{
 "fixture": "huggingface-mic/starcoder2-3b-mic",
 "key": "0b99c4e71ec5e3b22333a63d3cd2bb97d2071ea6fba36b6a6a1ec6b35c1666ff",
 "key_fields": {
  "architectures": [
   "Starcoder2ForCausalLM"
  ],
  "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": 49152,
  "hidden_size": 64,
  "intermediate_size": 128,
  "num_hidden_layers": 2,
  "num_attention_heads": 4,
  "num_key_value_heads": 2,
  "hidden_act": "gelu_pytorch_tanh",
  "max_position_embeddings": 16384,
  "initializer_range": 0.018042,
  "norm_epsilon": 1e-05,
  "use_cache": true,
  "bos_token_id": 0,
  "eos_token_id": 0,
  "pad_token_id": null,
  "rope_parameters": {
   "rope_theta": 999999.4420358813,
   "rope_type": "default"
  },
  "sliding_window": 4096,
  "attention_dropout": 0.1,
  "residual_dropout": 0.1,
  "embedding_dropout": 0.1,
  "use_bias": true,
  "tie_word_embeddings": true,
  "mlp_type": "default",
  "model_type": "starcoder2",
  "norm_type": "layer_norm",
  "output_attentions": false
 },
 "covers": {
  "bigcode/starcoder2-3b": "733247c55e3f73af49ce8e9c7949bf14af205928"
 },
 "built_from": {
  "repo": "bigcode/starcoder2-3b",
  "revision": "733247c55e3f73af49ce8e9c7949bf14af205928"
 },
 "seed": 17,
 "dtype": "float32",
 "builder": "example/build_fixtures.py",
 "builder_revision": "ed107e55ab1c53dc35febfa2a8ed87a0a5a42a3c",
 "recipe": {
  "seed": 17,
  "model_class": "Starcoder2ForCausalLM"
 },
 "shrink": "rule",
 "model_class": "Starcoder2ForCausalLM",
 "parameters": 3204480,
 "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": "Starcoder2ForCausalLM"
 },
 "weights": "random",
 "gate": {
  "random": {
   "pass": true,
   "accepted_insensitive": {},
   "threshold": 0.001,
   "deltas": {
    "input_ids:roll_batch": 1.2845580577850342,
    "input_ids:reverse": 1.1660540103912354,
    "attention_mask:all_ones": 0.823656439781189
   },
   "weakest": "attention_mask:all_ones",
   "no_perturbation_defined": [],
   "not_a_forward_parameter": []
  }
 },
 "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,
   "unscaled_gradients_finite": true,
   "finite": true
  }
 },
 "copied_unchanged": [
  "merges.txt",
  "special_tokens_map.json",
  "tokenizer.json",
  "tokenizer_config.json",
  "vocab.json"
 ],
 "environment": {
  "python": "3.12.12",
  "machine": "arm64",
  "torch": "2.14.1",
  "transformers": "5.22.0.dev0",
  "tokenizers": "0.23.3"
 },
 "sha256": {
  "config.json": "ad2f3bfbb3d898f702945bd9f78d314e20abf803618a852beddc6e39164a486c",
  "generation_config.json": "087c1528f0982539649c359e14be9f4fc9900f1a2ea39695d7baf632815056cb",
  "merges.txt": "ac9dfb2c5bd3b10bd46aa7fdaa8c54e8e1bcfe1618d9288320ebc82c518442fc",
  "model.safetensors": "9e02ed2bdffd25a57e35c3cebb45d5f51ddbecaf977919bfa46260d930e82247",
  "special_tokens_map.json": "1dd716f4648a359f7fca1221fc6a14edb1dd369239a4d7e0e5674c5f11f061d6",
  "tokenizer.json": "6190a6649d90a7fe35bfed8bff990c3e7aafc3cfa5beea185045f03f0890402b",
  "tokenizer_config.json": "8149d8e6b21275ad2cc346885ad92c0e9b5aa3e28a78bb5d39b7febc7e52545d",
  "vocab.json": "542ee3f8b27b3a825754728b4dc59eff31065fecce843a93f09b98f13dbd7de9"
 }
}