diff --git a/src/inference/hmc.jl b/src/inference/hmc.jl
index 6bffb390c..d1513ceed 100644
--- a/src/inference/hmc.jl
+++ b/src/inference/hmc.jl
@@ -206,13 +206,15 @@ end
 
 
 function sample(
-    model::Model, alg::Hamiltonian;
+    model::Model,
+    alg::Hamiltonian;
     save_state=false,                                   # flag for state saving
     resume_from=nothing,                                # chain to continue
     reuse_spl_n=0,                                      # flag for spl re-using
-    adaptor = AHMCAdaptor(),
+    # adaptor = AHMCAdaptor(alg),
     init_theta::Union{Nothing,Array{<:Any,1}}=nothing,
     rng::AbstractRNG=GLOBAL_RNG,
+    kwargs...
 )
     # Create sampler
     spl = reuse_spl_n > 0 ? resume_from.info[:spl] : Sampler(alg)
@@ -308,7 +310,7 @@ function step(
     spl::Sampler{<:AdaptiveHamiltonian},
     vi::VarInfo,
     is_first::Val{true};
-    adaptor=AHMCAdaptor(),
+    # adaptor=AHMCAdaptor(),
     kwargs...
 )
     spl.selector.tag != :default && link!(vi, spl)
@@ -329,7 +331,7 @@ function step(
 
     spl.info[:h] = h
     spl.info[:traj] = gen_traj(spl.alg, init_ϵ)
-    spl.info[:adaptor] = adaptor
+    spl.info[:adaptor] = AHMCAdaptor(spl.alg, metric)
 
     spl.selector.tag != :default && invlink!(vi, spl)
     return vi, true
@@ -576,9 +578,9 @@ observe(spl::Sampler{<:Hamiltonian},
 #### Default HMC stepsize and mass matrix adaptor
 ####
 
-function AHMCAdaptor()
+function AHMCAdaptor(alg::AdaptiveHamiltonian, metric)
         adaptor = AHMC.StanNUTSAdaptor(
-            spl.alg.n_adapts, AHMC.PreConditioner(metric),
-            AHMC.NesterovDualAveraging(spl.alg.δ, init_ϵ)
+            alg.n_adapts, AHMC.PreConditioner(metric),
+            AHMC.NesterovDualAveraging(alg.δ, alg.init_ϵ)
         )
 end
diff --git a/src/utilities/stan-interface.jl b/src/utilities/stan-interface.jl
index 2a7ed0b55..a06b220b4 100644
--- a/src/utilities/stan-interface.jl
+++ b/src/utilities/stan-interface.jl
@@ -15,34 +15,37 @@
 
 #   Ref
 #     http://goedman.github.io/Stan.jl/latest/index.html#Types-1
+
 function sample(mf::T, ss::CmdStan.Sample) where T
-    return sample(mf, ss.num_samples, ss.num_warmup, ss.save_warmup, ss.thin, ss.adapt, ss.alg)
+    return sample(mf, ss.num_samples, ss.num_warmup,
+                    ss.save_warmup, ss.thin, ss.adapt, ss.alg)
 end
-function sample(  mf::T,
-                  num_samples::Int,
-                  num_warmup::Int,
-                  save_warmup::Bool,
-                  thin::Int,
-                  ss::CmdStan.Sample
-                ) where T
+
+function sample(mf::T,
+    num_samples::Int,
+    num_warmup::Int,
+    save_warmup::Bool,
+    thin::Int,
+    ss::CmdStan.Sample
+) where T
     return sample(mf, num_samples, num_warmup, save_warmup, thin, ss.adapt, ss.alg)
 end
 
-function sample(  mf::T,
-                  num_samples::Int,
-                  num_warmup::Int,
-                  save_warmup::Bool,
-                  thin::Int,
-                  adapt::CmdStan.Adapt,
-                  alg::CmdStan.Hmc
-                ) where T
+function sample(mf::T,
+    num_samples::Int,
+    num_warmup::Int,
+    save_warmup::Bool,
+    thin::Int,
+    adapt::CmdStan.Adapt,
+    alg::CmdStan.Hmc
+) where T
     if alg.stepsize_jitter != 0.0
         @warn("[Turing.sample] Turing does not support adding noise to stepsize yet.")
     end
     if adapt.engaged == false
         if isa(alg.engine, CmdStan.Static)   # hmc
             stepnum = Int(round(alg.engine.int_time / alg.stepsize))
-            sample(mf, HMC(num_samples, alg.stepsize, stepnum); adapt_conf=adapt)
+            sample(mf, HMC(num_samples, alg.stepsize, stepnum); adaptor=NUTSAdaptor(adapt))
         elseif isa(alg.engine, CmdStan.Nuts) # error
             error("[Turing.sample] CmdStan.Nuts cannot be used with adapt.engaged set as false")
         end
@@ -52,7 +55,8 @@ function sample(  mf::T,
                     adaptor=NUTSAdaptor(adapt))
         elseif isa(alg.engine, CmdStan.Nuts) # nuts
             if isa(alg.metric, CmdStan.diag_e)
-                sample(mf, NUTS(num_samples, num_warmup, adapt.delta);  adaptor=NUTSAdaptor(adapt))
+                sample(mf, NUTS(num_samples, num_warmup, adapt.delta);
+                        adaptor=NUTSAdaptor(adapt))
             else # TODO: reove the following since Turing support this feature now.
                 @warn("[Turing.sample] Turing does not support full covariance matrix for pre-conditioning yet.")
             end
@@ -60,7 +64,9 @@ function sample(  mf::T,
     end
 end
 
-function Sampler(alg::Hamiltonian, adaptor::CmdStanAdaptorType) where CmdStanAdaptorType
+function Sampler(alg::Hamiltonian,
+    adaptor::CmdStanAdaptorType
+) where {CmdStanAdaptorType}
     _sampler(alg::Hamiltonian, AHMCAdaptor(adaptor))
 end