wip

README.md

@@ -1,3 +1,17 @@
+
+
+setup
+~~~sh
+# curl -LsSf https://astral.sh/uv/install.sh | sh
+uv sync
+~~~
+
+https://github.com/casper-hansen/OpenCoconut
+https://github.com/lucidrains/coconut-pytorch
+https://github.com/wassname/latent-gemma
+
+
+----
 # Gemma 2: Japanese-English Reasoning
 
 A fine-tuned version of Google's Gemma 2 model enhanced with continuous latent reasoning capabilities, based on the COCONUT (Chain of Continuous Thought) paradigm introduced by Hao et al. (2024).
@@ -151,4 +165,4 @@ Training parameters:
 
 ## License
 
-This model inherits Gemma's license terms. See [Gemma License](https://www.kaggle.com/models/google/gemma/license) for details.
+This model inherits Gemma's license terms. See [Gemma License](https://www.kaggle.com/models/google/gemma/license) for details.

pyproject.toml

@@ -1,3 +1,11 @@
+[dependency-groups]
+dev = [
+    "ipykernel>=6.29.5",
+    "ipywidgets>=8.1.5",
+    "pytest>=8.0.2",
+    "ruff>=0.8.3",
+]
+
 [build-system]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
@@ -5,16 +13,25 @@ build-backend = "hatchling.build"
 [project]
 name = "latent-gemma"
 version = "0.1.0"
-authors = [
-    { name = "Victor Umesiobi", email = "vicksemmanuel58@gmail.com" },
-]
 description = "Latent reasoning implementation for Gemma models"
 readme = "README.md"
-requires-python = ">=3.8"
+requires-python = ">=3.10"
 dependencies = [
-    "torch>=2.0.0",
-    "transformers>=4.0.0"
+    "numpy>=1.26.1",
+    "pandas>=2.1.1",
+    "matplotlib>=3.8.0",
+    "scikit-learn>=1.3.1",
+    "loguru>=0.7.2",
+    "tqdm>=4.66.1",
+    "transformers[torch]>=4.34.0",
+    # "einops>=0.7.0",
+    # "loguru>=0.7.2",
+    # "simple-parsing>=0.1.4",
+    # "torch >=2.1.0+cu124",
+    # "lightning>=2.1.0",
+    # "torchinfo>=1.8.0",
+    # "peft>=0.5.0",
 ]
 
 [project.urls]
-Homepage = "https://github.com/vicksEmmanuel/latent-gemma"
+Homepage = "https://github.com/wassname/latent-gemma"