RamTorch 0.1.0__tar.gz

ramtorch-0.1.0/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

ramtorch-0.1.0/PKG-INFO

@@ -0,0 +1,16 @@
+Metadata-Version: 2.4
+Name: RamTorch
+Version: 0.1.0
+Summary: RAM is All You Need
+Author-email: Lodestone <lodestone.rock@gmail.com>
+License: Apache-2.0
+Project-URL: Homepage, https://github.com/lodestone-rock/RamTorch
+Classifier: Programming Language :: Python :: 3
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Dynamic: license-file
+
+# RamTorch
+RAM is all you need

ramtorch-0.1.0/README.md

@@ -0,0 +1,2 @@
+# RamTorch
+RAM is all you need

ramtorch-0.1.0/RamTorch.egg-info/PKG-INFO

@@ -0,0 +1,16 @@
+Metadata-Version: 2.4
+Name: RamTorch
+Version: 0.1.0
+Summary: RAM is All You Need
+Author-email: Lodestone <lodestone.rock@gmail.com>
+License: Apache-2.0
+Project-URL: Homepage, https://github.com/lodestone-rock/RamTorch
+Classifier: Programming Language :: Python :: 3
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Dynamic: license-file
+
+# RamTorch
+RAM is all you need

ramtorch-0.1.0/RamTorch.egg-info/SOURCES.txt

@@ -0,0 +1,12 @@
+LICENSE
+README.md
+pyproject.toml
+RamTorch.egg-info/PKG-INFO
+RamTorch.egg-info/SOURCES.txt
+RamTorch.egg-info/dependency_links.txt
+RamTorch.egg-info/top_level.txt
+ramtorch/__init__.py
+ramtorch/helpers.py
+ramtorch/zero1.py
+ramtorch/modules/__init__.py
+ramtorch/modules/linear.py

ramtorch-0.1.0/RamTorch.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

ramtorch-0.1.0/RamTorch.egg-info/top_level.txt

@@ -0,0 +1 @@
+ramtorch

ramtorch-0.1.0/pyproject.toml

@@ -0,0 +1,19 @@
+[build-system]
+requires = ["setuptools>=45", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "RamTorch"
+version = "0.1.0"
+authors = [{name = "Lodestone", email = "lodestone.rock@gmail.com"}]
+description = "RAM is All You Need"
+readme = "README.md"
+license = {text = "Apache-2.0"}
+requires-python = ">=3.8"
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "Operating System :: OS Independent",
+]
+
+[project.urls]
+Homepage = "https://github.com/lodestone-rock/RamTorch"

ramtorch-0.1.0/ramtorch/modules/linear.py

@@ -0,0 +1,204 @@
+"""
+CPU Linear Module
+
+A memory-efficient linear layer implementation that keeps parameters on CPU
+and transfers them to GPU on-demand using asynchronous CUDA streams.
+
+This approach interleave compute and data transfer, making it useful for:
+- Very large models that don't fit in GPU memory
+- Scenarios where GPU memory is limited but CPU memory is abundant
+"""
+
+import os
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+# Global CUDA stream for asynchronous weight transfers
+# Using a dedicated stream allows transfers to overlap with computation
+TRANSFER_STREAM = torch.cuda.Stream()
+
+# Maximum number of in-flight transfers to prevent unbounded memory growth
+# Can be configured via environment variable
+MAX_INFLIGHT = int(os.getenv("MAX_INFLIGHT", 2))
+
+# Queue to track pending transfer events for synchronization
+PENDING_EVENTS = []
+
+
+class BouncingLinearFn(torch.autograd.Function):
+    """
+    Custom autograd function implementing the bouncing linear operation.
+
+    This function handles:
+    1. Asynchronous transfer of weights from CPU to GPU
+    2. Throttling of concurrent transfers to manage memory
+    3. Proper synchronization between transfer and compute streams
+    """
+
+    @staticmethod
+    def forward(ctx, x, weight_cpu, bias_cpu, device="cuda"):
+        """
+        Forward pass of bouncing linear layer.
+
+        Args:
+            ctx: PyTorch autograd context for saving backward pass info
+            x (torch.Tensor): Input tensor on GPU
+            weight_cpu (torch.Tensor): Weight matrix stored on CPU
+            bias_cpu (torch.Tensor, optional): Bias vector stored on CPU
+            device (str): Target GPU device for computation
+
+        Returns:
+            torch.Tensor: Linear transformation output (x @ weight.T + bias)
+
+        Flow:
+            1. Initiate async transfer of weights to GPU
+            2. Record completion event and add to pending queue
+            3. Throttle if too many transfers are in-flight
+            4. Wait for transfer completion before computation
+            5. Perform linear operation and return result
+        """
+        global PENDING_EVENTS
+
+        # enqueue transfer on transfer stream
+        with torch.cuda.stream(TRANSFER_STREAM):
+            w = weight_cpu.to(device, non_blocking=True)
+            b = bias_cpu.to(device, non_blocking=True) if bias_cpu is not None else None
+
+            # record event after transfer
+            evt = torch.cuda.Event()
+            evt.record(TRANSFER_STREAM)
+            PENDING_EVENTS.append(evt)
+
+        # throttle: wait if too many inflight
+        if len(PENDING_EVENTS) > MAX_INFLIGHT:
+            PENDING_EVENTS[0].synchronize()
+            PENDING_EVENTS.pop(0)
+
+        # make compute stream wait for this transfer
+        torch.cuda.current_stream().wait_event(evt)
+
+        out = F.linear(x, w, b)
+
+        # save for backward
+        ctx.save_for_backward(x, weight_cpu, bias_cpu)
+        ctx.device = device
+        return out
+
+    @staticmethod
+    def backward(ctx, grad_out):
+        """
+        Backward pass for gradient computation.
+
+        Args:
+            ctx: Autograd context containing saved forward pass data
+            grad_out (torch.Tensor): Gradient w.r.t. layer output
+
+        Returns:
+            tuple: Gradients w.r.t. (input, weight, bias, device)
+                  Device gradient is None (not differentiable)
+
+        Note:
+            Weights need to be transferred again for gradient computation
+            since they're not kept on GPU between forward and backward passes.
+        """
+        global PENDING_EVENTS
+        x, weight_cpu, bias_cpu = ctx.saved_tensors
+        device = ctx.device
+
+        # enqueue transfer on transfer stream
+        with torch.cuda.stream(TRANSFER_STREAM):
+            w = weight_cpu.to(device, non_blocking=True)
+            evt = torch.cuda.Event()
+            evt.record(TRANSFER_STREAM)
+            PENDING_EVENTS.append(evt)
+
+        # throttle: wait if too many inflight
+        if len(PENDING_EVENTS) > MAX_INFLIGHT:
+            PENDING_EVENTS[0].synchronize()
+            PENDING_EVENTS.pop(0)
+
+        torch.cuda.current_stream().wait_event(evt)
+
+        # grad computation
+        grad_input = grad_out @ w
+        grad_weight = grad_out.t() @ x
+        grad_bias = grad_out.sum(0) if bias_cpu is not None else None
+
+        return grad_input, grad_weight, grad_bias, None
+
+
+class CPUBouncingLinear(nn.Module):
+    """
+    Linear layer with CPU-stored parameters that bounce to GPU on demand.
+
+    This module provides a drop-in replacement for nn.Linear but with different
+    memory characteristics:
+    - Parameters stored on CPU (using shared memory for multiprocessing)
+    - Transferred to GPU only during forward/backward passes
+    - Automatic cleanup after each operation
+
+    Trade-offs:
+    + Drastically reduced GPU memory usage
+    + Enables training much larger models
+    - Requires batching to mask the latency
+
+    Best suited for:
+    - Models too large for GPU memory
+    - Inference scenarios with memory constraints
+    """
+
+    def __init__(self, in_features, out_features, bias=True, device="cuda"):
+        """
+        Initialize CPU linear layer.
+
+        Args:
+            in_features (int): Input feature dimension
+            out_features (int): Output feature dimension
+            bias (bool): Whether to include learnable bias term
+            device (str): Target GPU device for computation
+
+        Note:
+            Parameters are initialized on CPU with proper weight initialization.
+            share_memory_() enables efficient sharing in multiprocessing contexts.
+        """
+        super().__init__()
+        self.in_features = in_features
+        self.out_features = out_features
+        self.device = device
+
+        # parameters live on CPU
+        self.weight = nn.Parameter(
+            torch.empty(out_features, in_features, device="cpu").share_memory_()
+        )
+        self.bias = (
+            nn.Parameter(torch.empty(out_features, device="cpu").share_memory_())
+            if bias
+            else None
+        )
+
+        # init
+        nn.init.kaiming_uniform_(self.weight, a=5**0.5)
+        if self.bias is not None:
+            fan_in = in_features
+            bound = 1 / fan_in**0.5
+            nn.init.uniform_(self.bias, -bound, bound)
+
+    def forward(self, x):
+        """
+        Forward pass through CPU linear layer.
+
+        Args:
+            x (torch.Tensor): Input tensor (should be on GPU)
+
+        Returns:
+            torch.Tensor: Linear transformation output
+
+        Note:
+            Input tensor should already be on the target GPU device.
+            The autograd function handles all weight transfer logic.
+        """
+        return BouncingLinearFn.apply(x, self.weight, self.bias, self.device)
+
+
+Linear = CPUBouncingLinear

ramtorch-0.1.0/ramtorch/zero1.py

@@ -0,0 +1,66 @@
+from typing import List, Dict, Any, Tuple
+
+import torch
+import torch.distributed as dist
+
+
+def create_zero_param_groups(
+    param_groups: List[Dict[str, Any]], rank: int, world_size: int
+) -> Tuple[List[Dict[str, Any]], List[int]]:
+    """
+    Create parameter groups for ZeRO-1 optimizer sharding and generate a map
+    of parameter owners for broadcasting.
+
+    Args:
+        param_groups: Original parameter groups (list of dicts with 'params' key)
+        rank: Current process rank
+        world_size: Total number of processes
+
+    Returns:
+        A tuple containing:
+        - List of parameter groups containing only parameters owned by this rank.
+        - A list where the index corresponds to a parameter's global index and
+          the value is the rank of the process that owns it. This is the pre-computed
+          index for the broadcast function.
+    """
+    if not dist.is_initialized():
+        raise RuntimeError("torch.distributed must be initialized")
+
+    sharded_groups = []
+    owner_ranks = []
+    global_param_idx = 0
+
+    for group in param_groups:
+        # Copy all group settings except params
+        sharded_group = {k: v for k, v in group.items() if k != "params"}
+        sharded_group["params"] = []
+
+        # Add only parameters owned by this rank
+        for param in group["params"]:
+            owner_rank = global_param_idx % world_size
+            owner_ranks.append(owner_rank)
+
+            if owner_rank == rank:
+                sharded_group["params"].append(param)
+            global_param_idx += 1
+
+        # Only add group if it has parameters for this rank
+        if sharded_group["params"]:
+            sharded_groups.append(sharded_group)
+
+    return sharded_groups, owner_ranks
+
+
+def broadcast_zero_params(all_params: List[torch.Tensor], owner_ranks: List[int]):
+    """
+    Broadcast updated parameters from their owning ranks to all other ranks
+    using a pre-computed index of owner ranks.
+
+    Args:
+        all_params: List of all model parameters (in the same order across all ranks)
+        owner_ranks: A list mapping each parameter's global index to its owner rank.
+    """
+    with torch.no_grad():
+        for i, param in enumerate(all_params):
+            owner_rank = owner_ranks[i]
+            dist.broadcast(param.data, src=owner_rank)

ramtorch-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+