tarekziade HF Staff
Keep the real checkpoint's weight names, prefix and file format (tensors unchanged; reload verified bit-exact)
874f11d verified | { | |
| "fixture": "huggingface-mic/MiniCPM3-4B-mic", | |
| "key": "097e769bf0bad3690f7fe52afd1acfa15eba8a2cdb27a65623aed8fabb7038a8", | |
| "key_fields": { | |
| "architectures": [ | |
| "MiniCPM3ForCausalLM" | |
| ], | |
| "output_hidden_states": false, | |
| "return_dict": true, | |
| "chunk_size_feed_forward": 0, | |
| "is_encoder_decoder": false, | |
| "id2label": { | |
| "0": "LABEL_0", | |
| "1": "LABEL_1" | |
| }, | |
| "label2id": { | |
| "LABEL_0": 0, | |
| "LABEL_1": 1 | |
| }, | |
| "problem_type": null, | |
| "vocab_size": 73448, | |
| "hidden_size": 64, | |
| "intermediate_size": 128, | |
| "num_hidden_layers": 2, | |
| "num_attention_heads": 4, | |
| "num_key_value_heads": 4, | |
| "hidden_act": "silu", | |
| "max_position_embeddings": 32768, | |
| "initializer_range": 0.1, | |
| "rms_norm_eps": 1e-05, | |
| "use_cache": true, | |
| "pad_token_id": null, | |
| "bos_token_id": 1, | |
| "eos_token_id": [ | |
| 2, | |
| 73440 | |
| ], | |
| "pretraining_tp": 1, | |
| "tie_word_embeddings": true, | |
| "rope_parameters": { | |
| "type": "longrope", | |
| "long_factor": [ | |
| 1.0591234137867171, | |
| 2.093982484148734, | |
| 10.475458000005451, | |
| 16.461961959767827 | |
| ], | |
| "short_factor": [ | |
| 1.0591234137867171, | |
| 2.093982484148734, | |
| 10.475458000005451, | |
| 16.461961959767827 | |
| ], | |
| "original_max_position_embeddings": 32768, | |
| "rope_theta": 10000.0, | |
| "rope_type": "longrope" | |
| }, | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "mlp_bias": false, | |
| "head_dim": 8, | |
| "kv_lora_rank": 16, | |
| "q_lora_rank": 16, | |
| "qk_nope_head_dim": 16, | |
| "qk_rope_head_dim": 8, | |
| "v_head_dim": 16, | |
| "scale_emb": 12, | |
| "scale_depth": 1.4, | |
| "dim_model_base": 256, | |
| "qk_head_dim": 96, | |
| "auto_map": { | |
| "AutoConfig": "configuration_minicpm.MiniCPM3Config", | |
| "AutoModel": "modeling_minicpm.MiniCPM3Model", | |
| "AutoModelForCausalLM": "modeling_minicpm.MiniCPM3ForCausalLM", | |
| "AutoModelForSeq2SeqLM": "modeling_minicpm.MiniCPM3ForCausalLM", | |
| "AutoModelForSequenceClassification": "modeling_minicpm.MiniCPM3ForSequenceClassification" | |
| }, | |
| "model_type": "minicpm3", | |
| "output_attentions": false, | |
| "dtype": "bfloat16" | |
| }, | |
| "covers": { | |
| "openbmb/MiniCPM3-4B": "d6b14ddaefdb11c624dd75c3c779549bc90b08cb" | |
| }, | |
| "built_from": { | |
| "repo": "openbmb/MiniCPM3-4B", | |
| "revision": "d6b14ddaefdb11c624dd75c3c779549bc90b08cb" | |
| }, | |
| "seed": 17, | |
| "dtype": "bfloat16", | |
| "builder": "example/build_fixtures.py", | |
| "model_class": "MiniCPM3ForCausalLM", | |
| "parameters": 4770688, | |
| "missing_parameters": [], | |
| "checkpoint_inspection": "parameter headers/indexes; bounded binary fallback", | |
| "numeric_health_scope": "CPU float32/bfloat16/float16; autocast for low-precision losses", | |
| "captured": { | |
| "call": "training step", | |
| "inputs": [ | |
| "attention_mask", | |
| "input_ids", | |
| "labels" | |
| ], | |
| "class": "MiniCPM3ForCausalLM" | |
| }, | |
| "weights": "random", | |
| "gate": { | |
| "random": { | |
| "pass": true, | |
| "accepted_insensitive": {}, | |
| "threshold": 0.001, | |
| "deltas": { | |
| "input_ids:roll_batch": 28.734375, | |
| "input_ids:reverse": 28.125, | |
| "attention_mask:all_ones": 1.58203125 | |
| }, | |
| "weakest": "attention_mask:all_ones", | |
| "no_perturbation_defined": [] | |
| } | |
| }, | |
| "numeric_health": { | |
| "float32": { | |
| "outputs_finite": true, | |
| "loss_finite": true, | |
| "gradients_finite": true, | |
| "finite": true | |
| }, | |
| "bfloat16": { | |
| "outputs_finite": true, | |
| "loss_finite": true, | |
| "gradients_finite": true, | |
| "finite": true | |
| }, | |
| "float16": { | |
| "outputs_finite": true, | |
| "loss_finite": true, | |
| "gradients_finite": true, | |
| "finite": true | |
| } | |
| }, | |
| "copied_unchanged": [ | |
| "added_tokens.json", | |
| "configuration_minicpm.py", | |
| "generation_config.json", | |
| "modeling_minicpm.py", | |
| "special_tokens_map.json", | |
| "tokenizer.json", | |
| "tokenizer.model", | |
| "tokenizer_config.json" | |
| ], | |
| "environment": { | |
| "python": "3.11.14", | |
| "machine": "arm64", | |
| "torch": "2.14.1", | |
| "transformers": "5.21.0.dev0", | |
| "tokenizers": "0.23.3" | |
| }, | |
| "sha256": { | |
| "added_tokens.json": "4760fcbf90bc193f33827ffe02f2e7ba1af1ec43644cc02ac22fdd611f6cca15", | |
| "config.json": "d108dde1ec2adebe8710c003aed1981d8d4e05cb1ba1e429f45a9404a4016ad6", | |
| "configuration_minicpm.py": "60c9d2cba0d4aebcd64e573cb072e80caf6b067f9bdfcb06079990c0d61a3bba", | |
| "generation_config.json": "4ae895d10225b5323fbce89f59fcbb2866f74c9a1008f4b1d3c8d6f6a1749591", | |
| "modeling_minicpm.py": "a964ac7a8a2fc5a6192b9abbf4b78c548dd1c6097c83f60d57705b5f8da44846", | |
| "pytorch_model.bin": "2c1e5f63363031a736014ae6c222c66a4059ca64e2935b81cb20e60a38270e8a", | |
| "special_tokens_map.json": "068594063e37662c02b21acf42ebb334ef6a74fb810e68a2368f88f08351de76", | |
| "tokenizer.json": "b00802b71a613e3f7df3899fe9643a3ff949736d333a2b892448a974383fe372", | |
| "tokenizer.model": "bb74d51116831c3bf65db812c553f94ab0c88dcf97a5bbb37e3504f6d359c530", | |
| "tokenizer_config.json": "25620d5a3f5727bba2fb403624f2c9a7bba55a7d00205829650cd1e3c646aae0" | |
| }, | |
| "serialization": { | |
| "file": "pytorch_model.bin", | |
| "reverted_conversions": [], | |
| "migrated_from": "model.safetensors" | |
| } | |
| } | |