{ "_name_or_path": "distributed/llama-1b", "all_reduce_scores": { "0": "NOT_ALIVE", "1": "NON_PARTICIPATING", "10": "NOT_ALIVE", "100": "NOT_ALIVE", "101": "NOT_ALIVE", "102": "NOT_ALIVE", "103": "FAIL", "104": "NOT_ALIVE", "105": "NOT_ALIVE", "106": "NON_PARTICIPATING", "107": "NON_PARTICIPATING", "108": "NOT_ALIVE", "109": "NON_PARTICIPATING", "11": "FAIL", "110": "NOT_ALIVE", "111": "NOT_ALIVE", "112": "FAIL", "113": "NOT_ALIVE", "114": "SUCCESS", "115": "NON_PARTICIPATING", "116": "SUCCESS", "117": "NON_PARTICIPATING", "118": "NOT_ALIVE", "119": "NON_PARTICIPATING", "12": "NOT_ALIVE", "120": "NOT_ALIVE", "121": "NOT_ALIVE", "122": "NON_PARTICIPATING", "123": "FAIL", "124": "NOT_ALIVE", "125": "NOT_ALIVE", "126": "NOT_ALIVE", "127": "NON_PARTICIPATING", "128": "NON_PARTICIPATING", "129": "NON_PARTICIPATING", "13": "NON_PARTICIPATING", "130": "NOT_ALIVE", "131": "NOT_ALIVE", "132": "NOT_ALIVE", "133": "NON_PARTICIPATING", "134": "SUCCESS", "135": "NON_PARTICIPATING", "136": "NOT_ALIVE", "137": "NOT_ALIVE", "138": "NON_PARTICIPATING", "139": "NOT_ALIVE", "14": "NON_PARTICIPATING", "140": "NOT_ALIVE", "141": "NOT_ALIVE", "142": "NOT_ALIVE", "143": "FAIL", "144": "NOT_ALIVE", "145": "NOT_ALIVE", "146": "NON_PARTICIPATING", "147": "NON_PARTICIPATING", "148": "NON_PARTICIPATING", "149": "NOT_ALIVE", "15": "NON_PARTICIPATING", "150": "NON_PARTICIPATING", "151": "NOT_ALIVE", "152": "NON_PARTICIPATING", "153": "FAIL", "154": "SUCCESS", "155": "FAIL", "156": "NON_PARTICIPATING", "157": "NON_PARTICIPATING", "158": "NON_PARTICIPATING", "159": "NOT_ALIVE", "16": "NOT_ALIVE", "160": "FAIL", "161": "NOT_ALIVE", "162": "NOT_ALIVE", "163": "SUCCESS", "164": "NOT_ALIVE", "165": "NOT_ALIVE", "166": "NOT_ALIVE", "167": "FAIL", "168": "FAIL", "169": "NOT_ALIVE", "17": "NON_PARTICIPATING", "170": "NON_PARTICIPATING", "171": "NOT_ALIVE", "172": "NOT_ALIVE", "173": "NON_PARTICIPATING", "174": "NON_PARTICIPATING", "175": "FAIL", "176": "NOT_ALIVE", "177": "NOT_ALIVE", "178": "NOT_ALIVE", "179": "NON_PARTICIPATING", "18": "FAIL", "180": "NON_PARTICIPATING", "181": "NOT_ALIVE", "182": "NON_PARTICIPATING", "183": "NON_PARTICIPATING", "184": "NON_PARTICIPATING", "185": "NOT_ALIVE", "186": "FAIL", "187": "NOT_ALIVE", "188": "FAIL", "189": "NOT_ALIVE", "19": "NON_PARTICIPATING", "190": "NON_PARTICIPATING", "191": "NOT_ALIVE", "192": "NON_PARTICIPATING", "193": "NON_PARTICIPATING", "194": "NOT_ALIVE", "195": "FAIL", "196": "NON_PARTICIPATING", "197": "NON_PARTICIPATING", "198": "NON_PARTICIPATING", "199": "NON_PARTICIPATING", "2": "NON_PARTICIPATING", "20": "NON_PARTICIPATING", "200": "NON_PARTICIPATING", "201": "FAIL", "202": "NOT_ALIVE", "203": "NOT_ALIVE", "204": "NOT_ALIVE", "205": "NON_PARTICIPATING", "206": "FAIL", "207": "NOT_ALIVE", "208": "NOT_ALIVE", "209": "FAIL", "21": "NON_PARTICIPATING", "210": "NON_PARTICIPATING", "211": "FAIL", "212": "NON_PARTICIPATING", "213": "SUCCESS", "214": "NON_PARTICIPATING", "215": "NOT_ALIVE", "216": "NOT_ALIVE", "217": "NOT_ALIVE", "218": "FAIL", "219": "NOT_ALIVE", "22": "FAIL", "220": "NOT_ALIVE", "221": "NON_PARTICIPATING", "222": "NON_PARTICIPATING", "223": "FAIL", "224": "NOT_ALIVE", "225": "NON_PARTICIPATING", "226": "NON_PARTICIPATING", "227": "FAIL", "228": "NOT_ALIVE", "229": "NON_PARTICIPATING", "23": "NOT_ALIVE", "230": "NOT_ALIVE", "231": "NOT_ALIVE", "232": "NON_PARTICIPATING", "233": "FAIL", "234": "NOT_ALIVE", "235": "SUCCESS", "236": "FAIL", "237": "FAIL", "238": "NON_PARTICIPATING", "239": "NON_PARTICIPATING", "24": "NON_PARTICIPATING", "240": "NOT_ALIVE", "241": "SUCCESS", "242": "FAIL", "243": "NOT_ALIVE", "244": "NOT_ALIVE", "245": "NON_PARTICIPATING", "246": "FAIL", "247": "NOT_ALIVE", "248": "FAIL", "249": "FAIL", "25": "SUCCESS", "250": "NOT_ALIVE", "251": "NOT_ALIVE", "252": "FAIL", "253": "NOT_ALIVE", "254": "NOT_ALIVE", "255": "NOT_ALIVE", "26": "NOT_ALIVE", "27": "NON_PARTICIPATING", "28": "FAIL", "29": "FAIL", "3": "FAIL", "30": "NON_PARTICIPATING", "31": "NOT_ALIVE", "32": "FAIL", "33": "NOT_ALIVE", "34": "NON_PARTICIPATING", "35": "FAIL", "36": "NON_PARTICIPATING", "37": "NOT_ALIVE", "38": "FAIL", "39": "NON_PARTICIPATING", "4": "NON_PARTICIPATING", "40": "FAIL", "41": "NOT_ALIVE", "42": "NOT_ALIVE", "43": "NOT_ALIVE", "44": "NON_PARTICIPATING", "45": "FAIL", "46": "NOT_ALIVE", "47": "FAIL", "48": "NON_PARTICIPATING", "49": "NOT_ALIVE", "5": "NON_PARTICIPATING", "50": "NON_PARTICIPATING", "51": "NON_PARTICIPATING", "52": "NON_PARTICIPATING", "53": "NON_PARTICIPATING", "54": "NON_PARTICIPATING", "55": "NON_PARTICIPATING", "56": "FAIL", "57": "NON_PARTICIPATING", "58": "NON_PARTICIPATING", "59": "NON_PARTICIPATING", "6": "NOT_ALIVE", "60": "NON_PARTICIPATING", "61": "NOT_ALIVE", "62": "SUCCESS", "63": "NON_PARTICIPATING", "64": "NOT_ALIVE", "65": "NOT_ALIVE", "66": "NOT_ALIVE", "67": "NON_PARTICIPATING", "68": "NOT_ALIVE", "69": "FAIL", "7": "SUCCESS", "70": "NOT_ALIVE", "71": "NOT_ALIVE", "72": "NON_PARTICIPATING", "73": "NOT_ALIVE", "74": "NOT_ALIVE", "75": "NON_PARTICIPATING", "76": "NON_PARTICIPATING", "77": "NON_PARTICIPATING", "78": "NON_PARTICIPATING", "79": "NON_PARTICIPATING", "8": "NON_PARTICIPATING", "80": "NOT_ALIVE", "81": "NOT_ALIVE", "82": "NOT_ALIVE", "83": "NON_PARTICIPATING", "84": "NOT_ALIVE", "85": "NOT_ALIVE", "86": "FAIL", "87": "SUCCESS", "88": "FAIL", "89": "NOT_ALIVE", "9": "NOT_ALIVE", "90": "FAIL", "91": "SUCCESS", "92": "FAIL", "93": "NOT_ALIVE", "94": "NON_PARTICIPATING", "95": "NOT_ALIVE", "96": "NOT_ALIVE", "97": "NON_PARTICIPATING", "98": "NON_PARTICIPATING", "99": "NON_PARTICIPATING" }, "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "block_list": [ 6478739, 6478750 ], "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "inner_step": 49, "intermediate_size": 5632, "last_allreduce_block": 6478480, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.39.3", "use_cache": false, "vocab_size": 32000 }