| { | |
| "model_type": "hstu", | |
| "architectures": [ | |
| "HSTUForCausalLM" | |
| ], | |
| "num_hidden_layers": 12, | |
| "use_random_model": true, | |
| "max_seq_len": 8832, | |
| "hidden_size": 1024, | |
| "vocab_size": 1001024, | |
| "torch_dtype": "float32", | |
| "hstu_config": { | |
| "hidden_size": 1024, | |
| "num_layers": 12, | |
| "num_attention_heads": 8, | |
| "head_dim": 128, | |
| "norm_epsilon": 1e-05, | |
| "dtype": "float32", | |
| "residual": true, | |
| "is_causal": true, | |
| "has_ffn": true, | |
| "dropout_ratio": 0 | |
| }, | |
| "task_config": { | |
| "embedding_configs": [ | |
| { | |
| "feature_names": ["action_weights"], | |
| "table_name": "act", | |
| "vocab_size": 1024, | |
| "dim": 1024, | |
| "use_dynamicemb": false | |
| }, | |
| { | |
| "feature_names": ["video_id"], | |
| "table_name": "item", | |
| "vocab_size": 100000, | |
| "dim": 1024, | |
| "use_dynamicemb": true | |
| } | |
| ] | |
| }, | |
| "item_feature_name": "video_id", | |
| "action_feature_name": "action_weights" | |
| } |