v1.1.0

Changed

• Replace sync.Cond with a close of channel in the broadcasting mechanism of autograd ag package
• Refactor Matrix interface to extend from Tensor
• Replace ag.Node and ag.DualValue with mat.Tensor
• Breaking change: New single NewDense function accepting functional options (WithShape, WithBacking, WithGrad)
• Refactor BaseParam struct to embed mat.Matrix
• Remove Variable and enable gradient accumulation in Matrix
• Revise backward method and eliminate truncated backpropagation
• Backward function to require explicit output gradients to be accumulated beforehand
• Move package fn out of package ag and rename it to gradfn as mat subpackage

Removed

• Remove the parameter name to simplify the code and reduce unnecessary complexity (e.g., Introspect())
• Remove concurrent matrix-vector multiplication as it requires more careful considerations
• Remove the export feature in dot format
• Remove "spago" tag annotations on embeddings

v1.0.1

Added

• Method embeddings.Model.EmbeddingFast.

v1.0.0

First stable release!

Fixed

• Fix bug preventing the embeddings model from being traversed on nn.Apply.
• Fix incorrect use of self-attention cache when used for cross-attention.

Changed

• Optimize implementation of some Dense matrix functions, especially on
  amd64 with AVX.

v1.0.0-alpha

With this release we introduce breaking changes that bring significant
improvements to the project's structure, API and performance.

It would be difficult and confusing to list every single API change. Instead,
the following sections will broadly describe the most relevant changes,
arranged by topic.

Project structure

Until this release, the project was essentially a monorepo in disguise: the
core packages for handling matrices and computational graphs were accompanied
by many models implementations (from the very simple up to the most
sophisticated ones) and commands (models management utilities and servers).

We now prefer to keep in this very repository only the core components of spaGO,
only enriched with an (opinionated) set of popular models and functionalities.
Bigger sub-packages and related commands are moved to separate repositories.
The moved content includes, most notably, code related to Transformers
and Flair.
Please refer to the section Projects using Spago from the README for an
updated list of references to separate projects (note: some of them are still
work in progress).
If you have the feeling that something big is missing in spaGO, chances are
it was moved to one of these separate projects: just have a look there first.

The arrangement of packages has been simplified: there's no need anymore to
distinguish between cmd and pkg; all the main subpackages are located in
the project's root path. Similarly, many packages, previously nested under
pkg/ml, can now be found at root level too.

Go version and dependencies

The minimum required Go version is 1.18, primarily needed for the
introduction of type parameters (generics).

Thanks to the creation of separate projects, discussed above, and further
refactoring, the main set of required dependencies is limited to the ones
for testing.

Only the subpackage embeddings/store/diskstore requires something more, so
we defined it as "opt-in" submodule, with its own dependencies.

float32 vs. float64

Instead of separate packages mat32 and mat64, there is now a single unified
package mat. Many parts of the implementation make use of type parameters
(generics), however the package's public API makes a rather narrow use of them.

In particular, we abstained from adding type parameters to widely-used types,
such as the Matrix interface. Where suitable, we are simply favoring float64
values, the de-facto preferred floating point type in Go (just think about Go
math package). For other situations, we introduced a new subpackage
mat/float. It provides simple types, holding either float32 or float64
values, as scalars or slices, and makes it easy to convert values between
different precisions, all without making explicit use of generics.
This design prevents the excessive spreading of type arguments to tons of other
types that need to manipulate matrices, bot from other spaGO packages and
from your own code.

Matrices

• The type mat.Matrix is the primary interface for matrices and vectors
  throughout the project.
• The type mat.Dense is the concrete implementation for a dense matrix.
  Unlike the interface, it has a type argument to distinguish between float32
  and float64.
• We removed implementation and support for sparse matrices, since their
  efficacy and utility were marginal. A better implementation might come back
  in the future.
• A new dense matrix can be created "from scratch" by calling one of the several
  functions mat.New*** (NewDense, NewVecDense, ...). Here you must choose
  which data type to use, specifying it as type parameter (unless implicit).
• Once you have an existing matrix, you can create new instances preserving
  the same data type of the initial one: simply use one of the New*** methods
  on the matrix instance itself, rather than their top-level function
  counterparts.
• Any other operation performed on a matrix that creates a new instance will
  operate with the same type of the receiver, and returns an instance of that
  type too.
• Operations with matrices of different underlying data types are allowed, just
  beware the memory and computation overheads introduced by the necessary
  conversions.

Auto-grad package

• The package ag now implicitly works in "define-by-run" mode only.
  It's way more performant compared to the previous releases, and there would
  be no significant advantage in re-using a pre-defined graph ("define-and-run").
• There is no Graph anymore! At least, not as a first citizen: an implicit
  "virtual" graph is progressively formed each time an operation over some
  nodes is applied. The virtual graph can be observed by simply walking the
  tree of operations. Most methods of the former Graph are now simple
  functions in the ag package.
• We still provide a way to explicitly "free" some resources after use,
  both for helping the garbage collector and for returning some objects
  to their sync.Pool. The function ag.ReleaseGraph operates on the
  virtual graph described above, usually starting from the given output nodes.
• Forward operations are executed concurrently. As soon as an Operator is
  created (usually by calling one of the functions in ag, such as Add,
  Prod, etc.), the related Function's Forward procedure is performed
  on a new goroutine. Nevertheless, it's always safe to ask for the Operator's
  Value without worries: if it's called too soon, the function will lock
  until the result is computed, and only then return the value.
• To maximize performance, we removed the possibility to set a custom limit
  for concurrent computations. Thanks to the new design, we now let the Go
  runtime itself manage this problem for us, so that you can still limit
  and finetune concurrency with the GOMAXPROCS variable.
• The implementation of backpropagation is also redesigned and improved.
  Instead of invoking the backward procedure on an explicit Graph, you can call
  ag.Backward or ag.BackwardMany, specifying the output node (or nodes)
  of your computation (such as loss values, in traditional scenarios).
  The backward functions traverse the virtual graph and propagate the gradients,
  leveraging concurrency and making use of goroutines and locks in a way that's
  very similar to the forward procedure. The backward functions will lock and
  wait until the whole gradients propagation is complete before returning.
  The locking mechanism implemented in the nodes' Grad methods, will still
  prevent troubles in case your own code reads the gradients concurrently
  (that would be very uncommon).
• We also modified the implementation of time-steps handling and truncated
  backpropagation. Since we don't have the support of a concrete Graph
  structure anymore, we introduced a new dedicated type ag.TimeStepHandler,
  and related functions, such as NodeTimeStep. For performing a truncated
  backpropagation, we provide the function ag.BackwardT and
  ag.BackwardManyT: they work similarly to the normal backpropagation
  functions described above, only additionally requiring a time-step
  handler and the desired amount of back steps.
• We simplified and polished the API for creating new node-variables. Instead
  of having multiple functions for simple variables, scalars, constants,
  with/without name or grads, and various combination of those, you can now
  create any new variable with ag.Var, which accepts a Matrix value and
  creates a new node-variable with gradients accumulation disabled by default.
  To enable gradients propagation, or setting an explicit name (useful for
  model params or constants), you can use the Variable's chainable methods
  WithGrad and WithName. As a shortcut to create a scalar-matrix variable
  you can use ag.Scalar.
• The package ag/encoding provides generic structures and functions to obtain
  a sort of view of a virtual graph, with the goal of facilitating the
  encoding/marshaling of a graph in various formats.
  The package ag/encoding/dot is a rewriting of the former pkg/ml/graphviz,
  that uses the ag/encoding structures to represent a virtual graph in
  Graphviz DOT format.

Models

• As before, package nn provides types and functions for defining and
  handling models. Its subpackages are implementations of most common models.
  The set of built-in models has been remarkably revisited, moving some of them
  to separate projects, as previously explained.
• The Model interface has been extremely simplified: it only requires the
  special empty struct Module to be embedded in a model type. This is
  necessary only to distinguish an actual model from any other struct, which
  is especially useful for parameters traversal, or other similar operations.
• Since the Graph has been removed from ag, the models clearly don't need
  to hold a reference to it anymore. Similarly, there is no need for any other
  model-specific field, like the ones available from the former BaseModel.
  This implies the elimination of some seldomly used properties.
  Notable examples are the "processing mode" (from the old Graph) and the time
  step (from the old BaseModel).
  In situations where a removed value or feature is still needed, we suggest to
  either reintroduce the missing elements on the models that needs them, or
  to extract them to separate types and functions. An example of
  extracted behavior is the handling of time steps, already mentioned in the
  previous section.
• There is no distinction anymore between "pure" models and processors,
  making "reification" no longer necessary: once a model is created (or loaded),
  it can be immediately used, even for multiple concurrent inferences.
• A side effect of removing processor instances is that it's not possible
  to...

v0.7.0

Added

• New package ml/ag/encoding/dot, for simple serialization of a Graph to
  DOT (Graphviz) format.
• New package ml/nn/sgu, implementing a Spatial Gating Unit (SGU) model.
• New package ml/nn/conv1x1, implementing a simple 1-dimensional
  1-sized-kernel convolution model.
• New package ml/nn/gmlp, implementing a gMLP
  model.

Changed

• ml/nn/activation/Model.Forward now simply returns the input as it is if
  the activation function is the identity.

v0.6.0

Added

• ml/losses.WeightedCrossEntropy()
• ml/losses.FocalLoss()
• ml/losses.WeightedFocalLoss()
• nlp/sequencelabeler.LoadModel() (it replaces Load() and LoadEmbeddings())
• nlp/charlm.LoadModel()
• nlp/transformers/bert.Model.PredictMLM()
• nlp/transformers/bart/tasks package
• nlp/transformers/bert.Model.Vectorize()
• ml/ag.Graph.Nodes() and ml/ag.Nodes()
• ml/nn.Model.Close()
• ml/nn.ReifyForTraining() and ml/nn.ReifyForInference()
• ml/ag.Graph.Backward() now panics if it is executed with nodes belonging to
  different graphs.
• The new ml/graphviz package allows exporting a Graph to Graphviz
  DOT format. To make it possible,
  we introduced a new go-mod dependency gographviz.
• A custom name can be optionally set to a Graph's Variables. This can be
  useful for debugging purposes and visual graph representation.
  You can now use Graph.NewVariableWithName() and Graph.NewScalarWithName()
  to create named Variables, and get the name of a Variable with
  Variable.Name().

Changed

• All UnaryElementwise functions provided by the package ag/fn have been
  promoted to separate dedicated structs. This improves debuggability and you
  can get appropriate function names when using reflection. Here is the full
  list of the modified functions: Tan, Tanh, Sigmoid, HardSigmoid,
  HardTanh, ReLU, Softsign, Cos, Sin, Exp, Log, Neg,
  Reciprocal, Abs, Mish, GELU, Sqrt, Swish.
  For the same reason, a dedicated Square function is introduced, replacing
  Prod with both operands set to the same value.
• ml/ag types Operator, Variable, Wrapper are now public.
• ml/nn.Reify() now expects a Graph and a Processing Mode arguments
  instead of a Context object (removed).
• ml/nn.BaseModel has been modified, replacing the field Ctx Context with
  a direct reference to the model's Graph and the Processing Mode (fields G
  and ProcessingMode).
• Refactoring server implementation of nlp/sequencelabeler,
  nlp/transformers/bert, and nlp/transformers/bart.
• Upgrade various dependencies.
• Regenerate protocol buffers files (with protoc-gen-go v1.26.0 and
  protoc v3.16.0).

Removed

• nlp/sequencelabeler.Load() and LoadEmbeddings() (now replaced by
  nlp/sequencelabeler.LoadModel())
• ml/nn.Context (see related changes on Reify() and BaseModel)

v0.5.2

Added

• Handle multiple BERT pooling strategies (i.e. CLS_TOKEN, REDUCE_MEAN, REDUCE_MAX) in nlp.transformers.bert.server_encode.go.

v0.5.1

Added

• Add nlp.charlm.flair_converter.go to import Flair character language models.

Changed

• Improve nlp.transformer.generation algorithms:
  • optimize Generator.getTopKScoredTokens().
  • optimize Generator.updateTokensScores().
• Simplify mat32.Dense.Mul when doing Matrix-Vector multiplication.
• Refactor math32 functions using chewxy/math32 functions.
• Improve ag.Graph efficiency:
  • Use pre-computed cache doing ag.Graph.groupNodesByHeight().
  • Use sync.pool to reduce allocations of graph's operators.

Fixed

• Fix past key-values usage on self-attention and cross-attention

v0.5.0

Added

• Implement a beam-search algorithm for conditional generation:
  • nlp.transformer.generation package.
• Add implementation of the Sentence-Piece tokenizer:
  • nlp.tokenizers.sentencepiece package.
• BART improvements:
  • gRPC and HTTP API to perform Text Generation.
  • Add support for "Marian" architecture (used for translation tasks).
  • Add sinusoidal positional encoder (used by Marian).
  • Add "head" for conditional generation:
    • nlp.transformers.bart.head.conditionalgeneration package.
• Add nn.Closer interface (e.g. embeddings.Model needs to close the underlying key-value store).
• Add Swish act. function without trainable parameters.
• Add SiLU act. function (it is just an alias for Swish).
• New pe.SinusoidalPositionalEncoder (this implementation replaces unused pe.PositionalEncoder and pe.AxialPositionalEncoder)

Changed

• Update urfave/cli to v2.
• Update dgraph-io/badger to v3.
• Make the BART positional encoder an interface to support various encoding (i.e. trainable vs static).
• Rename to fn.NewSwish into fn.NewSwishB as this was the Swish variant with trainable parameters (B).
• Relax ag.GetOpName to match operator names in lower-case.
• Allow arbitrary activation function on BART encoder/decoder layers.
• Use precomputed "keys" and "values" in self-attention, multi-head attention and BART decoder.

Removed

• In relation to the aforementioned positional encoding changes:
  • pe.PositionalEncoder and related functions
  • pe.AxialPositionalEncoder and related functions

Fixed

• Fix causal-mask used by nn.ScaledDotProductAttention

v0.4.1

Added

• New function ReleaseMatrix to packages mat32 and mat64.
• New methods to Matrix interface, from mat32 and mat64: Minimum, Maximum, MulT, Inverse, DoNonZero. However, the implementation on sparse matrices is not implemented yet (it always panics).

Changed

• Prefer handling Matrix interface values over specific Dense or Sparse matrices, also avoiding unnecessary type casts. Relevant changes to the public API are listed below.
  • mat(32|64).Stack function's arguments and returned value are now Matrix interfaces, instead of explicit Dense matrices.
  • Dense.Minimum and Dense.Maximum, from packages mat32 and mat64, return a Matrix interface, instead of a specific Dense type.
  • The return values of fofe.EncodeDense, fofe.Encode, and fofe.BiEncode are slices of Matrix values, instead of Dense or Sparse.
  • The z argument of the function fofe.Decode is of type Matrix, instead of Dense.
  • ml.optimizers.de (Differential Evolution optimizer) API was changed handling Matrix values, instead of specific Dense matrices. Changes include: Member.TargetVector, Member.DonorVector, ScoredVector.Vector, the vector argument of NewMember function, the solution argument of score and validate functions passed to NewOptimizer.
  • PositionalEncoder.Cache and AxialPositionalEncoder.Cache are slices of Matrix, instead of slices of Dense.
  • AxialPositionalEncoder.EncodingAt returns a Matrix value, instead of Dense.
  • nn.DumpParamsVector returns a Matrix value, instead of Dense.
  • The vector argument of the function nn.LoadParamsVector is a Matrix, instead of Dense.
  • The value argument of the method embeddings.Model.SetEmbedding is of type Matrix, instead of Dense.
  • The type of the struct field evolvingembeddings.WordVectorPair.Vector is Matrix, instead of Dense.