CounselTrajLM / model_details.json
ZhiyangQi97's picture
Release merged BF16 CounselTrajLM based on Qwen3.5-9B
9f553e1 verified
Raw History Blame Contribute Delete
1.32 kB
{
"base_model": "Qwen/Qwen3.5-9B",
"base_revision": "c202236235762e1c871ad0ccb60c8ee5ba337b9a",
"rank": 32,
"alpha": 64,
"dropout": 0.05,
"adapter_parameters": 58195968,
"selection": "Final checkpoint after three epochs; no best-validation checkpoint reload.",
"epoch": 3.0,
"global_step": 79815,
"train_samples": 425679,
"validation_samples": 47528,
"validation_loss": 1.5481888055801392,
"training_hyperparameters": {
"max_seq_length": 2048,
"max_history_chars": 1800,
"learning_rate": 0.0002,
"warmup_ratio": 0.06,
"optim": "paged_adamw_8bit",
"lr_scheduler_type": "cosine",
"per_device_train_batch_size": 2,
"gradient_accumulation_steps": 1,
"num_train_epochs": 3.0,
"attn_implementation": "eager",
"seed": 20260704
},
"release_changes": [
"Replace the internal base-model path with its official Hub ID and pin the exact base revision.",
"Merge the published final adapter into the unquantized pinned base text language model."
],
"training_adapter_type": "LoRA",
"training_checkpoint": "final_adapter",
"release_format": "merged BF16 text language model",
"source_adapter_sha256": "c4d89ce3b6cfdb53dba2927683331e2182c8767f3380527a78f0533b0f31e5bd",
"source_adapter_hf_revision": "341e9b198c985ee75c982a43376015d806b15522"
}