Upload folder using huggingface_hub

Intel__neural-chat-7b-v3-1/results_2024-06-27T21-55-55.507233.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6323441545508863,
+      "acc_stderr,none": 0.004811815959388812,
+      "acc_norm,none": 0.7975502887870942,
+      "acc_norm_stderr,none": 0.004010043978333027,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 62.2626522660805,
+      "eqbench_stderr,none": 2.2134366454600554,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=Intel/neural-chat-7b-v3-1,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.float16",
+    "model_revision": "main",
+    "model_sha": "c0d379a49c1c0579529d5e6f2e936ddb759552a8",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719523797.1185606,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "Intel/neural-chat-7b-v3-1",
+  "model_name_sanitized": "Intel__neural-chat-7b-v3-1",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 76180.270750278,
+  "end_time": 77745.57190531,
+  "total_evaluation_time_seconds": "1565.301155032008"
+}

KoboldAI__Mistral-7B-Erebus-v3/results_2024-06-28T02-01-18.290687.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.5837482573192591,
+      "acc_stderr,none": 0.0049192891130275095,
+      "acc_norm,none": 0.7665803624775941,
+      "acc_norm_stderr,none": 0.004221424792919153,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 18.19761609584577,
+      "eqbench_stderr,none": 3.474273216617232,
+      "percent_parseable,none": 97.6608187134503,
+      "percent_parseable_stderr,none": 1.1592247905734945,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=KoboldAI/Mistral-7B-Erebus-v3,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.float16",
+    "model_revision": "main",
+    "model_sha": "476c2eed031028f2fcfb9b8d0115b83363cec904",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719538494.5098195,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "KoboldAI/Mistral-7B-Erebus-v3",
+  "model_name_sanitized": "KoboldAI__Mistral-7B-Erebus-v3",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 90877.776869387,
+  "end_time": 92468.355496828,
+  "total_evaluation_time_seconds": "1590.578627440991"
+}

KoboldAI__Mistral-7B-Holodeck-1/results_2024-06-28T01-04-59.368025.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6026687910774746,
+      "acc_stderr,none": 0.004883455188908956,
+      "acc_norm,none": 0.7918741286596296,
+      "acc_norm_stderr,none": 0.0040513767194979506,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 2.099910527905425,
+      "eqbench_stderr,none": 2.491702523648299,
+      "percent_parseable,none": 98.24561403508773,
+      "percent_parseable_stderr,none": 1.0069193740062292,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=KoboldAI/Mistral-7B-Holodeck-1,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.float16",
+    "model_revision": "main",
+    "model_sha": "76057cc5c1923921162133c81ae7ca0e92755810",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719535119.7065547,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "KoboldAI/Mistral-7B-Holodeck-1",
+  "model_name_sanitized": "KoboldAI__Mistral-7B-Holodeck-1",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 87502.926965946,
+  "end_time": 89089.432844292,
+  "total_evaluation_time_seconds": "1586.5058783459972"
+}

NousResearch__Hermes-2-Pro-Mistral-7B/results_2024-06-28T00-36-44.931474.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6267675761800439,
+      "acc_stderr,none": 0.004826746160830164,
+      "acc_norm,none": 0.8055168293168692,
+      "acc_norm_stderr,none": 0.003949933997955457,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 65.92538496559615,
+      "eqbench_stderr,none": 2.278499695256187,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=NousResearch/Hermes-2-Pro-Mistral-7B,trust_remote_code=True",
+    "model_num_parameters": 7241994240,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "09317b1d8da639b5d9af77c06aa17cde0f0f91c0",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719533491.7175071,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_eos_token": [
+    "<|im_end|>",
+    32000
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 32000,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "NousResearch/Hermes-2-Pro-Mistral-7B",
+  "model_name_sanitized": "NousResearch__Hermes-2-Pro-Mistral-7B",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 85874.893829605,
+  "end_time": 87394.996271898,
+  "total_evaluation_time_seconds": "1520.1024422929913"
+}

Open-Orca__Mistral-7B-OpenOrca/results_2024-06-27T21-00-54.306241.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6379207329217288,
+      "acc_stderr,none": 0.004796193584930065,
+      "acc_norm,none": 0.8166699860585541,
+      "acc_norm_stderr,none": 0.0038614605262315377,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 63.978950638437865,
+      "eqbench_stderr,none": 2.3824356593314344,
+      "percent_parseable,none": 99.41520467836257,
+      "percent_parseable_stderr,none": 0.5847953216374284,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=Open-Orca/Mistral-7B-OpenOrca,trust_remote_code=True",
+    "model_num_parameters": 7241748480,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "4a37328cef00f524d3791b1c0cc559a3cc6af14d",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719520557.5287726,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "<|im_end|>",
+    32000
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 32000,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "Open-Orca/Mistral-7B-OpenOrca",
+  "model_name_sanitized": "Open-Orca__Mistral-7B-OpenOrca",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 72940.705278236,
+  "end_time": 74444.371073833,
+  "total_evaluation_time_seconds": "1503.6657955970004"
+}

SanjiWatsuki__Kunoichi-7B/results_2024-06-27T20-34-47.197919.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6803425612427804,
+      "acc_stderr,none": 0.004653907471785688,
+      "acc_norm,none": 0.8525194184425413,
+      "acc_norm_stderr,none": 0.003538596773704852,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 72.35673723130577,
+      "eqbench_stderr,none": 1.842888264461036,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=SanjiWatsuki/Kunoichi-7B,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "b2c23b9d0036e6e74e5f61de74776e9091956c83",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719519064.691441,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 8192,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "SanjiWatsuki/Kunoichi-7B",
+  "model_name_sanitized": "SanjiWatsuki__Kunoichi-7B",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 71447.838088771,
+  "end_time": 72877.26274353,
+  "total_evaluation_time_seconds": "1429.424654759001"
+}

Undi95__Toppy-M-7B/results_2024-06-28T02-28-16.478931.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6571400119498108,
+      "acc_stderr,none": 0.00473695081061781,
+      "acc_norm,none": 0.8351921927902808,
+      "acc_norm_stderr,none": 0.003702487662126953,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 66.56565114431275,
+      "eqbench_stderr,none": 2.1832557339862837,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=Undi95/Toppy-M-7B,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "5d0c492effbb2e52ea04e5100c6ce02eba48a793",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719540198.0639265,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "Undi95/Toppy-M-7B",
+  "model_name_sanitized": "Undi95__Toppy-M-7B",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 92581.210827571,
+  "end_time": 94086.543624108,
+  "total_evaluation_time_seconds": "1505.3327965369972"
+}

athirdpath__NSFW_DPO_vmgb-7b/results_2024-06-28T02-55-12.160237.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6730730930093607,
+      "acc_stderr,none": 0.0046813160644444095,
+      "acc_norm,none": 0.8536148177653854,
+      "acc_norm_stderr,none": 0.0035276951498235012,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 74.82935321697907,
+      "eqbench_stderr,none": 1.6591997145588517,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=athirdpath/NSFW_DPO_vmgb-7b,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "b667d7810267685cd4f32a8f82044e419c010abe",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719541817.2830012,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "athirdpath/NSFW_DPO_vmgb-7b",
+  "model_name_sanitized": "athirdpath__NSFW_DPO_vmgb-7b",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 94200.513127129,
+  "end_time": 95702.225070561,
+  "total_evaluation_time_seconds": "1501.7119434320048"
+}

jondurbin__airoboros-m-7b-3.1.2/results_2024-06-27T21-27-37.734965.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6262696673969329,
+      "acc_stderr,none": 0.0048280457747349,
+      "acc_norm,none": 0.8133837880900219,
+      "acc_norm_stderr,none": 0.0038880689432920544,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 38.51541437922841,
+      "eqbench_stderr,none": 3.511567923871361,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=jondurbin/airoboros-m-7b-3.1.2,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "e9a7f0271fa442d65bf6be87feeb3f4de2f5760e",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719522153.290284,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "jondurbin/airoboros-m-7b-3.1.2",
+  "model_name_sanitized": "jondurbin__airoboros-m-7b-3.1.2",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 74536.468290869,
+  "end_time": 76047.79978283,
+  "total_evaluation_time_seconds": "1511.3314919609984"
+}

jondurbin__cinematika-7b-v0.1/results_2024-06-27T23-16-51.732979.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6138219478191596,
+      "acc_stderr,none": 0.004858771963468838,
+      "acc_norm,none": 0.8031268671579367,
+      "acc_norm_stderr,none": 0.00396822985262125,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 44.84948931109151,
+      "eqbench_stderr,none": 3.1571076496385277,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=jondurbin/cinematika-7b-v0.1,trust_remote_code=True",
+    "model_num_parameters": 7241756672,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "6df1846af7de7ab8e2201ad87071ed661e3b0de2",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719528705.1154015,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "[PAD]",
+    32000
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "jondurbin/cinematika-7b-v0.1",
+  "model_name_sanitized": "jondurbin__cinematika-7b-v0.1",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 81088.287076101,
+  "end_time": 82601.797810444,
+  "total_evaluation_time_seconds": "1513.510734342999"
+}

migtissera__Synthia-7B-v3.0/results_2024-06-27T22-50-03.654626.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6188010356502689,
+      "acc_stderr,none": 0.004846886929763468,
+      "acc_norm,none": 0.8173670583549094,
+      "acc_norm_stderr,none": 0.0038557568514416335,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 15.026068685500109,
+      "eqbench_stderr,none": 2.8920476087351132,
+      "percent_parseable,none": 94.73684210526316,
+      "percent_parseable_stderr,none": 1.7126088775157098,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=migtissera/Synthia-7B-v3.0,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.float16",
+    "model_revision": "main",
+    "model_sha": "93c2e8b8055b42779f2b68059ebe38af6f2789c4",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719527019.219691,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "migtissera/Synthia-7B-v3.0",
+  "model_name_sanitized": "migtissera__Synthia-7B-v3.0",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 79402.540235241,
+  "end_time": 80993.719447117,
+  "total_evaluation_time_seconds": "1591.1792118759913"
+}

mlabonne__NeuralBeagle14-7B/results_2024-06-28T00-10-47.687175.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.7003584943238399,
+      "acc_stderr,none": 0.004571647137441099,
+      "acc_norm,none": 0.8645688109938259,
+      "acc_norm_stderr,none": 0.003414842236516961,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 74.20803234078544,
+      "eqbench_stderr,none": 1.9057062958788094,
+      "percent_parseable,none": 99.41520467836257,
+      "percent_parseable_stderr,none": 0.5847953216374271,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=mlabonne/NeuralBeagle14-7B,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.float16",
+    "model_revision": "main",
+    "model_sha": "1567ad618a0998139654cb355738bb9bc018ca64",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719531860.3285184,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "mlabonne/NeuralBeagle14-7B",
+  "model_name_sanitized": "mlabonne__NeuralBeagle14-7B",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 84243.603381348,
+  "end_time": 85837.751968409,
+  "total_evaluation_time_seconds": "1594.1485870609904"
+}

rwitz__go-bruins/results_2024-06-27T22-21-09.060416.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6664011153156741,
+      "acc_stderr,none": 0.0047053471376996584,
+      "acc_norm,none": 0.8492332204740092,
+      "acc_norm_stderr,none": 0.0035709011883580865,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 73.61594324522352,
+      "eqbench_stderr,none": 1.8000447804156592,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=rwitz/go-bruins,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "27b510cc158d83cad4f4df4f5cee65353647e080",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719525369.5356786,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "rwitz/go-bruins",
+  "model_name_sanitized": "rwitz__go-bruins",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 77752.802678043,
+  "end_time": 79259.12526582,
+  "total_evaluation_time_seconds": "1506.322587777002"
+}

senseable__WestLake-7B-v2/results_2024-06-28T01-32-26.319492.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.7048396733718383,
+      "acc_stderr,none": 0.0045518262729780865,
+      "acc_norm,none": 0.874228241386178,
+      "acc_norm_stderr,none": 0.0033091427273509244,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 77.87295612615341,
+      "eqbench_stderr,none": 1.577200540645531,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=senseable/WestLake-7B-v2,trust_remote_code=True",
+    "model_num_parameters": 7241732096,
+    "model_dtype": "torch.float16",
+    "model_revision": "main",
+    "model_sha": "41625004c47628837678859753b94c50c82f3bec",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719536781.997623,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "</s>",
+    2
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 2,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "senseable/WestLake-7B-v2",
+  "model_name_sanitized": "senseable__WestLake-7B-v2",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 89165.24699567,
+  "end_time": 90736.384333217,
+  "total_evaluation_time_seconds": "1571.1373375470139"
+}

teknium__OpenHermes-2.5-Mistral-7B/results_2024-06-27T23-43-07.467674.json

@@ -0,0 +1,177 @@
+{
+  "results": {
+    "hellaswag": {
+      "acc,none": 0.6302529376618203,
+      "acc_stderr,none": 0.004817495546789561,
+      "acc_norm,none": 0.8167695678151763,
+      "acc_norm_stderr,none": 0.003860646998897285,
+      "alias": "hellaswag"
+    },
+    "eq_bench": {
+      "eqbench,none": 65.75110483136034,
+      "eqbench_stderr,none": 2.270775919439369,
+      "percent_parseable,none": 100.0,
+      "percent_parseable_stderr,none": 0.0,
+      "alias": "eq_bench"
+    }
+  },
+  "group_subtasks": {
+    "eq_bench": [],
+    "hellaswag": []
+  },
+  "configs": {
+    "eq_bench": {
+      "task": "eq_bench",
+      "dataset_path": "pbevan11/EQ-Bench",
+      "validation_split": "validation",
+      "doc_to_text": "prompt",
+      "doc_to_target": "reference_answer_fullscale",
+      "process_results": "def calculate_score_fullscale(docs, results):\n    reference = eval(docs[\"reference_answer_fullscale\"])\n    user = dict(re.findall(r\"(\\w+):\\s+(\\d+)\", results[0]))\n    # First check that the emotions specified in the answer match those in the reference\n    if len(user.items()) != 4:\n        # print('! Error: 4 emotions were not returned')\n        # print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n    emotions_dict = {}\n    for emotion, user_emotion_score in user.items():\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                emotions_dict[emotion] = True\n    if len(emotions_dict) != 4:\n        print(\"! Error: emotions did not match reference\")\n        print(user)\n        return {\"eqbench\": 0, \"percent_parseable\": 0}\n\n    difference_tally = (\n        0  # Tally of differerence from reference answers for this question\n    )\n\n    # Iterate over each emotion in the user's answers.\n    for emotion, user_emotion_score in user.items():\n        # If this emotion is in the reference, calculate the difference between the user's score and the reference score.\n        for i in range(1, 5):\n            if emotion == reference[f\"emotion{i}\"]:\n                d = abs(\n                    float(user_emotion_score) - float(reference[f\"emotion{i}_score\"])\n                )\n                # this will be a value between 0 and 10\n                if d == 0:\n                    scaled_difference = 0\n                elif d <= 5:\n                    # S-shaped scaling function\n                    # https://www.desmos.com/calculator\n                    # 6.5\\cdot\\ \\frac{1}{\\left(1\\ +\\ e^{\\left(-1.2\\cdot\\left(x-4\\right)\\right)}\\right)}\n                    scaled_difference = 6.5 * (1 / (1 + math.e ** (-1.2 * (d - 4))))\n\n                else:\n                    scaled_difference = d\n                difference_tally += scaled_difference\n\n    # Inverting the difference tally so that the closer the answer is to reference, the higher the score.\n    # The adjustment constant is chosen such that answering randomly produces a score of zero.\n    adjust_const = 0.7477\n    final_score = 10 - (difference_tally * adjust_const)\n    final_score_percent = final_score * 10\n\n    return {\"eqbench\": final_score_percent, \"percent_parseable\": 100}\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "eqbench",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "percent_parseable",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "do_sample": false,
+        "temperature": 0.0,
+        "max_gen_toks": 80,
+        "until": [
+          "\n\n"
+        ]
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 2.1
+      }
+    },
+    "hellaswag": {
+      "task": "hellaswag",
+      "group": [
+        "multiple_choice"
+      ],
+      "dataset_path": "hellaswag",
+      "training_split": "train",
+      "validation_split": "validation",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc):\n        ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n        out_doc = {\n            \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n            \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n            \"gold\": int(doc[\"label\"]),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "{{query}}",
+      "doc_to_target": "{{label}}",
+      "doc_to_choice": "choices",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        },
+        {
+          "metric": "acc_norm",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "eq_bench": 2.1,
+    "hellaswag": 1.0
+  },
+  "n-shot": {
+    "eq_bench": 0,
+    "hellaswag": 0
+  },
+  "higher_is_better": {
+    "eq_bench": {
+      "eqbench": true,
+      "percent_parseable": true
+    },
+    "hellaswag": {
+      "acc": true,
+      "acc_norm": true
+    }
+  },
+  "n-samples": {
+    "hellaswag": {
+      "original": 10042,
+      "effective": 10042
+    },
+    "eq_bench": {
+      "original": 171,
+      "effective": 171
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=teknium/OpenHermes-2.5-Mistral-7B,trust_remote_code=True",
+    "model_num_parameters": 7241748480,
+    "model_dtype": "torch.bfloat16",
+    "model_revision": "main",
+    "model_sha": "24c0bea14d53e6f67f1fbe2eca5bfe7cae389b33",
+    "batch_size": "auto",
+    "batch_sizes": [
+      64
+    ],
+    "device": null,
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": null,
+  "date": 1719530289.0024347,
+  "pretty_env_info": "PyTorch version: 2.3.1+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.4 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-1022-gcp-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 12.1.105\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA L4\nGPU 1: NVIDIA L4\n\nNvidia driver version: 555.42.02\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:                       x86_64\nCPU op-mode(s):                     32-bit, 64-bit\nAddress sizes:                      46 bits physical, 48 bits virtual\nByte Order:                         Little Endian\nCPU(s):                             24\nOn-line CPU(s) list:                0-23\nVendor ID:                          GenuineIntel\nModel name:                         Intel(R) Xeon(R) CPU @ 2.20GHz\nCPU family:                         6\nModel:                              85\nThread(s) per core:                 2\nCore(s) per socket:                 12\nSocket(s):                          1\nStepping:                           7\nBogoMIPS:                           4400.47\nFlags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves arat avx512_vnni md_clear arch_capabilities\nHypervisor vendor:                  KVM\nVirtualization type:                full\nL1d cache:                          384 KiB (12 instances)\nL1i cache:                          384 KiB (12 instances)\nL2 cache:                           12 MiB (12 instances)\nL3 cache:                           38.5 MiB (1 instance)\nNUMA node(s):                       1\nNUMA node0 CPU(s):                  0-23\nVulnerability Gather data sampling: Not affected\nVulnerability Itlb multihit:        Not affected\nVulnerability L1tf:                 Not affected\nVulnerability Mds:                  Mitigation; Clear CPU buffers; SMT Host state unknown\nVulnerability Meltdown:             Not affected\nVulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown\nVulnerability Retbleed:             Mitigation; Enhanced IBRS\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2:           Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds:                Not affected\nVulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT Host state unknown\n\nVersions of relevant libraries:\n[pip3] numpy==2.0.0\n[pip3] torch==2.3.1\n[pip3] triton==2.3.1\n[conda] Could not collect",
+  "transformers_version": "4.41.2",
+  "upper_git_hash": null,
+  "tokenizer_pad_token": [
+    "<unk>",
+    0
+  ],
+  "tokenizer_eos_token": [
+    "<|im_end|>",
+    32000
+  ],
+  "tokenizer_bos_token": [
+    "<s>",
+    1
+  ],
+  "eot_token_id": 32000,
+  "max_length": 32768,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "teknium/OpenHermes-2.5-Mistral-7B",
+  "model_name_sanitized": "teknium__OpenHermes-2.5-Mistral-7B",
+  "system_instruction": null,
+  "system_instruction_sha": null,
+  "fewshot_as_multiturn": false,
+  "chat_template": null,
+  "chat_template_sha": null,
+  "start_time": 82672.316515428,
+  "end_time": 84177.532482507,
+  "total_evaluation_time_seconds": "1505.2159670789988"
+}