markovgpu-rane 0.1.0__py3-none-any.whl

markovgpu/__init__.py

@@ -0,0 +1,3 @@
+from .backend import MarkovEngine
+
+__all__ = ["MarkovEngine"]

markovgpu/backend.py

@@ -0,0 +1,326 @@
+import pyopencl as cl
+import numpy as np
+import os
+
+# Locate the kernel file
+MODULE_PATH = os.path.dirname(os.path.abspath(__file__))
+KERNEL_PATH = os.path.join(MODULE_PATH, "kernels.cl")
+
+# Threshold: Use GPU if states >= 64, otherwise CPU is faster
+GPU_THRESHOLD = 64
+
+
+class MarkovEngine:
+    def __init__(self):
+        self.use_gpu = False
+        self.ctx = None
+        self.queue = None
+        self.prg = None
+
+        # 1. Try to Connect to GPU
+        try:
+            platforms = cl.get_platforms()
+            gpu_devices = []
+            for p in platforms:
+                gpu_devices.extend(p.get_devices(device_type=cl.device_type.GPU))
+
+            if gpu_devices:
+                # Pick the discrete GPU (highest compute units)
+                best_dev = sorted(
+                    gpu_devices, key=lambda d: d.max_compute_units, reverse=True
+                )[0]
+                self.ctx = cl.Context([best_dev])
+                print(
+                    f"🔌 Connected to Accelerator: {best_dev.name} ({best_dev.max_compute_units} CUs)"
+                )
+            else:
+                self.ctx = cl.create_some_context(interactive=False)
+                print(f"⚠️ No Dedicated GPU found. Using: {self.ctx.devices[0].name}")
+
+            self.queue = cl.CommandQueue(self.ctx)
+
+            # 2. Compile Kernels
+            if not os.path.exists(KERNEL_PATH):
+                raise FileNotFoundError(f"Kernel file missing at: {KERNEL_PATH}")
+
+            with open(KERNEL_PATH, "r") as f:
+                self.prg = cl.Program(self.ctx, f.read()).build()
+
+            # 3. Cache Kernels (Robust Retrieval)
+            self.use_gpu = True
+            try:
+                # Basic
+                self.k_markov = self.prg.markov_step
+                self.k_hmm_basic = self.prg.hmm_forward_step
+
+                # Advanced / Viterbi
+                self.k_hmm_log = self.prg.hmm_forward_log
+                self.k_viterbi = self.prg.viterbi_step
+
+                # Training
+                self.k_hmm_back = self.prg.hmm_backward_log
+                self.k_acc_trans = self.prg.accumulate_transitions
+                self.k_acc_gamma = self.prg.accumulate_gammas
+
+            except AttributeError as e:
+                print(f"❌ Kernel Warning: {e}")
+                print("⚠️ Some GPU features may be disabled.")
+
+        except Exception as e:
+            print(f"⚠️ OpenCL Initialization failed: {e}")
+            print("⚠️ Running in CPU-Only Mode (NumPy).")
+            self.use_gpu = False
+
+    # --- 1. Simulation ---
+    def step(self, P, v):
+        """Runs one step: v_new = v * P"""
+        N = len(v)
+
+        if not self.use_gpu or N < GPU_THRESHOLD:
+            return v.dot(P)
+
+        mf = cl.mem_flags
+        P = np.ascontiguousarray(P, dtype=np.float32)
+        v = np.ascontiguousarray(v, dtype=np.float32)
+        result = np.empty_like(v)
+
+        d_P = cl.Buffer(self.ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=P)
+        d_v = cl.Buffer(self.ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=v)
+        d_res = cl.Buffer(self.ctx, mf.WRITE_ONLY, size=result.nbytes)
+
+        self.k_markov(self.queue, (N,), None, np.int32(N), d_v, d_P, d_res)
+        cl.enqueue_copy(self.queue, result, d_res)
+        return result
+
+    def converge(self, P, start_v, tolerance=1e-5, max_steps=1000):
+        N = len(start_v)
+
+        # CPU Path
+        if not self.use_gpu or N < GPU_THRESHOLD:
+            # print(f"🔄 Converging on CPU (N={N})...")
+            current_v = start_v.copy()
+            for i in range(max_steps):
+                new_v = current_v.dot(P)
+                if np.sum(np.abs(new_v - current_v)) < tolerance:
+                    return new_v
+                current_v = new_v
+            return current_v
+
+        # GPU Path
+        # print(f"🔄 Converging on GPU (N={N})...")
+        mf = cl.mem_flags
+        P = np.ascontiguousarray(P, dtype=np.float32)
+        start_v = np.ascontiguousarray(start_v, dtype=np.float32)
+
+        d_P = cl.Buffer(self.ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=P)
+        d_v_read = cl.Buffer(
+            self.ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=start_v
+        )
+        d_v_write = cl.Buffer(self.ctx, mf.READ_WRITE, size=start_v.nbytes)
+
+        current_v = start_v.copy()
+
+        for i in range(max_steps):
+            self.k_markov(self.queue, (N,), None, np.int32(N), d_v_read, d_P, d_v_write)
+
+            if i % 10 == 0:
+                new_v = np.empty_like(current_v)
+                cl.enqueue_copy(self.queue, new_v, d_v_write)
+                if np.sum(np.abs(new_v - current_v)) < tolerance:
+                    return new_v
+                current_v = new_v
+
+            d_v_read, d_v_write = d_v_write, d_v_read
+
+        print("⚠️ Reached max steps without full convergence.")
+        return current_v
+
+    # --- 2. Inference & Viterbi ---
+    def hmm_filter(self, transition_matrix, observation_probs):
+        """Standard HMM Filter (Returns Probabilities)"""
+        # Simplification: Running basic HMM forward pass
+        # For production use, usually prefer Log-Space to avoid underflow.
+        # This wrapper can be upgraded to use k_hmm_log if needed.
+        pass
+
+    def decode_regime(self, transition_matrix, observation_probs):
+        """Viterbi Algorithm (Finds Most Likely Path)"""
+        T, N = observation_probs.shape
+        epsilon = 1e-20
+
+        # CPU Path
+        if not self.use_gpu or N < GPU_THRESHOLD:
+            log_trans = np.log(transition_matrix + epsilon)
+            log_emis = np.log(observation_probs + epsilon)
+            log_delta = np.zeros((T, N))
+            backpointers = np.zeros((T, N), dtype=int)
+
+            log_delta[0] = -np.log(N) + log_emis[0]
+
+            for t in range(1, T):
+                for j in range(N):
+                    vals = log_delta[t - 1] + log_trans[:, j]
+                    best_prev = np.argmax(vals)
+                    backpointers[t, j] = best_prev
+                    log_delta[t, j] = vals[best_prev] + log_emis[t, j]
+
+            path = np.zeros(T, dtype=int)
+            path[-1] = np.argmax(log_delta[-1])
+            for t in range(T - 2, -1, -1):
+                path[t] = backpointers[t + 1, path[t + 1]]
+            return path
+
+        # GPU Path
+        mf = cl.mem_flags
+        log_trans = np.log(transition_matrix + epsilon).astype(np.float32)
+        log_emis = np.log(observation_probs + epsilon).astype(np.float32)
+        log_delta = np.full(N, -np.log(N), dtype=np.float32)
+
+        d_trans = cl.Buffer(
+            self.ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=log_trans
+        )
+        d_delta_in = cl.Buffer(
+            self.ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=log_delta
+        )
+        d_delta_out = cl.Buffer(self.ctx, mf.READ_WRITE, size=log_delta.nbytes)
+
+        full_backpointer_history = np.zeros((T, N), dtype=np.int32)
+        d_backpointers = cl.Buffer(
+            self.ctx, mf.WRITE_ONLY, size=full_backpointer_history.nbytes // T
+        )
+
+        print(f"🕵️ Decoding {T} days (GPU Accelerated)...")
+
+        for t in range(T):
+            d_emis = cl.Buffer(
+                self.ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=log_emis[t]
+            )
+
+            self.k_viterbi(
+                self.queue,
+                (N,),
+                None,
+                np.int32(N),
+                d_delta_in,
+                d_trans,
+                d_emis,
+                d_delta_out,
+                d_backpointers,
+            )
+
+            step_pointers = np.empty(N, dtype=np.int32)
+            cl.enqueue_copy(self.queue, step_pointers, d_backpointers)
+            full_backpointer_history[t] = step_pointers
+
+            d_delta_in, d_delta_out = d_delta_out, d_delta_in
+
+        final_log_probs = np.empty(N, dtype=np.float32)
+        cl.enqueue_copy(self.queue, final_log_probs, d_delta_in)
+
+        best_path = np.zeros(T, dtype=np.int32)
+        best_path[-1] = np.argmax(final_log_probs)
+
+        for t in range(T - 2, -1, -1):
+            next_state = best_path[t + 1]
+            best_path[t] = full_backpointer_history[t + 1][next_state]
+
+        return best_path
+
+    # --- 3. Training (Baum-Welch) ---
+    def fit(self, observations, n_states, n_iters=10, tolerance=1e-4):
+        """Baum-Welch Expectation Maximization (Training)"""
+        T = observations.shape[0]
+        N = n_states
+
+        # Random Init
+        log_trans = np.log(
+            np.full((N, N), 1.0 / N) + np.random.rand(N, N) * 0.01
+        ).astype(np.float32)
+        log_emis = np.log(observations + 1e-20).astype(np.float32)
+
+        mf = cl.mem_flags
+        d_trans = cl.Buffer(
+            self.ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=log_trans
+        )
+        d_alpha = cl.Buffer(self.ctx, mf.READ_WRITE, size=T * N * 4)  # Full history
+        d_beta = cl.Buffer(self.ctx, mf.READ_WRITE, size=T * N * 4)  # Full history
+        d_emis = cl.Buffer(self.ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=log_emis)
+
+        d_new_trans = cl.Buffer(self.ctx, mf.READ_WRITE, size=log_trans.nbytes)
+        d_gamma_sums = cl.Buffer(self.ctx, mf.READ_WRITE, size=N * 4)
+
+        prev_score = -np.inf
+
+        print(f"🧠 Training HMM ({N} States, {T} Steps)...")
+
+        for i in range(n_iters):
+            # 1. CPU Forward/Backward (Latency Optimized)
+            alpha_full, log_likelihood = self._cpu_forward(log_trans, log_emis)
+            beta_full = self._cpu_backward(log_trans, log_emis)
+
+            # 2. GPU Accumulation (Throughput Optimized)
+            cl.enqueue_copy(self.queue, d_alpha, alpha_full)
+            cl.enqueue_copy(self.queue, d_beta, beta_full)
+            cl.enqueue_copy(self.queue, d_trans, log_trans)
+
+            self.k_acc_trans(
+                self.queue,
+                (N, N),
+                None,
+                np.int32(T),
+                np.int32(N),
+                d_alpha,
+                d_beta,
+                d_emis,
+                d_trans,
+                d_new_trans,
+            )
+
+            self.k_acc_gamma(
+                self.queue,
+                (N,),
+                None,
+                np.int32(T),
+                np.int32(N),
+                d_alpha,
+                d_beta,
+                d_gamma_sums,
+            )
+
+            # 3. Update
+            new_log_trans_counts = np.empty_like(log_trans)
+            log_gamma_sums = np.empty(N, dtype=np.float32)
+
+            cl.enqueue_copy(self.queue, new_log_trans_counts, d_new_trans)
+            cl.enqueue_copy(self.queue, log_gamma_sums, d_gamma_sums)
+
+            log_trans = new_log_trans_counts - log_gamma_sums[:, None]
+
+            change = log_likelihood - prev_score
+            print(
+                f"   Iter {i + 1}: Likelihood {log_likelihood:.2f} (Delta: {change:.4f})"
+            )
+            if abs(change) < tolerance:
+                break
+            prev_score = log_likelihood
+
+        return np.exp(log_trans)
+
+    def _cpu_forward(self, log_trans, log_emis):
+        T, N = log_emis.shape
+        alpha = np.zeros((T, N), dtype=np.float32)
+        alpha[0] = -np.log(N) + log_emis[0]
+        for t in range(1, T):
+            for j in range(N):
+                prev = alpha[t - 1] + log_trans[:, j]
+                alpha[t, j] = np.logaddexp.reduce(prev) + log_emis[t, j]
+        return alpha, np.logaddexp.reduce(alpha[-1])
+
+    def _cpu_backward(self, log_trans, log_emis):
+        T, N = log_emis.shape
+        beta = np.zeros((T, N), dtype=np.float32)
+        for t in range(T - 2, -1, -1):
+            for i in range(N):
+                terms = log_trans[i, :] + log_emis[t + 1] + beta[t + 1]
+                beta[t, i] = np.logaddexp.reduce(terms)
+        return beta

markovgpu/kernels.cl

@@ -0,0 +1,169 @@
+// kernels.cl - The Complete Suite
+
+// --- HELPER: Log-Sum-Exp Trick ---
+// Prevents overflow when adding log-probabilities
+float log_add(float log_a, float log_b) {
+    float max_val = max(log_a, log_b);
+    float min_val = min(log_a, log_b);
+    return max_val + log1p(exp(min_val - max_val));
+}
+
+// --- SECTION 1: Basic Markov Operations ---
+
+// 1. Standard Markov Step: Next = Current * Matrix
+__kernel void markov_step(
+    const int N,
+    __global const float *current_state,
+    __global const float *transition_mat,
+    __global float *next_state)
+{
+    int id = get_global_id(0);
+    if (id < N) {
+        float sum = 0.0f;
+        for (int k = 0; k < N; k++) {
+            sum += current_state[k] * transition_mat[k * N + id];
+        }
+        next_state[id] = sum;
+    }
+}
+
+// 2. Standard HMM Filter (Probability Space)
+// Used for simple "What state am I in?" queries without log-space
+__kernel void hmm_forward_step(
+    const int N,
+    __global const float *alpha_prev,
+    __global const float *trans_mat,
+    __global const float *emissions,
+    __global float *alpha_new)
+{
+    int id = get_global_id(0);
+    if (id < N) {
+        float sum = 0.0f;
+        for (int k = 0; k < N; k++) {
+            sum += alpha_prev[k] * trans_mat[k * N + id];
+        }
+        alpha_new[id] = sum * emissions[id];
+    }
+}
+
+// --- SECTION 2: Advanced Log-Space Operations (Stable) ---
+
+// 3. Log-Space Forward (For Viterbi & Training)
+__kernel void hmm_forward_log(
+    const int N,
+    __global const float *log_alpha_prev, 
+    __global const float *log_trans_mat,  
+    __global const float *log_emissions,  
+    __global float *log_alpha_new)        
+{
+    int id = get_global_id(0);
+    if (id < N) {
+        float log_sum = -INFINITY;
+        for (int k = 0; k < N; k++) {
+            float val = log_alpha_prev[k] + log_trans_mat[k * N + id];
+            if (k == 0) log_sum = val;
+            else log_sum = log_add(log_sum, val);
+        }
+        log_alpha_new[id] = log_sum + log_emissions[id];
+    }
+}
+
+// 4. Log-Space Backward (For Training)
+__kernel void hmm_backward_log(
+    const int N, 
+    __global const float *beta_future, 
+    __global const float *trans, 
+    __global const float *emis_future, 
+    __global float *beta_curr) 
+{
+    int id = get_global_id(0); // State 'i'
+    if (id < N) {
+        float log_sum = -INFINITY;
+        for (int j=0; j<N; j++) {
+            // transition i->j + emission(t+1) + beta(t+1)
+            float val = trans[id*N + j] + emis_future[j] + beta_future[j];
+            if (j==0) log_sum = val;
+            else log_sum = log_add(log_sum, val);
+        }
+        beta_curr[id] = log_sum;
+    }
+}
+
+// 5. Viterbi Algorithm (Finds best path)
+__kernel void viterbi_step(
+    const int N,
+    __global const float *log_delta_prev,
+    __global const float *log_trans_mat,
+    __global const float *log_emissions,
+    __global float *log_delta_new,       
+    __global int *backpointers)          
+{
+    int id = get_global_id(0);
+    if (id < N) {
+        float max_prob = -INFINITY;
+        int best_prev_state = 0;
+
+        for (int k = 0; k < N; k++) {
+            float prob = log_delta_prev[k] + log_trans_mat[k * N + id];
+            if (prob > max_prob) {
+                max_prob = prob;
+                best_prev_state = k;
+            }
+        }
+        log_delta_new[id] = max_prob + log_emissions[id];
+        backpointers[id] = best_prev_state; 
+    }
+}
+
+// --- SECTION 3: Learning Accumulators (Baum-Welch) ---
+
+// 6. Accumulate Transitions (E-Step)
+// Condenses time T into N*N summary matrix
+__kernel void accumulate_transitions(
+    const int T, const int N,
+    __global const float *alpha_full, 
+    __global const float *beta_full,  
+    __global const float *emis_full,  
+    __global const float *trans_mat,  
+    __global float *new_trans_counts) 
+{
+    int row = get_global_id(1); // From State i
+    int col = get_global_id(0); // To State j
+
+    if (row < N && col < N) {
+        float log_sum_xi = -INFINITY;
+        float log_trans_val = trans_mat[row * N + col];
+
+        // Loop over time 0 to T-2
+        for (int t = 0; t < T - 1; t++) {
+            float log_xi = alpha_full[t*N + row] + 
+                           log_trans_val + 
+                           emis_full[(t+1)*N + col] + 
+                           beta_full[(t+1)*N + col];
+            
+            if (t == 0) log_sum_xi = log_xi;
+            else log_sum_xi = log_add(log_sum_xi, log_xi);
+        }
+        new_trans_counts[row * N + col] = log_sum_xi;
+    }
+}
+
+// 7. Accumulate Gammas (E-Step)
+// Condenses time T into N summary counts
+__kernel void accumulate_gammas(
+    const int T, const int N,
+    __global const float *alpha_full,
+    __global const float *beta_full,
+    __global float *log_gamma_sums) 
+{
+    int id = get_global_id(0);
+    if (id < N) {
+        float log_sum_gamma = -INFINITY;
+        for (int t = 0; t < T; t++) {
+            float val = alpha_full[t*N + id] + beta_full[t*N + id];
+            if (t == 0) log_sum_gamma = val;
+            else log_sum_gamma = log_add(log_sum_gamma, val);
+        }
+        log_gamma_sums[id] = log_sum_gamma;
+    }
+}

markovgpu_rane-0.1.0.dist-info/METADATA

@@ -0,0 +1,22 @@
+Metadata-Version: 2.4
+Name: markovgpu-rane
+Version: 0.1.0
+Summary: High-performance Markov Chains & HMMs using OpenCL
+Author-email: Sahil Rane <sahilrane249@gmail.com>
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Financial and Insurance Industry
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Scientific/Engineering :: Mathematics
+Requires-Python: >=3.12
+Requires-Dist: matplotlib>=3.8.0
+Requires-Dist: numpy>=1.26.0
+Requires-Dist: pyopencl>=2024.1
+Requires-Dist: scipy>=1.11.0
+Description-Content-Type: text/markdown
+
+hello

markovgpu_rane-0.1.0.dist-info/RECORD

@@ -0,0 +1,7 @@
+markovgpu/__init__.py,sha256=aGWvFGT6VaLCuFNO9T3ubnlhz2qgkBmNIcCy976YrqE,62
+markovgpu/backend.py,sha256=bfYnge9MgMcDHmJ7CcCG2VGqVfsGxsCzXavGLUFdB2w,11733
+markovgpu/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+markovgpu/kernels.cl,sha256=RXpt2jD6IRdh5YTunB_lwfajT1Cw9M95v3uxwMMmMvs,5141
+markovgpu_rane-0.1.0.dist-info/METADATA,sha256=-CqRDK-d95CjNbsFpBIBut--zqLGHvvXPgsWdUe0Mtg,840
+markovgpu_rane-0.1.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+markovgpu_rane-0.1.0.dist-info/RECORD,,

markovgpu_rane-0.1.0.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.28.0
+Root-Is-Purelib: true
+Tag: py3-none-any