V3 gpu flash attention triton solution fix

[jeniag]

Existing solution in v3/gpu-systems/flash-attention-triton/flash-attention-triton_SOLN.ipynb only has a flash_attention_kernel not a full wrapper. This adds the wrapper.

To test:
from torchleet import check; check("flash-attention-triton", flash_attention_pytorch,flash_attention_triton=flash_attention_triton)

Passes with

torchleet: flash-attention-triton
   ✅ matches standard attention
   ✅ block size does not change the answer
   ✅ sequence not divisible by block
   ✅ numerically stable on large scores
   ✅ attention weights are a distribution
   ✅ gradients flow
   ✅ triton kernel
   all checks passed

True

[vercel]

@jeniag is attempting to deploy a commit to the exorust's projects Team on Vercel.

A member of the Team first needs to authorize it.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB