{ "mic": "0.3", "name": "PegasusX summarization", "usage": "AutoTokenizer + AutoModelForSeq2SeqLM, batched, padded, generate + one training step", "architecture": "PegasusXForConditionalGeneration", "checkpoints": { "google/pegasus-x-base": { "fixture": "huggingface-mic/pegasus-x-base-mic" } }, "fixtures": { "huggingface-mic/pegasus-x-base-mic": { "built_from": "google/pegasus-x-base", "covers": [ "google/pegasus-x-base" ], "why_separate": "The full checkpoint is 12 layers and 768-d; the fixture is a slim 2-layer 64-d model for CPU tests.", "build": { "seed": 17, "model_class": "PegasusXForConditionalGeneration", "config_overrides": { "fields": { "encoder_layers": 2, "decoder_layers": 2, "d_model": 64, "encoder_attention_heads": 4, "decoder_attention_heads": 4, "encoder_ffn_dim": 128, "decoder_ffn_dim": 128 }, "reason": "Shrink the 12-layer 768-d model to a 2-layer 64-d fixture for CPU tests; keep vocab and token ids intact." } } } }, "integrations": { "transformers": { "entry_point": "integrations/transformers/usage.md", "format": "runnable-markdown", "requires": [ "transformers>=5", "torch" ], "targets": [ "fixtures", "checkpoints" ], "variants": { "default": {}, "eager": { "attn_implementation": "eager" }, "bfloat16": { "dtype": "bfloat16" } }, "training": true, "resources": { "fixture": { "download_mb": 20 }, "google/pegasus-x-base": { "memory_gb": 8, "download_mb": 2200, "gpu_memory_gb": 12 } } } }, "outputs": { "runtime": { "compare": "exact" }, "summary_ids": { "compare": "exact" }, "summary_text": { "compare": "exact" }, "logits.probe": { "compare": "allclose", "atol": 0.0001, "rtol": 0.0001 }, "training.loss": { "compare": "allclose", "atol": 0.0001, "rtol": 0.0001 }, "training.gradients": { "compare": "all_true" }, "training.updated": { "compare": "true" } } }