transformers
77f95d61 - [InternVL] Fix stale expected values for Llama integration tests (cuda sm_80)

Commit
93 days ago
[InternVL] Fix stale expected values for Llama integration tests (cuda sm_80) The expected logits and generated text for `("cuda", 8)` (Ampere GPUs) in `test_llama_small_model_integration_forward` and `test_llama_small_model_integration_generate_text_only` were stale after the CI docker was updated to torch 2.10 (commit c0fe6164d0, "Update Nvidia CI docker file to use torch 2.10"). The actual model outputs are identical across all code commits; the drift is purely due to the torch version change. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Author
Parents
Loading