f48e98f73b404d67c66a518067c5bf2063e5db757b239d40cf96c381dbc580f4

config.json

@@ -1,39 +1,39 @@
 {
-  "_name_or_path": "/tmp/models/tmpt4h4qgbw/tmpy2_5t8rr",
+  "_name_or_path": "/tmp/models/tmpw4q4h9hb/tmp6lhbyvd4",
   "activation_function": "gelu_new",
   "all_reduce_scores": {
     "0": "NON_PARTICIPATING",
-    "1": "SUCCESS",
+    "1": "NON_PARTICIPATING",
     "10": "SUCCESS",
     "100": "NON_PARTICIPATING",
     "101": "NON_PARTICIPATING",
     "102": "SUCCESS",
-    "103": "SUCCESS",
-    "104": "NON_PARTICIPATING",
+    "103": "NON_PARTICIPATING",
+    "104": "SUCCESS",
     "105": "NON_PARTICIPATING",
-    "106": "SUCCESS",
+    "106": "NON_PARTICIPATING",
     "107": "NON_PARTICIPATING",
-    "108": "NON_PARTICIPATING",
-    "109": "NON_PARTICIPATING",
-    "11": "NON_PARTICIPATING",
-    "110": "NON_PARTICIPATING",
+    "108": "SUCCESS",
+    "109": "SUCCESS",
+    "11": "SUCCESS",
+    "110": "SUCCESS",
     "111": "SUCCESS",
-    "112": "SUCCESS",
+    "112": "NON_PARTICIPATING",
     "113": "NON_PARTICIPATING",
     "114": "NON_PARTICIPATING",
-    "115": "NON_PARTICIPATING",
-    "116": "NON_PARTICIPATING",
+    "115": "SUCCESS",
+    "116": "SUCCESS",
     "117": "NON_PARTICIPATING",
     "118": "NON_PARTICIPATING",
     "119": "SUCCESS",
-    "12": "SUCCESS",
-    "120": "SUCCESS",
-    "121": "NON_PARTICIPATING",
+    "12": "NON_PARTICIPATING",
+    "120": "NON_PARTICIPATING",
+    "121": "SUCCESS",
     "122": "NON_PARTICIPATING",
     "123": "SUCCESS",
     "124": "NON_PARTICIPATING",
     "125": "NON_PARTICIPATING",
-    "126": "NON_PARTICIPATING",
+    "126": "SUCCESS",
     "127": "NON_PARTICIPATING",
     "128": "NON_PARTICIPATING",
     "129": "SUCCESS",
@@ -41,223 +41,223 @@
     "130": "SUCCESS",
     "131": "NON_PARTICIPATING",
     "132": "NON_PARTICIPATING",
-    "133": "SUCCESS",
-    "134": "SUCCESS",
-    "135": "NON_PARTICIPATING",
+    "133": "NON_PARTICIPATING",
+    "134": "NON_PARTICIPATING",
+    "135": "SUCCESS",
     "136": "NON_PARTICIPATING",
-    "137": "SUCCESS",
+    "137": "NON_PARTICIPATING",
     "138": "SUCCESS",
     "139": "NON_PARTICIPATING",
-    "14": "NON_PARTICIPATING",
+    "14": "SUCCESS",
     "140": "NON_PARTICIPATING",
     "141": "NON_PARTICIPATING",
-    "142": "NON_PARTICIPATING",
+    "142": "SUCCESS",
     "143": "NON_PARTICIPATING",
     "144": "NON_PARTICIPATING",
-    "145": "NON_PARTICIPATING",
-    "146": "NON_PARTICIPATING",
+    "145": "SUCCESS",
+    "146": "SUCCESS",
     "147": "NON_PARTICIPATING",
-    "148": "SUCCESS",
-    "149": "SUCCESS",
+    "148": "NON_PARTICIPATING",
+    "149": "NON_PARTICIPATING",
     "15": "SUCCESS",
-    "150": "SUCCESS",
-    "151": "NON_PARTICIPATING",
+    "150": "NON_PARTICIPATING",
+    "151": "SUCCESS",
     "152": "SUCCESS",
     "153": "NON_PARTICIPATING",
-    "154": "NON_PARTICIPATING",
-    "155": "NON_PARTICIPATING",
+    "154": "SUCCESS",
+    "155": "SUCCESS",
     "156": "SUCCESS",
-    "157": "NON_PARTICIPATING",
-    "158": "SUCCESS",
+    "157": "SUCCESS",
+    "158": "NON_PARTICIPATING",
     "159": "NON_PARTICIPATING",
-    "16": "SUCCESS",
+    "16": "NON_PARTICIPATING",
     "160": "NON_PARTICIPATING",
-    "161": "SUCCESS",
+    "161": "NON_PARTICIPATING",
     "162": "NON_PARTICIPATING",
-    "163": "NON_PARTICIPATING",
+    "163": "SUCCESS",
     "164": "NON_PARTICIPATING",
     "165": "NON_PARTICIPATING",
     "166": "SUCCESS",
-    "167": "NON_PARTICIPATING",
+    "167": "SUCCESS",
     "168": "NON_PARTICIPATING",
     "169": "NON_PARTICIPATING",
     "17": "NON_PARTICIPATING",
     "170": "SUCCESS",
     "171": "NON_PARTICIPATING",
     "172": "SUCCESS",
-    "173": "SUCCESS",
-    "174": "SUCCESS",
+    "173": "NON_PARTICIPATING",
+    "174": "NON_PARTICIPATING",
     "175": "NON_PARTICIPATING",
-    "176": "NON_PARTICIPATING",
-    "177": "NON_PARTICIPATING",
+    "176": "SUCCESS",
+    "177": "SUCCESS",
     "178": "NON_PARTICIPATING",
-    "179": "NON_PARTICIPATING",
+    "179": "SUCCESS",
     "18": "SUCCESS",
-    "180": "SUCCESS",
+    "180": "NON_PARTICIPATING",
     "181": "NON_PARTICIPATING",
-    "182": "SUCCESS",
+    "182": "NON_PARTICIPATING",
     "183": "SUCCESS",
-    "184": "NON_PARTICIPATING",
+    "184": "SUCCESS",
     "185": "NON_PARTICIPATING",
     "186": "SUCCESS",
     "187": "NON_PARTICIPATING",
     "188": "NON_PARTICIPATING",
-    "189": "NON_PARTICIPATING",
+    "189": "SUCCESS",
     "19": "SUCCESS",
-    "190": "NON_PARTICIPATING",
-    "191": "SUCCESS",
+    "190": "SUCCESS",
+    "191": "NON_PARTICIPATING",
     "192": "SUCCESS",
     "193": "NON_PARTICIPATING",
     "194": "NON_PARTICIPATING",
-    "195": "NON_PARTICIPATING",
+    "195": "SUCCESS",
     "196": "NON_PARTICIPATING",
     "197": "NON_PARTICIPATING",
-    "198": "NON_PARTICIPATING",
+    "198": "SUCCESS",
     "199": "NON_PARTICIPATING",
     "2": "NON_PARTICIPATING",
     "20": "NON_PARTICIPATING",
-    "200": "NON_PARTICIPATING",
+    "200": "SUCCESS",
     "201": "NON_PARTICIPATING",
-    "202": "SUCCESS",
-    "203": "SUCCESS",
+    "202": "NON_PARTICIPATING",
+    "203": "NON_PARTICIPATING",
     "204": "NON_PARTICIPATING",
-    "205": "NON_PARTICIPATING",
+    "205": "SUCCESS",
     "206": "NON_PARTICIPATING",
     "207": "NON_PARTICIPATING",
     "208": "NON_PARTICIPATING",
     "209": "SUCCESS",
     "21": "SUCCESS",
     "210": "NON_PARTICIPATING",
-    "211": "NON_PARTICIPATING",
-    "212": "NON_PARTICIPATING",
+    "211": "SUCCESS",
+    "212": "SUCCESS",
     "213": "NON_PARTICIPATING",
     "214": "NON_PARTICIPATING",
     "215": "SUCCESS",
-    "216": "SUCCESS",
+    "216": "NON_PARTICIPATING",
     "217": "NON_PARTICIPATING",
-    "218": "SUCCESS",
-    "219": "NON_PARTICIPATING",
+    "218": "NON_PARTICIPATING",
+    "219": "SUCCESS",
     "22": "NON_PARTICIPATING",
-    "220": "SUCCESS",
+    "220": "NON_PARTICIPATING",
     "221": "NON_PARTICIPATING",
     "222": "NON_PARTICIPATING",
     "223": "NON_PARTICIPATING",
     "224": "NON_PARTICIPATING",
-    "225": "SUCCESS",
+    "225": "NON_PARTICIPATING",
     "226": "NON_PARTICIPATING",
     "227": "SUCCESS",
-    "228": "SUCCESS",
+    "228": "NON_PARTICIPATING",
     "229": "NON_PARTICIPATING",
     "23": "NON_PARTICIPATING",
     "230": "NON_PARTICIPATING",
-    "231": "NON_PARTICIPATING",
-    "232": "NON_PARTICIPATING",
-    "233": "SUCCESS",
+    "231": "SUCCESS",
+    "232": "SUCCESS",
+    "233": "NON_PARTICIPATING",
     "234": "NON_PARTICIPATING",
-    "235": "NON_PARTICIPATING",
+    "235": "SUCCESS",
     "236": "NON_PARTICIPATING",
     "237": "SUCCESS",
     "238": "NON_PARTICIPATING",
     "239": "NON_PARTICIPATING",
-    "24": "NON_PARTICIPATING",
+    "24": "SUCCESS",
     "240": "NON_PARTICIPATING",
     "241": "SUCCESS",
     "242": "SUCCESS",
-    "243": "SUCCESS",
+    "243": "NON_PARTICIPATING",
     "244": "NON_PARTICIPATING",
-    "245": "NON_PARTICIPATING",
-    "246": "NON_PARTICIPATING",
+    "245": "SUCCESS",
+    "246": "SUCCESS",
     "247": "NON_PARTICIPATING",
     "248": "NON_PARTICIPATING",
-    "249": "SUCCESS",
+    "249": "NON_PARTICIPATING",
     "25": "NON_PARTICIPATING",
     "250": "NON_PARTICIPATING",
     "251": "NON_PARTICIPATING",
     "252": "NON_PARTICIPATING",
-    "253": "SUCCESS",
-    "254": "NON_PARTICIPATING",
+    "253": "NON_PARTICIPATING",
+    "254": "SUCCESS",
     "255": "SUCCESS",
     "26": "SUCCESS",
-    "27": "NON_PARTICIPATING",
-    "28": "SUCCESS",
+    "27": "SUCCESS",
+    "28": "NON_PARTICIPATING",
     "29": "NON_PARTICIPATING",
     "3": "SUCCESS",
-    "30": "SUCCESS",
-    "31": "NON_PARTICIPATING",
-    "32": "NON_PARTICIPATING",
+    "30": "NON_PARTICIPATING",
+    "31": "SUCCESS",
+    "32": "SUCCESS",
     "33": "NON_PARTICIPATING",
-    "34": "NON_PARTICIPATING",
+    "34": "SUCCESS",
     "35": "NON_PARTICIPATING",
-    "36": "SUCCESS",
+    "36": "NON_PARTICIPATING",
     "37": "NON_PARTICIPATING",
     "38": "NON_PARTICIPATING",
-    "39": "NON_PARTICIPATING",
-    "4": "NON_PARTICIPATING",
-    "40": "SUCCESS",
-    "41": "SUCCESS",
-    "42": "SUCCESS",
+    "39": "SUCCESS",
+    "4": "SUCCESS",
+    "40": "NON_PARTICIPATING",
+    "41": "NON_PARTICIPATING",
+    "42": "NON_PARTICIPATING",
     "43": "NON_PARTICIPATING",
-    "44": "SUCCESS",
+    "44": "NON_PARTICIPATING",
     "45": "SUCCESS",
     "46": "NON_PARTICIPATING",
     "47": "NON_PARTICIPATING",
     "48": "NON_PARTICIPATING",
-    "49": "NON_PARTICIPATING",
+    "49": "SUCCESS",
     "5": "NON_PARTICIPATING",
-    "50": "NON_PARTICIPATING",
-    "51": "SUCCESS",
-    "52": "SUCCESS",
-    "53": "SUCCESS",
-    "54": "SUCCESS",
-    "55": "SUCCESS",
+    "50": "SUCCESS",
+    "51": "NON_PARTICIPATING",
+    "52": "NON_PARTICIPATING",
+    "53": "NON_PARTICIPATING",
+    "54": "NON_PARTICIPATING",
+    "55": "NON_PARTICIPATING",
     "56": "NON_PARTICIPATING",
     "57": "NON_PARTICIPATING",
-    "58": "SUCCESS",
-    "59": "NON_PARTICIPATING",
-    "6": "NON_PARTICIPATING",
+    "58": "NON_PARTICIPATING",
+    "59": "SUCCESS",
+    "6": "SUCCESS",
     "60": "SUCCESS",
     "61": "NON_PARTICIPATING",
-    "62": "SUCCESS",
-    "63": "NON_PARTICIPATING",
-    "64": "SUCCESS",
-    "65": "SUCCESS",
+    "62": "NON_PARTICIPATING",
+    "63": "SUCCESS",
+    "64": "NON_PARTICIPATING",
+    "65": "NON_PARTICIPATING",
     "66": "NON_PARTICIPATING",
     "67": "NON_PARTICIPATING",
     "68": "SUCCESS",
-    "69": "SUCCESS",
+    "69": "NON_PARTICIPATING",
     "7": "NON_PARTICIPATING",
     "70": "NON_PARTICIPATING",
     "71": "NON_PARTICIPATING",
-    "72": "NON_PARTICIPATING",
-    "73": "SUCCESS",
+    "72": "SUCCESS",
+    "73": "NON_PARTICIPATING",
     "74": "NON_PARTICIPATING",
-    "75": "NON_PARTICIPATING",
+    "75": "SUCCESS",
     "76": "NON_PARTICIPATING",
-    "77": "NON_PARTICIPATING",
+    "77": "SUCCESS",
     "78": "SUCCESS",
     "79": "SUCCESS",
     "8": "SUCCESS",
-    "80": "NON_PARTICIPATING",
+    "80": "SUCCESS",
     "81": "SUCCESS",
     "82": "NON_PARTICIPATING",
-    "83": "SUCCESS",
-    "84": "SUCCESS",
+    "83": "NON_PARTICIPATING",
+    "84": "NON_PARTICIPATING",
     "85": "SUCCESS",
     "86": "NON_PARTICIPATING",
-    "87": "SUCCESS",
+    "87": "NON_PARTICIPATING",
     "88": "NON_PARTICIPATING",
     "89": "NON_PARTICIPATING",
     "9": "NON_PARTICIPATING",
     "90": "NON_PARTICIPATING",
-    "91": "SUCCESS",
-    "92": "NON_PARTICIPATING",
+    "91": "NON_PARTICIPATING",
+    "92": "SUCCESS",
     "93": "NON_PARTICIPATING",
-    "94": "NON_PARTICIPATING",
-    "95": "SUCCESS",
+    "94": "SUCCESS",
+    "95": "NON_PARTICIPATING",
     "96": "NON_PARTICIPATING",
-    "97": "NON_PARTICIPATING",
-    "98": "NON_PARTICIPATING",
-    "99": "NON_PARTICIPATING"
+    "97": "SUCCESS",
+    "98": "SUCCESS",
+    "99": "SUCCESS"
   },
   "architectures": [
     "GPTOptim"

smash_config.json

@@ -26,7 +26,7 @@
     "quant_hqq_weight_bits": 8,
     "max_batch_size": 1,
     "device": "cuda",
-    "cache_dir": "/tmp/models/tmpt4h4qgbw",
+    "cache_dir": "/tmp/models/tmpw4q4h9hb",
     "task": "",
     "save_load_fn": "hqq",
     "save_load_fn_args": {},