pilot.jsonl

{"adamw_implementation": "auto", "architecture": "gfx942:sramecc+:xnack-", "batch": 1, "bf16_training_mirror_bytes": 715653120, "bf16_training_mirror_tensors": 168, "bf16_weight_mirrors_enabled": true, "compute_dtype": "bf16_linear_fp32_master", "context": 512, "device": "hip:0", "device_name": "AMD Instinct MI300X VF", "engine_allocation_calls": 3392, "engine_backend_allocation_calls": 1350, "engine_backend_deallocation_calls": 0, "engine_cache_reuse_calls": 2042, "engine_cached_bytes": 5400799244, "engine_current_bytes": 8620181504, "engine_deallocation_calls": 3392, "engine_peak_bytes": 12673058824, "engine_reserved_bytes": 14020980748, "engine_total_allocated_bytes": 14876806168, "final_loss": 0.016907709, "first_loss": 0.049376152, "fp32_weight_bytes": 1976131072, "framework": "microllm", "hip_driver_version": 71399004, "hip_runtime_version": 71399004, "load_current_engine_bytes": 1976131072, "load_device_to_device_calls": 0, "load_device_to_host_calls": 0, "load_host_to_device_bytes": 988065536, "load_host_to_device_calls": 290, "load_ms": 553.549702, "load_peak_engine_bytes": 2248400384, "loaded_tensors": 290, "loss": 0.016907709, "mean_optimizer_ms": 599.9886335, "mean_step_ms": 743.3004405, "measured_ms": 1486.600881, "measurement_profile": "comparison", "milliseconds_per_token": 1.451758673, "model": "qwen2.5-0.5b", "observed_parameter_after": 7.468718052, "observed_parameter_before": 7.468739033, "optimizer_device_to_host_calls": 0, "optimizer_host_to_device_calls": 0, "optimizer_ms": 1199.977267, "pair_order": ["microllm", "pytorch"], "parameter_changed": true, "parameter_count": 494032768, "peak_bytes": 12673058824, "precision": "bf16", "process_run": 1, "record_type": "official_training_shape_measurement", "revision": "060db6499f32faf8b98477b0a26969ef7d8b9987", "schema_version": 1, "status": "pass", "step_ms": 1486.600881, "steps": 2, "tokens_per_second": 688.81971825, "trained_tokens": 1024, "warmup": 1, "warmup_ms": 8433.130711}
{"architecture": "gfx942:sramecc+:xnack-", "batch": 1, "compute_dtype": "bf16_amp", "context": 512, "device": "cuda", "device_current_allocated_bytes": 8080038912, "device_current_reserved_bytes": 11085545472, "device_discovery_workaround": "none", "device_name": "AMD Instinct MI300X VF", "device_peak_allocated_bytes": 10227678208, "device_peak_reserved_bytes": 11085545472, "final_loss": 0.018056392669677734, "first_loss": 0.0496223047375679, "fp32_weight_bytes": 1976131072, "framework": "pytorch", "load_ms": 1623.265370260924, "loaded_tensors": 290, "loss": 0.018056392669677734, "mean_optimizer_ms": 14.424551976844668, "mean_step_ms": 61.554206535220146, "measured_ms": 123.10841307044029, "measurement_profile": "comparison", "milliseconds_per_token": 0.12022305963910185, "mode": "train", "model": "qwen2.5-0.5b", "observed_parameter_after": 7.4687180519104, "observed_parameter_before": 7.468739032745361, "optimizer_ms": 28.849103953689337, "pair_order": ["microllm", "pytorch"], "parameter_changed": true, "parameter_count": 494032768, "peak_bytes": 10227678208, "precision": "bf16", "process_run": 1, "pytorch_state_tensors": 291, "record_type": "official_training_shape_measurement", "resident_weight_bytes": 1976131072, "revision": "060db6499f32faf8b98477b0a26969ef7d8b9987", "schema_version": 1, "status": "pass", "step_ms": 123.10841307044029, "steps": 2, "tokens_per_second": 8317.871820946037, "torch_hip_version": "7.13.99004", "torch_version": "2.10.0+rocm7.13.0rc2", "trained_tokens": 1024, "transformers_version": "5.8.1", "warmup": 1, "warmup_ms": 4107.168472837657}