v0.2.0: Separate Vision/Text/Overall Models, and better sharding!

What's Changed

• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #42
• feat: update docs to have the VisionModels, without the save_pretrained method for some reason by @pythoncrazy in #43
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #44
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #45
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #46
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #47
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #48
• Add claude GitHub actions 1760404803436 by @pythoncrazy in #52
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #49
• feat: switched to having the model parameter logic in a separate params.py by @pythoncrazy in #51
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #53
• feat: fixed the training loop code by @pythoncrazy in #54
• Logical axis annotations by @pythoncrazy in #55
• Logical axis annotations by @pythoncrazy in #56
• feat: made the model loading simpler, also implemented the CLIPTextMo… by @pythoncrazy in #57
• fix: got the training loop working, with the int issue by @pythoncrazy in #58
• Fix int issues by @pythoncrazy in #60
• feat: added from_config by @pythoncrazy in #63
• Distributed initialize fix by changing the RNGs to lazily init in the methods by @pythoncrazy in #65
• Splash attention by @pythoncrazy in #67
• Better sharding by @pythoncrazy in #68

Full Changelog: 0.1.0...0.2.0