 2.2500330487673637e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.920166015625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07803016155958176
Predicted free_energy: 0.2631157338619232
[DiffTRe] Epoch 107
Epoch loss = 0.00071
Gradient norm: 2.2081849238020368e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9230346679688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.0779755637049675
Predicted free_energy: 0.2630595862865448
[DiffTRe] Epoch 108
Epoch loss = 0.00071
Gradient norm: 2.1638687286440472e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9225463867188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07791997492313385
Predicted free_energy: 0.2630022168159485
[DiffTRe] Epoch 109
Epoch loss = 0.00071
Gradient norm: 2.1171487674109812e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9235229492188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07786084711551666
Predicted free_energy: 0.2629411816596985
[DiffTRe] Epoch 110
Epoch loss = 0.00071
Gradient norm: 2.073821292469802e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9249267578125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07780183851718903
Predicted free_energy: 0.2628801465034485
[DiffTRe] Epoch 111
Epoch loss = 0.00071
Gradient norm: 2.0261699873458383e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9259033203125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.0777401551604271
Predicted free_energy: 0.2628166675567627
[DiffTRe] Epoch 112
Epoch loss = 0.00071
Gradient norm: 1.974937191562276e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9268188476562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07767515629529953
Predicted free_energy: 0.2627495229244232
[DiffTRe] Epoch 113
Epoch loss = 0.00071
Gradient norm: 1.9278514784559775e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9287719726562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07760992646217346
Predicted free_energy: 0.26268240809440613
[DiffTRe] Epoch 114
Epoch loss = 0.00071
Gradient norm: 1.8760985653898388e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9296875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.0775412768125534
Predicted free_energy: 0.26261159777641296
[DiffTRe] Epoch 115
Epoch loss = 0.00071
Gradient norm: 1.8244682209456187e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9325561523438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07747005671262741
Predicted free_energy: 0.262538343667984
[DiffTRe] Epoch 116
Epoch loss = 0.00071
Gradient norm: 1.773890723200111e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9339599609375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07739809900522232
Predicted free_energy: 0.26246389746665955
[DiffTRe] Epoch 117
Epoch loss = 0.00071
Gradient norm: 1.719429221225255e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9339599609375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07732322067022324
Predicted free_energy: 0.2623869776725769
[DiffTRe] Epoch 118
Epoch loss = 0.00071
Gradient norm: 1.668340487981368e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9368286132812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07724747806787491
Predicted free_energy: 0.26230886578559875
[DiffTRe] Epoch 119
Epoch loss = 0.00071
Gradient norm: 1.6096707966539725e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9373168945312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.0771670937538147
Predicted free_energy: 0.262225866317749
[DiffTRe] Epoch 120
Epoch loss = 0.00071
Gradient norm: 1.5611709258678275e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.939697265625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07708635181188583
Predicted free_energy: 0.2621428370475769
[DiffTRe] Epoch 121
Epoch loss = 0.00071
Gradient norm: 1.4977523221659794e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9401245117188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07700180262327194
Predicted free_energy: 0.2620549499988556
[DiffTRe] Epoch 122
Epoch loss = 0.00071
Gradient norm: 1.453823816888189e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9425048828125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07691691815853119
Predicted free_energy: 0.2619682848453522
[DiffTRe] Epoch 123
Epoch loss = 0.00071
Gradient norm: 1.3809451360202729e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9453735351562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07682637870311737
Predicted free_energy: 0.2618742883205414
[DiffTRe] Epoch 124
Epoch loss = 0.00071
Gradient norm: 1.3573588475424003e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9463500976562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07674223929643631
Predicted free_energy: 0.26178884506225586
[DiffTRe] Epoch 125
Epoch loss = 0.00071
Gradient norm: 1.2705710261684544e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9506225585938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07663822174072266
Predicted free_energy: 0.2616789638996124
[DiffTRe] Epoch 126
Epoch loss = 0.00071
Gradient norm: 1.3911871654670449e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.94775390625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07656872272491455
Predicted free_energy: 0.26161307096481323
[DiffTRe] Epoch 127
Epoch loss = 0.00071
Gradient norm: 1.5282802579008603e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.9553833007812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07641508430242538
Predicted free_energy: 0.2614409327507019
[DiffTRe] Epoch 128
Epoch loss = 0.00071
Gradient norm: 4.077440962646506e-08
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.947265625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07645521312952042
Predicted free_energy: 0.26151540875434875
[DiffTRe] Epoch 129
Epoch loss = 0.00071
Gradient norm: 1.6623734211407282e-07
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.975830078125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07599751651287079
Predicted free_energy: 0.2609587609767914
[DiffTRe] Epoch 130
Epoch loss = 0.00071
Gradient norm: 1.1493574447740684e-06
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.902587890625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07687459141016006
Predicted free_energy: 0.26209524273872375
[DiffTRe] Epoch 131
Epoch loss = 0.00071
Gradient norm: 8.69480300025316e-06
Elapsed time = 0.000 min
[Propagate] Effective sample size: 997.1236572265625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.07391263544559479
Predicted free_energy: 0.2583635449409485
[DiffTRe] Epoch 132
Epoch loss = 0.00071
Gradient norm: 7.656068919459358e-05
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.4278564453125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.0826638713479042
Predicted free_energy: 0.26948538422584534
[DiffTRe] Epoch 133
Epoch loss = 0.00072
Gradient norm: 0.0007659340044483542
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.4458618164062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.654 ref_kT = 2.560
Predicted entropy: -0.05340646952390671
Predicted free_energy: 0.2323918640613556
[DiffTRe] Epoch 134
Epoch loss = 0.00082
Gradient norm: 0.008118154481053352
Elapsed time = 0.000 min
[Propagate] Effective sample size: 988.4835815429688 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.477 ref_kT = 2.560
Predicted entropy: -0.0008668985101394355
Predicted free_energy: 0.34405583143234253
[DiffTRe] Epoch 135
Epoch loss = 0.00525
Gradient norm: 0.42401373386383057
Elapsed time = 0.003 min
[Propagate] Effective sample size: 917.0904541015625 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.529 ref_kT = 2.560
Predicted entropy: 0.00014898554945830256
Predicted free_energy: -0.1065322756767273
[DiffTRe] Epoch 136
Epoch loss = 0.01836
Gradient norm: 0.8522506356239319
Elapsed time = 0.003 min
[Propagate] Effective sample size: 925.8302001953125 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: -0.0007419296307489276
Predicted free_energy: 0.3226367235183716
[DiffTRe] Epoch 137
Epoch loss = 0.00202
Gradient norm: 0.08977305889129639
Elapsed time = 0.003 min
[Propagate] Effective sample size: 992.126220703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.102389395236969
Predicted free_energy: 0.16870972514152527
[DiffTRe] Epoch 138
Epoch loss = 0.00150
Gradient norm: 0.04170217737555504
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.2360229492188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.03537546470761299
Predicted free_energy: 0.2704431116580963
[DiffTRe] Epoch 139
Epoch loss = 0.00108
Gradient norm: 0.017471421509981155
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.1566772460938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.08445154130458832
Predicted free_energy: 0.19701294600963593
[DiffTRe] Epoch 140
Epoch loss = 0.00098
Gradient norm: 0.010351981967687607
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.6315307617188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.04895415157079697
Predicted free_energy: 0.250946044921875
[DiffTRe] Epoch 141
Epoch loss = 0.00091
Gradient norm: 0.0066974530927836895
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.741943359375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.08231963962316513
Predicted free_energy: 0.2009582668542862
[DiffTRe] Epoch 142
Epoch loss = 0.00089
Gradient norm: 0.00549100199714303
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.4268188476562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.054617274552583694
Predicted free_energy: 0.2431604117155075
[DiffTRe] Epoch 143
Epoch loss = 0.00088
Gradient norm: 0.00510498508810997
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.588134765625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.08620687574148178
Predicted free_energy: 0.1957067996263504
[DiffTRe] Epoch 144
Epoch loss = 0.00088
Gradient norm: 0.00571803143247962
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.5453491210938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.05533357337117195
Predicted free_energy: 0.2427649050951004
[DiffTRe] Epoch 145
Epoch loss = 0.00089
Gradient norm: 0.007371718063950539
Elapsed time = 0.000 min
[Propagate] Effective sample size: 994.7155151367188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.09565610438585281
Predicted free_energy: 0.18189820647239685
[DiffTRe] Epoch 146
Epoch loss = 0.00094
Gradient norm: 0.010818101465702057
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.0487670898438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.049000684171915054
Predicted free_energy: 0.25291261076927185
[DiffTRe] Epoch 147
Epoch loss = 0.00103
Gradient norm: 0.018559440970420837
Elapsed time = 0.000 min
[Propagate] Effective sample size: 992.1683349609375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.11533694714307785
Predicted free_energy: 0.1520984023809433
[DiffTRe] Epoch 148
Epoch loss = 0.00128
Gradient norm: 0.032545559108257294
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7898559570312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.580 ref_kT = 2.560
Predicted entropy: 0.02950919046998024
Predicted free_energy: 0.28254273533821106
[DiffTRe] Epoch 149
Epoch loss = 0.00164
Gradient norm: 0.06591632217168808
Elapsed time = 0.000 min
[Propagate] Effective sample size: 985.651123046875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.542 ref_kT = 2.560
Predicted entropy: -0.00020104279974475503
Predicted free_energy: 0.09425345063209534
[DiffTRe] Epoch 150
Epoch loss = 0.00191
Gradient norm: 0.06818103045225143
Elapsed time = 0.003 min
[Propagate] Effective sample size: 987.3516235351562 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.457 ref_kT = 2.560
Predicted entropy: -0.0005822874954901636
Predicted free_energy: 0.28466230630874634
[DiffTRe] Epoch 151
Epoch loss = 0.00295
Gradient norm: 0.027164828032255173
Elapsed time = 0.003 min
[Propagate] Effective sample size: 995.5477905273438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.457 ref_kT = 2.560
Predicted entropy: 0.052186720073223114
Predicted free_energy: 0.17181074619293213
[DiffTRe] Epoch 152
Epoch loss = 0.00357
Gradient norm: 0.0710373967885971
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.20068359375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.457 ref_kT = 2.560
Predicted entropy: -0.04876452311873436
Predicted free_energy: 0.3730815052986145
[DiffTRe] Epoch 153
Epoch loss = 0.00488
Gradient norm: 0.22696836292743683
Elapsed time = 0.000 min
[Propagate] Effective sample size: 984.9689331054688 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.695 ref_kT = 2.560
Predicted entropy: -0.00016174474149011075
Predicted free_energy: 0.0737295001745224
[DiffTRe] Epoch 154
Epoch loss = 0.00642
Gradient norm: 0.23641805350780487
Elapsed time = 0.003 min
[Propagate] Effective sample size: 974.6869506835938 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: -0.0008455367642454803
Predicted free_energy: 0.36718541383743286
[DiffTRe] Epoch 155
Epoch loss = 0.00342
Gradient norm: 0.06104237586259842
Elapsed time = 0.003 min
[Propagate] Effective sample size: 993.48681640625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.10443101823329926
Predicted free_energy: 0.21145543456077576
[DiffTRe] Epoch 156
Epoch loss = 0.00296
Gradient norm: 0.029089538380503654
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.2864990234375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.04175901040434837
Predicted free_energy: 0.3076260983943939
[DiffTRe] Epoch 157
Epoch loss = 0.00260
Gradient norm: 0.01191561296582222
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.7633666992188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.09322860091924667
Predicted free_energy: 0.23184484243392944
[DiffTRe] Epoch 158
Epoch loss = 0.00249
Gradient norm: 0.00716664781793952
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.611083984375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.061803948134183884
Predicted free_energy: 0.2809073328971863
[DiffTRe] Epoch 159
Epoch loss = 0.00240
Gradient norm: 0.0043516685254871845
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.089111328125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.0965779572725296
Predicted free_energy: 0.22988438606262207
[DiffTRe] Epoch 160
Epoch loss = 0.00235
Gradient norm: 0.0036036306992173195
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.2511596679688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.07324721664190292
Predicted free_energy: 0.266700804233551
[DiffTRe] Epoch 161
Epoch loss = 0.00231
Gradient norm: 0.003106089076027274
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.8065185546875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.10493109375238419
Predicted free_energy: 0.2200516313314438
[DiffTRe] Epoch 162
Epoch loss = 0.00229
Gradient norm: 0.0034764402080327272
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.154541015625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.07947061955928802
Predicted free_energy: 0.2600955665111542
[DiffTRe] Epoch 163
Epoch loss = 0.00227
Gradient norm: 0.004168632905930281
Elapsed time = 0.000 min
[Propagate] Effective sample size: 994.980224609375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.11748353391885757
Predicted free_energy: 0.20354406535625458
[DiffTRe] Epoch 164
Epoch loss = 0.00228
Gradient norm: 0.006046382710337639
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.4087524414062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.07920067757368088
Predicted free_energy: 0.26312413811683655
[DiffTRe] Epoch 165
Epoch loss = 0.00231
Gradient norm: 0.009758592583239079
Elapsed time = 0.000 min
[Propagate] Effective sample size: 992.9734497070312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.13799113035202026
Predicted free_energy: 0.17455601692199707
[DiffTRe] Epoch 166
Epoch loss = 0.00243
Gradient norm: 0.017025666311383247
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.0602416992188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.525 ref_kT = 2.560
Predicted entropy: 0.06736026704311371
Predicted free_energy: 0.28356847167015076
[DiffTRe] Epoch 167
Epoch loss = 0.00263
Gradient norm: 0.033410828560590744
Elapsed time = 0.000 min
[Propagate] Effective sample size: 988.103759765625 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.675 ref_kT = 2.560
Predicted entropy: -0.00018765449931379408
Predicted free_energy: 0.10769063234329224
[DiffTRe] Epoch 168
Epoch loss = 0.00222
Gradient norm: 0.07649091631174088
Elapsed time = 0.003 min
[Propagate] Effective sample size: 984.5120239257812 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.552 ref_kT = 2.560
Predicted entropy: -0.0007877166499383748
Predicted free_energy: 0.353839635848999
[DiffTRe] Epoch 169
Epoch loss = 0.00397
Gradient norm: 0.24211327731609344
Elapsed time = 0.003 min
[Propagate] Effective sample size: 971.1356201171875 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.384 ref_kT = 2.560
Predicted entropy: -0.0001316229026997462
Predicted free_energy: 0.05019447207450867
[DiffTRe] Epoch 170
Epoch loss = 0.00483
Gradient norm: 0.1499573290348053
Elapsed time = 0.002 min
[Propagate] Effective sample size: 986.0756225585938 (990.0) -> Recompute is True
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.0005531908245757222
Predicted free_energy: 0.26362335681915283
[DiffTRe] Epoch 171
Epoch loss = 0.00175
Gradient norm: 0.00026600121054798365
Elapsed time = 0.003 min
[Propagate] Effective sample size: 999.9719848632812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.007627727929502726
Predicted free_energy: 0.27506500482559204
[DiffTRe] Epoch 172
Epoch loss = 0.00174
Gradient norm: 0.00018478021956980228
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9991455078125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.0016428017988801003
Predicted free_energy: 0.26535430550575256
[DiffTRe] Epoch 173
Epoch loss = 0.00174
Gradient norm: 0.00015287283167708665
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.973876953125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.007530405186116695
Predicted free_energy: 0.2748696804046631
[DiffTRe] Epoch 174
Epoch loss = 0.00174
Gradient norm: 0.00015550546231679618
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9986572265625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.0015242935623973608
Predicted free_energy: 0.2651248276233673
[DiffTRe] Epoch 175
Epoch loss = 0.00174
Gradient norm: 0.00018407453899271786
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9662475585938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.008573605678975582
Predicted free_energy: 0.2765176296234131
[DiffTRe] Epoch 176
Epoch loss = 0.00175
Gradient norm: 0.00026405672542750835
Elapsed time = 0.000 min
[Propagate] Effective sample size: 1000.0001220703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 2.6657478883862495e-05
Predicted free_energy: 0.2625699043273926
[DiffTRe] Epoch 177
Epoch loss = 0.00175
Gradient norm: 0.0004350962408352643
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.933349609375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.011757948435842991
Predicted free_energy: 0.28162017464637756
[DiffTRe] Epoch 178
Epoch loss = 0.00175
Gradient norm: 0.0008573969826102257
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.9767456054688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.005238635465502739
Predicted free_energy: 0.2540542781352997
[DiffTRe] Epoch 179
Epoch loss = 0.00177
Gradient norm: 0.0019025783985853195
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.76220703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02152797393500805
Predicted free_energy: 0.2972854673862457
[DiffTRe] Epoch 180
Epoch loss = 0.00182
Gradient norm: 0.005039378069341183
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6568603515625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02286083810031414
Predicted free_energy: 0.22512482106685638
[DiffTRe] Epoch 181
Epoch loss = 0.00196
Gradient norm: 0.013912174850702286
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.5006103515625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.05351552367210388
Predicted free_energy: 0.34800079464912415
[DiffTRe] Epoch 182
Epoch loss = 0.00237
Gradient norm: 0.04496917128562927
Elapsed time = 0.000 min
[Propagate] Effective sample size: 996.5499267578125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.07267089933156967
Predicted free_energy: 0.14139436185359955
[DiffTRe] Epoch 183
Epoch loss = 0.00347
Gradient norm: 0.10086008906364441
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.3650512695312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.09481312334537506
Predicted free_energy: 0.4123440682888031
[DiffTRe] Epoch 184
Epoch loss = 0.00391
Gradient norm: 0.16379888355731964
Elapsed time = 0.000 min
[Propagate] Effective sample size: 995.3370361328125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.08569223433732986
Predicted free_energy: 0.1194986030459404
[DiffTRe] Epoch 185
Epoch loss = 0.00400
Gradient norm: 0.12839940190315247
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.3178100585938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.05400052294135094
Predicted free_energy: 0.3496255576610565
[DiffTRe] Epoch 186
Epoch loss = 0.00249
Gradient norm: 0.05311335623264313
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.12548828125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.04000628739595413
Predicted free_energy: 0.1976870745420456
[DiffTRe] Epoch 187
Epoch loss = 0.00221
Gradient norm: 0.028799109160900116
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4642944335938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.028896942734718323
Predicted free_energy: 0.3100235164165497
[DiffTRe] Epoch 188
Epoch loss = 0.00196
Gradient norm: 0.014702939428389072
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7531127929688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02235882356762886
Predicted free_energy: 0.22696198523044586
[DiffTRe] Epoch 189
Epoch loss = 0.00189
Gradient norm: 0.009488023817539215
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.72021484375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02035915106534958
Predicted free_energy: 0.2963150143623352
[DiffTRe] Epoch 190
Epoch loss = 0.00184
Gradient norm: 0.006859931629151106
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8599243164062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.016988523304462433
Predicted free_energy: 0.2357071042060852
[DiffTRe] Epoch 191
Epoch loss = 0.00184
Gradient norm: 0.005892251152545214
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.751220703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.019432060420513153
Predicted free_energy: 0.2947281002998352
[DiffTRe] Epoch 192
Epoch loss = 0.00183
Gradient norm: 0.005915910936892033
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8389282226562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.017826788127422333
Predicted free_energy: 0.23422761261463165
[DiffTRe] Epoch 193
Epoch loss = 0.00185
Gradient norm: 0.006746315862983465
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6434936523438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.024011757224798203
Predicted free_energy: 0.3019571006298065
[DiffTRe] Epoch 194
Epoch loss = 0.00187
Gradient norm: 0.009005422703921795
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6592407226562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.024996720254421234
Predicted free_energy: 0.22232696413993835
[DiffTRe] Epoch 195
Epoch loss = 0.00195
Gradient norm: 0.012877039611339569
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.2574462890625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.03570021316409111
Predicted free_energy: 0.3204959034919739
[DiffTRe] Epoch 196
Epoch loss = 0.00204
Gradient norm: 0.020911768078804016
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.036376953125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.040889181196689606
Predicted free_energy: 0.1958853155374527
[DiffTRe] Epoch 197
Epoch loss = 0.00226
Gradient norm: 0.0320969820022583
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.4063720703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.053264934569597244
Predicted free_energy: 0.3482290804386139
[DiffTRe] Epoch 198
Epoch loss = 0.00244
Gradient norm: 0.04957028105854988
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.0393676757812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.05755646899342537
Predicted free_energy: 0.16793976724147797
[DiffTRe] Epoch 199
Epoch loss = 0.00274
Gradient norm: 0.0597454309463501
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.1683349609375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.0567590706050396
Predicted free_energy: 0.3538796901702881
[DiffTRe] Epoch 200
Epoch loss = 0.00256
Gradient norm: 0.058301981538534164
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.4234619140625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.052543964236974716
Predicted free_energy: 0.17661651968955994
[DiffTRe] Epoch 201
Epoch loss = 0.00255
Gradient norm: 0.04874968156218529
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.8796997070312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.043175261467695236
Predicted free_energy: 0.3326736390590668
[DiffTRe] Epoch 202
Epoch loss = 0.00222
Gradient norm: 0.03387156501412392
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.244140625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.03759397566318512
Predicted free_energy: 0.20178131759166718
[DiffTRe] Epoch 203
Epoch loss = 0.00215
Gradient norm: 0.025170188397169113
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.352294921875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.03200574591755867
Predicted free_energy: 0.3149954378604889
[DiffTRe] Epoch 204
Epoch loss = 0.00201
Gradient norm: 0.018303917720913887
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.58154296875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02860076166689396
Predicted free_energy: 0.2166970819234848
[DiffTRe] Epoch 205
Epoch loss = 0.00198
Gradient norm: 0.014958413317799568
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5157470703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02746734768152237
Predicted free_energy: 0.30771270394325256
[DiffTRe] Epoch 206
Epoch loss = 0.00193
Gradient norm: 0.013090403750538826
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6558837890625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.025900790467858315
Predicted free_energy: 0.2210916131734848
[DiffTRe] Epoch 207
Epoch loss = 0.00194
Gradient norm: 0.012654143385589123
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.50048828125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02816326729953289
Predicted free_energy: 0.3087441921234131
[DiffTRe] Epoch 208
Epoch loss = 0.00194
Gradient norm: 0.013453827239573002
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.58056640625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.0281553752720356
Predicted free_energy: 0.21728791296482086
[DiffTRe] Epoch 209
Epoch loss = 0.00198
Gradient norm: 0.015161029994487762
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.3380126953125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.03303053602576256
Predicted free_energy: 0.3164224326610565
[DiffTRe] Epoch 210
Epoch loss = 0.00201
Gradient norm: 0.018498210236430168
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.351318359375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.03439982235431671
Predicted free_energy: 0.20688019692897797
[DiffTRe] Epoch 211
Epoch loss = 0.00210
Gradient norm: 0.022376567125320435
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.057373046875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.04001018777489662
Predicted free_energy: 0.32746246457099915
[DiffTRe] Epoch 212
Epoch loss = 0.00214
Gradient norm: 0.027612194418907166
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.0502319335938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.04112241789698601
Predicted free_energy: 0.1956741362810135
[DiffTRe] Epoch 213
Epoch loss = 0.00225
Gradient norm: 0.03128916025161743
Elapsed time = 0.000 min
[Propagate] Effective sample size: 998.890625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.04350772500038147
Predicted free_energy: 0.33301910758018494
[DiffTRe] Epoch 214
Epoch loss = 0.00221
Gradient norm: 0.03318024054169655
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.018798828125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.04189695417881012
Predicted free_energy: 0.19442658126354218
[DiffTRe] Epoch 215
Epoch loss = 0.00226
Gradient norm: 0.03210335224866867
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.0387573242188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.04012291878461838
Predicted free_energy: 0.327720046043396
[DiffTRe] Epoch 216
Epoch loss = 0.00215
Gradient norm: 0.02827434428036213
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.2741088867188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.0365321971476078
Predicted free_energy: 0.2034182846546173
[DiffTRe] Epoch 217
Epoch loss = 0.00214
Gradient norm: 0.024558175355196
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.2789306640625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.034363191574811935
Predicted free_energy: 0.3186050355434418
[DiffTRe] Epoch 218
Epoch loss = 0.00204
Gradient norm: 0.020473718643188477
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4947509765625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.03084920160472393
Predicted free_energy: 0.2128555327653885
[DiffTRe] Epoch 219
Epoch loss = 0.00203
Gradient norm: 0.01786050759255886
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4227905273438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.03057454712688923
Predicted free_energy: 0.3125552535057068
[DiffTRe] Epoch 220
Epoch loss = 0.00197
Gradient norm: 0.01586994342505932
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5877075195312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.027883559465408325
Predicted free_energy: 0.21772614121437073
[DiffTRe] Epoch 221
Epoch loss = 0.00198
Gradient norm: 0.014939622953534126
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4571533203125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02974815107882023
Predicted free_energy: 0.31119048595428467
[DiffTRe] Epoch 222
Epoch loss = 0.00196
Gradient norm: 0.014748288318514824
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5829467773438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.027881575748324394
Predicted free_energy: 0.21767853200435638
[DiffTRe] Epoch 223
Epoch loss = 0.00198
Gradient norm: 0.015165572986006737
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4032592773438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.03140055388212204
Predicted free_energy: 0.3137710690498352
[DiffTRe] Epoch 224
Epoch loss = 0.00198
Gradient norm: 0.01631847769021988
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5047607421875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.030079327523708344
Predicted free_energy: 0.21399566531181335
[DiffTRe] Epoch 225
Epoch loss = 0.00202
Gradient norm: 0.017673185095191002
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.3036499023438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.034241676330566406
Predicted free_energy: 0.3182522654533386
[DiffTRe] Epoch 226
Epoch loss = 0.00202
Gradient norm: 0.019474787637591362
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4061279296875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.03274628520011902
Predicted free_energy: 0.20954987406730652
[DiffTRe] Epoch 227
Epoch loss = 0.00208
Gradient norm: 0.020808152854442596
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.232666015625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.036092668771743774
Predicted free_energy: 0.32117828726768494
[DiffTRe] Epoch 228
Epoch loss = 0.00205
Gradient norm: 0.02174541726708412
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.3742065429688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.03352247551083565
Predicted free_energy: 0.20825715363025665
[DiffTRe] Epoch 229
Epoch loss = 0.00209
Gradient norm: 0.021747710183262825
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.26171875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.035376109182834625
Predicted free_energy: 0.32004424929618835
[DiffTRe] Epoch 230
Epoch loss = 0.00204
Gradient norm: 0.020838145166635513
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4475708007812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.031656570732593536
Predicted free_energy: 0.2113565057516098
[DiffTRe] Epoch 231
Epoch loss = 0.00206
Gradient norm: 0.019552649930119514
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.3560791015625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.032922420650720596
Predicted free_energy: 0.31614288687705994
[DiffTRe] Epoch 232
Epoch loss = 0.00200
Gradient norm: 0.01782418042421341
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5467529296875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02876231260597706
Predicted free_energy: 0.21614167094230652
[DiffTRe] Epoch 233
Epoch loss = 0.00200
Gradient norm: 0.016434838995337486
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.43994140625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.030633268877863884
Predicted free_energy: 0.31248199939727783
[DiffTRe] Epoch 234
Epoch loss = 0.00196
Gradient norm: 0.015155640430748463
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6139526367188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02661435678601265
Predicted free_energy: 0.21966828405857086
[DiffTRe] Epoch 235
Epoch loss = 0.00197
Gradient norm: 0.014370421878993511
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4785766601562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.029590707272291183
Predicted free_energy: 0.310790091753006
[DiffTRe] Epoch 236
Epoch loss = 0.00194
Gradient norm: 0.013911803252995014
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6325073242188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02588626556098461
Predicted free_energy: 0.22083649039268494
[DiffTRe] Epoch 237
Epoch loss = 0.00196
Gradient norm: 0.013811017386615276
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4728393554688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.029877621680498123
Predicted free_energy: 0.3112112581729889
[DiffTRe] Epoch 238
Epoch loss = 0.00195
Gradient norm: 0.014050908386707306
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6148681640625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.026374561712145805
Predicted free_energy: 0.2199966460466385
[DiffTRe] Epoch 239
Epoch loss = 0.00197
Gradient norm: 0.014431674964725971
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.44189453125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.030920391902327538
Predicted free_energy: 0.31284087896347046
[DiffTRe] Epoch 240
Epoch loss = 0.00196
Gradient norm: 0.01501269731670618
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5848388671875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.027290550991892815
Predicted free_energy: 0.21845611929893494
[DiffTRe] Epoch 241
Epoch loss = 0.00199
Gradient norm: 0.01546627376228571
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4137573242188 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.031765129417181015
Predicted free_energy: 0.3141641318798065
[DiffTRe] Epoch 242
Epoch loss = 0.00197
Gradient norm: 0.015833789482712746
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.571533203125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.027625923976302147
Predicted free_energy: 0.2178848385810852
[DiffTRe] Epoch 243
Epoch loss = 0.00199
Gradient norm: 0.015886781737208366
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4194946289062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.03164578974246979
Predicted free_energy: 0.3139602839946747
[DiffTRe] Epoch 244
Epoch loss = 0.00197
Gradient norm: 0.01563776284456253
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5963134765625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.026866255328059196
Predicted free_energy: 0.21912872791290283
[DiffTRe] Epoch 245
Epoch loss = 0.00198
Gradient norm: 0.01516709290444851
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.4571533203125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.03060743398964405
Predicted free_energy: 0.31229522824287415
[DiffTRe] Epoch 246
Epoch loss = 0.00195
Gradient norm: 0.014462215825915337
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6377563476562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.025402924045920372
Predicted free_energy: 0.22153107821941376
[DiffTRe] Epoch 247
Epoch loss = 0.00196
Gradient norm: 0.01378859207034111
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.029345408082008362
Predicted free_energy: 0.3102676570415497
[DiffTRe] Epoch 248
Epoch loss = 0.00193
Gradient norm: 0.013088238425552845
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6768798828125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02401617169380188
Predicted free_energy: 0.22379913926124573
[DiffTRe] Epoch 249
Epoch loss = 0.00194
Gradient norm: 0.012567493133246899
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.530517578125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02850301004946232
Predicted free_energy: 0.30890166759490967
[DiffTRe] Epoch 250
Epoch loss = 0.00192
Gradient norm: 0.012162074446678162
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6959228515625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02321126125752926
Predicted free_energy: 0.22510163486003876
[DiffTRe] Epoch 251
Epoch loss = 0.00193
Gradient norm: 0.011935462243855
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5391235351562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.028296586126089096
Predicted free_energy: 0.3085476756095886
[DiffTRe] Epoch 252
Epoch loss = 0.00191
Gradient norm: 0.011859605088829994
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6992797851562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.023035060614347458
Predicted free_energy: 0.2253665179014206
[DiffTRe] Epoch 253
Epoch loss = 0.00193
Gradient norm: 0.011879022233188152
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5333862304688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.028547869995236397
Predicted free_energy: 0.3089236617088318
[DiffTRe] Epoch 254
Epoch loss = 0.00192
Gradient norm: 0.011998646892607212
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.691650390625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.023179585114121437
Predicted free_energy: 0.22510528564453125
[DiffTRe] Epoch 255
Epoch loss = 0.00193
Gradient norm: 0.012099952436983585
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5271606445312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.028841344639658928
Predicted free_energy: 0.3093704283237457
[DiffTRe] Epoch 256
Epoch loss = 0.00192
Gradient norm: 0.012194310314953327
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.691650390625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02319018915295601
Predicted free_energy: 0.22506989538669586
[DiffTRe] Epoch 257
Epoch loss = 0.00193
Gradient norm: 0.012183820828795433
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.531005859375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02878928743302822
Predicted free_energy: 0.30927032232284546
[DiffTRe] Epoch 258
Epoch loss = 0.00192
Gradient norm: 0.012073732912540436
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.70166015625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.022761918604373932
Predicted free_energy: 0.2257595807313919
[DiffTRe] Epoch 259
Epoch loss = 0.00193
Gradient norm: 0.011861461214721203
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.5476684570312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.0282980315387249
Predicted free_energy: 0.3084719777107239
[DiffTRe] Epoch 260
Epoch loss = 0.00191
Gradient norm: 0.011544710956513882
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.720703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02195085771381855
Predicted free_energy: 0.22708038985729218
[DiffTRe] Epoch 261
Epoch loss = 0.00192
Gradient norm: 0.011205408722162247
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.571044921875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.027592264115810394
Predicted free_energy: 0.3073306381702423
[DiffTRe] Epoch 262
Epoch loss = 0.00190
Gradient norm: 0.010828757658600807
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.74169921875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.021059514954686165
Predicted free_energy: 0.22853058576583862
[DiffTRe] Epoch 263
Epoch loss = 0.00191
Gradient norm: 0.010506071150302887
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.591064453125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.026985498145222664
Predicted free_energy: 0.30634430050849915
[DiffTRe] Epoch 264
Epoch loss = 0.00189
Gradient norm: 0.010215135291218758
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7578735351562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.02037207782268524
Predicted free_energy: 0.22964264452457428
[DiffTRe] Epoch 265
Epoch loss = 0.00190
Gradient norm: 0.01000273134559393
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6024780273438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02665608935058117
Predicted free_energy: 0.30579864978790283
[DiffTRe] Epoch 266
Epoch loss = 0.00189
Gradient norm: 0.00985272228717804
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7650146484375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.019987570121884346
Predicted free_energy: 0.23025543987751007
[DiffTRe] Epoch 267
Epoch loss = 0.00190
Gradient norm: 0.009758569300174713
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6077270507812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02657749317586422
Predicted free_energy: 0.3056546151638031
[DiffTRe] Epoch 268
Epoch loss = 0.00188
Gradient norm: 0.009714044630527496
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7693481445312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.019806882366538048
Predicted free_energy: 0.23053376376628876
[DiffTRe] Epoch 269
Epoch loss = 0.00189
Gradient norm: 0.009679406881332397
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6091918945312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.026583997532725334
Predicted free_energy: 0.3056472837924957
[DiffTRe] Epoch 270
Epoch loss = 0.00188
Gradient norm: 0.00965423695743084
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.772216796875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.019634053111076355
Predicted free_energy: 0.23080231249332428
[DiffTRe] Epoch 271
Epoch loss = 0.00189
Gradient norm: 0.009598271921277046
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.611572265625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.026485521346330643
Predicted free_energy: 0.3054739534854889
[DiffTRe] Epoch 272
Epoch loss = 0.00188
Gradient norm: 0.00951145775616169
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7778930664062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.019298555329442024
Predicted free_energy: 0.23133820295333862
[DiffTRe] Epoch 273
Epoch loss = 0.00189
Gradient norm: 0.009382855147123337
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6229858398438 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.026190444827079773
Predicted free_energy: 0.3049893081188202
[DiffTRe] Epoch 274
Epoch loss = 0.00188
Gradient norm: 0.009210917167365551
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7893676757812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.018775932490825653
Predicted free_energy: 0.23218293488025665
[DiffTRe] Epoch 275
Epoch loss = 0.00188
Gradient norm: 0.009019371122121811
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6358642578125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.025760099291801453
Predicted free_energy: 0.3042874038219452
[DiffTRe] Epoch 276
Epoch loss = 0.00187
Gradient norm: 0.00880349799990654
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.80078125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.018171006813645363
Predicted free_energy: 0.23316192626953125
[DiffTRe] Epoch 277
Epoch loss = 0.00188
Gradient norm: 0.008600380271673203
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6492309570312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.025327283889055252
Predicted free_energy: 0.3035818636417389
[DiffTRe] Epoch 278
Epoch loss = 0.00186
Gradient norm: 0.0084022656083107
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.812255859375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.01762300170958042
Predicted free_energy: 0.23404693603515625
[DiffTRe] Epoch 279
Epoch loss = 0.00187
Gradient norm: 0.008235456421971321
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.66015625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02500266209244728
Predicted free_energy: 0.30304718017578125
[DiffTRe] Epoch 280
Epoch loss = 0.00186
Gradient norm: 0.008092297241091728
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8203735351562 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.01721014268696308
Predicted free_energy: 0.23470978438854218
[DiffTRe] Epoch 281
Epoch loss = 0.00187
Gradient norm: 0.00797810684889555
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6654052734375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.024804025888442993
Predicted free_energy: 0.30271515250205994
[DiffTRe] Epoch 282
Epoch loss = 0.00186
Gradient norm: 0.007887446321547031
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8270263671875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.016914574429392815
Predicted free_energy: 0.23517975211143494
[DiffTRe] Epoch 283
Epoch loss = 0.00186
Gradient norm: 0.0078094471246004105
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.671630859375 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.024676477536559105
Predicted free_energy: 0.30249664187431335
[DiffTRe] Epoch 284
Epoch loss = 0.00185
Gradient norm: 0.0077409204095602036
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8289184570312 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.016657153144478798
Predicted free_energy: 0.2355886846780777
[DiffTRe] Epoch 285
Epoch loss = 0.00186
Gradient norm: 0.007666534278541803
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6754150390625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02452925778925419
Predicted free_energy: 0.3022476136684418
[DiffTRe] Epoch 286
Epoch loss = 0.00185
Gradient norm: 0.007584180682897568
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8341674804688 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.016348911449313164
Predicted free_energy: 0.23608063161373138
[DiffTRe] Epoch 287
Epoch loss = 0.00186
Gradient norm: 0.007486959919333458
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.6830444335938 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.024301964789628983
Predicted free_energy: 0.3018728792667389
[DiffTRe] Epoch 288
Epoch loss = 0.00185
Gradient norm: 0.0073730903677642345
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8408203125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.015960395336151123
Predicted free_energy: 0.23670685291290283
[DiffTRe] Epoch 289
Epoch loss = 0.00185
Gradient norm: 0.00724849384278059
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.691650390625 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.024000221863389015
Predicted free_energy: 0.3013784885406494
[DiffTRe] Epoch 290
Epoch loss = 0.00185
Gradient norm: 0.007111131679266691
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8499145507812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.015513913705945015
Predicted free_energy: 0.23742584884166718
[DiffTRe] Epoch 291
Epoch loss = 0.00185
Gradient norm: 0.006975288037210703
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.701171875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.023677293211221695
Predicted free_energy: 0.30084991455078125
[DiffTRe] Epoch 292
Epoch loss = 0.00184
Gradient norm: 0.006838047411292791
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8565673828125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.01507781632244587
Predicted free_energy: 0.23812897503376007
[DiffTRe] Epoch 293
Epoch loss = 0.00185
Gradient norm: 0.006712732836604118
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7083129882812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.023389894515275955
Predicted free_energy: 0.3003775179386139
[DiffTRe] Epoch 294
Epoch loss = 0.00184
Gradient norm: 0.006595405284315348
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.86279296875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.014697391539812088
Predicted free_energy: 0.23874054849147797
[DiffTRe] Epoch 295
Epoch loss = 0.00184
Gradient norm: 0.006492304150015116
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.7173461914062 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02316209487617016
Predicted free_energy: 0.3000015318393707
[DiffTRe] Epoch 296
Epoch loss = 0.00184
Gradient norm: 0.006399841979146004
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.8684692382812 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.014378974214196205
Predicted free_energy: 0.2392495721578598
[DiffTRe] Epoch 297
Epoch loss = 0.00184
Gradient norm: 0.006316690240055323
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.720703125 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: -0.02297964319586754
Predicted free_energy: 0.2996987998485565
[DiffTRe] Epoch 298
Epoch loss = 0.00183
Gradient norm: 0.006239705719053745
Elapsed time = 0.000 min
[Propagate] Effective sample size: 999.871826171875 (990.0) -> Recompute is False
[Statepoint 0]
kT = 2.535 ref_kT = 2.560
Predicted entropy: 0.01409659069031477
Predicted free_energy: 0.23970124125480652
[DiffTRe] Epoch 299
Epoch loss = 0.00184
Gradient norm: 0.006164095364511013
Elapsed time = 0.000 min
Results#
plt.plot(trainer.epoch_losses)
plt.title("Loss History")
plt.xlabel("Iterations")
plt.ylabel("Loss")
Text(0, 0.5, 'Loss')
The plot shows, that DiffTRe is able to learn the correct ensemble average.
last_epoch = len(trainer.predictions[0]) - 1
plt.plot(onp.linspace(0.00, 1.0, 50), trainer.predictions[0][0]['rdf'], label="Initial")
plt.plot(onp.linspace(0.00, 1.0, 50), trainer.predictions[0][last_epoch]['rdf'], label="Final")
plt.plot(r, radial_distribution(r), label="Reference")
plt.legend()
plt.title("Radial Distribution Function")
plt.show()
Let’s also take a look at the inferred parameters.
print(trainer.params)
{'r_0': Array(0.34808064, dtype=float32), 'scaled_b': Array(2.4722724, dtype=float32)}
Further Reading#
Examples#
Publications#
Thaler, S., Zavadlav, J. Learning neural network potentials from experimental data via Differentiable Trajectory Reweighting. Nat Commun 12, 6884 (2021). https://doi.org/10.1038/s41467-021-27241-4
Carles Navarro and Maciej Majewski and Gianni de Fabritiis Top-down machine learning of coarse-grained protein force-fields. arXiv (2023). https://arxiv.org/abs/2306.11375