Training in progress, step 12

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a4b766f685049ba65fd5771463826b62c7df1d1a4a20f21847a78bc102475f5
+oid sha256:05d589b2a1c6787ffb348050f6fdc49c0ecc2fe8ea1488b498164e8671751a97
 size 137970488

run-0/checkpoint-12/config.json

@@ -0,0 +1,26 @@
+{
+  "_name_or_path": "mistralai/Mistral-7B-v0.1",
+  "architectures": [
+    "MistralForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 0,
+  "eos_token_id": 0,
+  "hidden_act": "silu",
+  "hidden_size": 432,
+  "initializer_range": 0.02,
+  "intermediate_size": 1600,
+  "max_position_embeddings": 75,
+  "model_type": "mistral",
+  "num_attention_heads": 8,
+  "num_hidden_layers": 12,
+  "num_key_value_heads": 8,
+  "rms_norm_eps": 1e-05,
+  "rope_theta": 10000.0,
+  "sliding_window": 75,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.38.2",
+  "use_cache": true,
+  "vocab_size": 738
+}

run-0/checkpoint-12/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "eos_token_id": 0,
+  "transformers_version": "4.38.2"
+}

run-0/checkpoint-12/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05d589b2a1c6787ffb348050f6fdc49c0ecc2fe8ea1488b498164e8671751a97
+size 137970488

run-0/checkpoint-12/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ee76df05622695c8a2fd73d806908e2fd642b9517e65ee721a9eef1f34c7aae
+size 276010426

run-0/checkpoint-12/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70c9f6de7e0c53595531a7bd28932b385562e6645002bd519aa4d5ecd618f456
+size 14244

run-0/checkpoint-12/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a48a9bf789897d06ca14032ecff94dd18bf2e25751bae26b0f33281cf4f92d42
+size 1064

run-0/checkpoint-12/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<|endoftext|>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}

run-0/checkpoint-12/tokenizer_config.json

@@ -0,0 +1,20 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "model_max_length": 75,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "AISTokenizer",
+  "unk_token": "<|endoftext|>"
+}

run-0/checkpoint-12/trainer_state.json

@@ -0,0 +1,38 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.6,
+  "eval_steps": 10,
+  "global_step": 12,
+  "is_hyper_param_search": true,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.5,
+      "grad_norm": 200694.375,
+      "learning_rate": 0.00011654557407940855,
+      "loss": 5.3823,
+      "step": 10
+    },
+    {
+      "epoch": 0.5,
+      "eval_loss": 4.528500556945801,
+      "eval_runtime": 17.0848,
+      "eval_samples_per_second": 585.315,
+      "eval_steps_per_second": 1.112,
+      "step": 10
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 20,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "save_steps": 6,
+  "total_flos": 83650040064000.0,
+  "train_batch_size": 544,
+  "trial_name": null,
+  "trial_params": {
+    "learning_rate": 0.00011654557407940855
+  }
+}

run-0/checkpoint-12/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c0adb69a864270aa8734e61821561a06ca3d35d3cd81446c72576c4c967a811
+size 4920

run-0/checkpoint-12/vocab_file.txt

@@ -0,0 +1,738 @@
+<|endoftext|>
+#
+(
+)
+-
+/
+1
+2
+3
+4
+5
+6
+7
+=
+[Br;!R;C]
+[Br;!R;N]
+[C;!R;BrBrBrC]
+[C;!R;BrBrCF]
+[C;!R;BrBrCS]
+[C;!R;BrCCC]
+[C;!R;BrCC]
+[C;!R;BrCFF]
+[C;!R;BrCN]
+[C;!R;CCCC]
+[C;!R;CCCCl]
+[C;!R;CCCF]
+[C;!R;CCCN]
+[C;!R;CCCO]
+[C;!R;CCCP]
+[C;!R;CCCS]
+[C;!R;CCC]
+[C;!R;CCClCl]
+[C;!R;CCClO]
+[C;!R;CCCl]
+[C;!R;CCFF]
+[C;!R;CCF]
+[C;!R;CCI]
+[C;!R;CCNN]
+[C;!R;CCNO]
+[C;!R;CCN]
+[C;!R;CCOO]
+[C;!R;CCO]
+[C;!R;CCP]
+[C;!R;CCSS]
+[C;!R;CCS]
+[C;!R;CC]
+[C;!R;CClClCl]
+[C;!R;CClClF]
+[C;!R;CClClO]
+[C;!R;CClClS]
+[C;!R;CClCl]
+[C;!R;CClFF]
+[C;!R;CClN]
+[C;!R;CFFF]
+[C;!R;CFFI]
+[C;!R;CFFN]
+[C;!R;CFFO]
+[C;!R;CFFP]
+[C;!R;CFFS]
+[C;!R;CFF]
+[C;!R;CFS]
+[C;!R;CNN]
+[C;!R;CNO]
+[C;!R;CNP]
+[C;!R;CNS]
+[C;!R;CN]
+[C;!R;COO]
+[C;!R;COS]
+[C;!R;COSi]
+[C;!R;CO]
+[C;!R;CSSS]
+[C;!R;CSS]
+[C;!R;ClClClO]
+[C;!R;ClClClS]
+[C;!R;ClClFS]
+[C;!R;ClClNO]
+[C;!R;ClFFO]
+[C;!R;FFFN]
+[C;!R;FFFO]
+[C;!R;FFFS]
+[C;!R;FFNO]
+[C;!R;FFNS]
+[C;!R;NNN]
+[C;!R;NNO]
+[C;!R;NNS]
+[C;!R;NN]
+[C;!R;NOO]
+[C;!R;NOP]
+[C;!R;NOS]
+[C;!R;NSS]
+[C;!R;NS]
+[C;!R;OOO]
+[C;!R;OOS]
+[C;!R;OSS]
+[C;R;BrBrCC]
+[C;R;BrBrCS]
+[C;R;BrCCC]
+[C;R;BrCC]
+[C;R;BrCN]
+[C;R;BrCS]
+[C;R;CCCC]
+[C;R;CCCCl]
+[C;R;CCCF]
+[C;R;CCCN]
+[C;R;CCCO]
+[C;R;CCCS]
+[C;R;CCCSi]
+[C;R;CCC]
+[C;R;CCClCl]
+[C;R;CCClO]
+[C;R;CCCl]
+[C;R;CCFF]
+[C;R;CCF]
+[C;R;CCNN]
+[C;R;CCNO]
+[C;R;CCNP]
+[C;R;CCNS]
+[C;R;CCN]
+[C;R;CCOO]
+[C;R;CCOS]
+[C;R;CCO]
+[C;R;CCSS]
+[C;R;CCS]
+[C;R;CClN]
+[C;R;CClS]
+[C;R;CFFO]
+[C;R;CIN]
+[C;R;CNNO]
+[C;R;CNN]
+[C;R;CNOO]
+[C;R;CNO]
+[C;R;CNS]
+[C;R;COO]
+[C;R;COS]
+[C;R;CSS]
+[C;R;ClClNS]
+[C;R;ClNN]
+[C;R;FFOO]
+[C;R;NNN]
+[C;R;NNO]
+[C;R;NNS]
+[C;R;NOO]
+[C;R;NOS]
+[C;R;NSS]
+[C;R;OOO]
+[C;R;OOS]
+[CH2;!R;BrC]
+[CH2;!R;BrS]
+[CH2;!R;CC]
+[CH2;!R;CCl]
+[CH2;!R;CF]
+[CH2;!R;CI]
+[CH2;!R;CN]
+[CH2;!R;CO]
+[CH2;!R;CP]
+[CH2;!R;CS]
+[CH2;!R;CSi]
+[CH2;!R;C]
+[CH2;!R;FO]
+[CH2;!R;NN]
+[CH2;!R;NO]
+[CH2;!R;NP]
+[CH2;!R;NS]
+[CH2;!R;NSi]
+[CH2;!R;OO]
+[CH2;!R;OP]
+[CH2;!R;OS]
+[CH2;!R;OSi]
+[CH2;!R;PS]
+[CH2;!R;SS]
+[CH2;!R;SSi]
+[CH2;R;CC]
+[CH2;R;CN]
+[CH2;R;CO]
+[CH2;R;CS]
+[CH2;R;CSi]
+[CH2;R;NN]
+[CH2;R;NO]
+[CH2;R;NP]
+[CH2;R;NS]
+[CH2;R;NSi]
+[CH2;R;OO]
+[CH2;R;OP]
+[CH2;R;OS]
+[CH2;R;SS]
+[CH3;!R;C]
+[CH3;!R;N]
+[CH3;!R;O]
+[CH3;!R;P]
+[CH3;!R;S]
+[CH3;!R;Si]
+[CH;!R;BrC]
+[CH;!R;CCC]
+[CH;!R;CCF]
+[CH;!R;CCN]
+[CH;!R;CCO]
+[CH;!R;CCP]
+[CH;!R;CCS]
+[CH;!R;CCSi]
+[CH;!R;CC]
+[CH;!R;CCl]
+[CH;!R;CFF]
+[CH;!R;CF]
+[CH;!R;CI]
+[CH;!R;CNN]
+[CH;!R;CNO]
+[CH;!R;CN]
+[CH;!R;COO]
+[CH;!R;CO]
+[CH;!R;CP]
+[CH;!R;CSS]
+[CH;!R;CS]
+[CH;!R;CSi]
+[CH;!R;C]
+[CH;!R;FFN]
+[CH;!R;FFO]
+[CH;!R;FFS]
+[CH;!R;NNN]
+[CH;!R;NN]
+[CH;!R;NOO]
+[CH;!R;NO]
+[CH;!R;NSiSi]
+[CH;!R;N]
+[CH;!R;OO]
+[CH;!R;PSS]
+[CH;R;CCC]
+[CH;R;CCCl]
+[CH;R;CCF]
+[CH;R;CCN]
+[CH;R;CCO]
+[CH;R;CCP]
+[CH;R;CCS]
+[CH;R;CC]
+[CH;R;CClN]
+[CH;R;CNN]
+[CH;R;CNO]
+[CH;R;CNS]
+[CH;R;CN]
+[CH;R;COO]
+[CH;R;CO]
+[CH;R;CP]
+[CH;R;CSS]
+[CH;R;CS]
+[CH;R;NNN]
+[CH;R;NNO]
+[CH;R;NNS]
+[CH;R;NN]
+[CH;R;NOO]
+[CH;R;NO]
+[CH;R;NS]
+[Cl;!R;C]
+[Cl;!R;N]
+[Cl;!R;S]
+[F;!R;C]
+[F;!R;N]
+[I;!R;C]
+[N;!R;CCC]
+[N;!R;CCN]
+[N;!R;CCO]
+[N;!R;CCP]
+[N;!R;CCS]
+[N;!R;CC]
+[N;!R;CNS]
+[N;!R;CN]
+[N;!R;COS]
+[N;!R;CO]
+[N;!R;CP]
+[N;!R;CSS]
+[N;!R;CS]
+[N;!R;CSiSi]
+[N;!R;C]
+[N;!R;ClClS]
+[N;!R;NNO]
+[N;!R;NO]
+[N;!R;N]
+[N;!R;SS]
+[N;R;BrCC]
+[N;R;CCC]
+[N;R;CCCl]
+[N;R;CCN]
+[N;R;CCO]
+[N;R;CCP]
+[N;R;CCS]
+[N;R;CC]
+[N;R;CNN]
+[N;R;CNO]
+[N;R;CNS]
+[N;R;CN]
+[N;R;COO]
+[N;R;COS]
+[N;R;CO]
+[N;R;CP]
+[N;R;CS]
+[N;R;NNO]
+[N;R;NN]
+[N;R;NO]
+[N;R;PP]
+[NH2;!R;C]
+[NH2;!R;N]
+[NH2;!R;O]
+[NH2;!R;S]
+[NH;!R;CC]
+[NH;!R;CN]
+[NH;!R;CO]
+[NH;!R;CP]
+[NH;!R;CS]
+[NH;!R;C]
+[NH;!R;NP]
+[NH;!R;NS]
+[NH;!R;OS]
+[NH;!R;S]
+[NH;R;CC]
+[NH;R;CN]
+[NH;R;CO]
+[NH;R;CS]
+[NH;R;NN]
+[O;!R;CC]
+[O;!R;CN]
+[O;!R;CP]
+[O;!R;CS]
+[O;!R;C]
+[O;!R;NN]
+[O;!R;N]
+[O;!R;P]
+[O;!R;S]
+[O;R;CC]
+[O;R;CN]
+[O;R;CP]
+[O;R;CS]
+[O;R;NN]
+[OH;!R;C]
+[OH;!R;N]
+[OH;!R;P]
+[P;!R;CCCO]
+[P;!R;CCC]
+[P;!R;CCN]
+[P;!R;CNNN]
+[P;!R;CNNO]
+[P;!R;COOO]
+[P;!R;NNOO]
+[P;!R;NOOO]
+[P;!R;OOOO]
+[P;!R;OOO]
+[P;R;CCCN]
+[P;R;CCNO]
+[P;R;CCOO]
+[P;R;COOOO]
+[P;R;COOO]
+[P;R;NNNN]
+[P;R;NNN]
+[P;R;NOOO]
+[P;R;NOO]
+[P;R;OOOOO]
+[P;R;OOO]
+[PH;!R;CC]
+[S;!R;CCNO]
+[S;!R;CCOO]
+[S;!R;CCO]
+[S;!R;CC]
+[S;!R;CClOO]
+[S;!R;CNOO]
+[S;!R;CN]
+[S;!R;COOO]
+[S;!R;CO]
+[S;!R;C]
+[S;!R;NNOO]
+[S;!R;NN]
+[S;!R;NOOO]
+[S;!R;OOOO]
+[S;R;CCNO]
+[S;R;CCOO]
+[S;R;CCO]
+[S;R;CC]
+[S;R;CNOO]
+[S;R;CN]
+[S;R;COOO]
+[S;R;CS]
+[S;R;NNN]
+[S;R;NNOO]
+[S;R;NN]
+[SH;!R;C]
+[SH;!R;NOO]
+[Si;!R;CCCC]
+[Si;!R;CCCN]
+[Si;R;CCCC]
+[[C@@H];!R;BrCC]
+[[C@@H];!R;CCC]
+[[C@@H];!R;CCF]
+[[C@@H];!R;CCN]
+[[C@@H];!R;CCO]
+[[C@@H];!R;CCP]
+[[C@@H];!R;CCS]
+[[C@@H];!R;CClS]
+[[C@@H];!R;CFN]
+[[C@@H];!R;CFO]
+[[C@@H];!R;CFS]
+[[C@@H];!R;CNN]
+[[C@@H];!R;CNO]
+[[C@@H];!R;CNP]
+[[C@@H];!R;CNS]
+[[C@@H];!R;COO]
+[[C@@H];!R;COP]
+[[C@@H];!R;COS]
+[[C@@H];!R;FNO]
+[[C@@H];!R;NOO]
+[[C@@H];R;CCC]
+[[C@@H];R;CCF]
+[[C@@H];R;CCN]
+[[C@@H];R;CCO]
+[[C@@H];R;CCS]
+[[C@@H];R;CNN]
+[[C@@H];R;CNO]
+[[C@@H];R;CNS]
+[[C@@H];R;COO]
+[[C@@H];R;COS]
+[[C@@H];R;CSS]
+[[C@@H];R;NNN]
+[[C@@H];R;NNS]
+[[C@@H];R;NOO]
+[[C@@];!R;BrCCCl]
+[[C@@];!R;BrCCI]
+[[C@@];!R;BrCCS]
+[[C@@];!R;BrCClS]
+[[C@@];!R;CCCC]
+[[C@@];!R;CCCF]
+[[C@@];!R;CCCN]
+[[C@@];!R;CCCO]
+[[C@@];!R;CCCS]
+[[C@@];!R;CCClF]
+[[C@@];!R;CCClS]
+[[C@@];!R;CCFN]
+[[C@@];!R;CCNN]
+[[C@@];!R;CCNO]
+[[C@@];!R;CCNP]
+[[C@@];!R;CCOO]
+[[C@@];!R;CCOP]
+[[C@@];!R;CClFI]
+[[C@@];!R;CClFO]
+[[C@@];!R;CClIS]
+[[C@@];R;BrCCC]
+[[C@@];R;BrCCN]
+[[C@@];R;BrCCO]
+[[C@@];R;CCCC]
+[[C@@];R;CCCCl]
+[[C@@];R;CCCF]
+[[C@@];R;CCCN]
+[[C@@];R;CCCO]
+[[C@@];R;CCCS]
+[[C@@];R;CCClF]
+[[C@@];R;CCClN]
+[[C@@];R;CCFN]
+[[C@@];R;CCNN]
+[[C@@];R;CCNO]
+[[C@@];R;CCNS]
+[[C@@];R;CCOO]
+[[C@@];R;CCOP]
+[[C@@];R;CCOS]
+[[C@@];R;CCSSi]
+[[C@@];R;CFOO]
+[[C@@];R;CNNO]
+[[C@@];R;CNOO]
+[[C@@];R;NNSS]
+[[C@H];!R;BrCC]
+[[C@H];!R;CCC]
+[[C@H];!R;CCF]
+[[C@H];!R;CCN]
+[[C@H];!R;CCO]
+[[C@H];!R;CCP]
+[[C@H];!R;CCS]
+[[C@H];!R;CClS]
+[[C@H];!R;CFN]
+[[C@H];!R;CNN]
+[[C@H];!R;CNO]
+[[C@H];!R;CNP]
+[[C@H];!R;CNS]
+[[C@H];!R;COO]
+[[C@H];!R;COP]
+[[C@H];!R;COS]
+[[C@H];!R;FNO]
+[[C@H];R;BrCC]
+[[C@H];R;CCC]
+[[C@H];R;CCF]
+[[C@H];R;CCN]
+[[C@H];R;CCO]
+[[C@H];R;CCP]
+[[C@H];R;CCS]
+[[C@H];R;CNN]
+[[C@H];R;CNO]
+[[C@H];R;CNS]
+[[C@H];R;COO]
+[[C@H];R;COS]
+[[C@H];R;NNO]
+[[C@H];R;NNS]
+[[C@H];R;NOO]
+[[C@H];R;OOO]
+[[C@];!R;BrCCF]
+[[C@];!R;BrCCS]
+[[C@];!R;BrCClS]
+[[C@];!R;BrCIS]
+[[C@];!R;CCCC]
+[[C@];!R;CCCCl]
+[[C@];!R;CCCF]
+[[C@];!R;CCCN]
+[[C@];!R;CCCO]
+[[C@];!R;CCCS]
+[[C@];!R;CCClF]
+[[C@];!R;CCClO]
+[[C@];!R;CCClS]
+[[C@];!R;CCFN]
+[[C@];!R;CCFO]
+[[C@];!R;CCNN]
+[[C@];!R;CCNO]
+[[C@];!R;CClFO]
+[[C@];!R;CClIS]
+[[C@];R;BrCCC]
+[[C@];R;BrCCCl]
+[[C@];R;CCCC]
+[[C@];R;CCCCl]
+[[C@];R;CCCF]
+[[C@];R;CCCI]
+[[C@];R;CCCN]
+[[C@];R;CCCO]
+[[C@];R;CCCS]
+[[C@];R;CCClF]
+[[C@];R;CCFN]
+[[C@];R;CCNN]
+[[C@];R;CCNO]
+[[C@];R;CCNS]
+[[C@];R;CCOO]
+[[C@];R;CCOP]
+[[C@];R;CCOSi]
+[[C@];R;CNNO]
+[[C@];R;CNNS]
+[[C@];R;CNOO]
+[[C@];R;NNNS]
+[[C@];R;NNSS]
+[[N+];!R;CCC]
+[[N+];!R;CC]
+[[N+];!R;CN]
+[[N+];!R;COO]
+[[N+];!R;NN]
+[[N+];!R;NOO]
+[[N+];!R;OOO]
+[[N+];R;CCC]
+[[N+];R;CCN]
+[[N+];R;CCO]
+[[N+];R;NNO]
+[[N+];R;NOO]
+[[N-];!R;CC]
+[[N-];!R;CN]
+[[N-];!R;CO]
+[[N-];!R;CS]
+[[N-];!R;NS]
+[[N-];!R;N]
+[[N-];R;CC]
+[[N-];R;CN]
+[[N-];R;CO]
+[[N-];R;CS]
+[[NH+];!R;CCC]
+[[NH+];!R;CC]
+[[NH+];!R;CN]
+[[NH+];!R;CO]
+[[NH+];!R;CP]
+[[NH+];R;CCC]
+[[NH+];R;CC]
+[[NH+];R;CN]
+[[NH+];R;CO]
+[[NH2+];!R;CC]
+[[NH2+];!R;CN]
+[[NH2+];!R;C]
+[[NH2+];R;CC]
+[[NH2+];R;CN]
+[[NH3+];!R;C]
+[[NH3+];!R;N]
+[[O+];R;CN]
+[[O-];!R;C]
+[[O-];!R;N]
+[[O-];!R;P]
+[[O-];!R;S]
+[[P@@];!R;CCCO]
+[[P@@];!R;CCNN]
+[[P@@];!R;CCNO]
+[[P@@];!R;CCOO]
+[[P@@];!R;CNOO]
+[[P@@];!R;COOO]
+[[P@@];R;CCCO]
+[[P@@];R;CCOO]
+[[P@@];R;NOOO]
+[[P@@];R;OOOO]
+[[P@H];!R;OOO]
+[[P@];!R;CCNN]
+[[P@];!R;CCNO]
+[[P@];!R;CCOO]
+[[P@];!R;CNOO]
+[[P@];!R;COOO]
+[[P@];!R;NNOO]
+[[P@];!R;NOOO]
+[[P@];R;CCCO]
+[[P@];R;CNNO]
+[[PH+];!R;CCC]
+[[PH+];!R;CCN]
+[[S+];!R;CCC]
+[[S-];!R;C]
+[[S@@];!R;CCNO]
+[[S@@];!R;CCN]
+[[S@@];!R;CCO]
+[[S@@];!R;CNNO]
+[[S@@];!R;CNOO]
+[[S@@];!R;CNO]
+[[S@@];!R;COO]
+[[S@@];R;CCNO]
+[[S@@];R;CCO]
+[[S@];!R;CCNO]
+[[S@];!R;CCN]
+[[S@];!R;CCO]
+[[S@];!R;CNNO]
+[[S@];!R;CNOO]
+[[S@];!R;CNO]
+[[S@];!R;NOO]
+[[S@];R;CCNO]
+[[S@];R;CCO]
+[[[N@@H]+];!R;CCC]
+[[[N@@H]+];R;CCC]
+[[[N@@H]+];R;CCN]
+[[[N@H]+];!R;CCC]
+[[[N@H]+];R;CCC]
+[[n+];R;CCC]
+[[n+];R;CCN]
+[[n+];R;CCO]
+[[n+];R;CCS]
+[[n+];R;CNN]
+[[n+];R;CNO]
+[[n+];R;COO]
+[[n+];R;NNN]
+[[n-];R;CC]
+[[n-];R;CN]
+[[n-];R;CO]
+[[n-];R;CS]
+[[n-];R;NN]
+[[n-];R;NO]
+[[n-];R;NS]
+[[nH+];R;CC]
+[[nH+];R;CN]
+[[o+];R;CC]
+[[o+];R;CN]
+[[s+];R;CC]
+[[s+];R;CN]
+[c;R;BrCC]
+[c;R;BrCN]
+[c;R;BrCO]
+[c;R;BrCS]
+[c;R;BrNN]
+[c;R;BrNO]
+[c;R;BrNS]
+[c;R;CCC]
+[c;R;CCCl]
+[c;R;CCF]
+[c;R;CCI]
+[c;R;CCN]
+[c;R;CCO]
+[c;R;CCP]
+[c;R;CCS]
+[c;R;CCSi]
+[c;R;CClN]
+[c;R;CClO]
+[c;R;CClS]
+[c;R;CFN]
+[c;R;CFO]
+[c;R;CFS]
+[c;R;CIN]
+[c;R;CIO]
+[c;R;CIS]
+[c;R;CNN]
+[c;R;CNO]
+[c;R;CNP]
+[c;R;CNS]
+[c;R;CNSi]
+[c;R;COO]
+[c;R;COP]
+[c;R;COS]
+[c;R;CSS]
+[c;R;ClNN]
+[c;R;ClNO]
+[c;R;ClNS]
+[c;R;FNN]
+[c;R;INN]
+[c;R;INO]
+[c;R;INS]
+[c;R;NNN]
+[c;R;NNO]
+[c;R;NNP]
+[c;R;NNS]
+[c;R;NOO]
+[c;R;NOS]
+[c;R;NSS]
+[c;R;OOO]
+[c;R;OOS]
+[c;R;OSS]
+[c;R;SSS]
+[cH;R;CC]
+[cH;R;CN]
+[cH;R;CO]
+[cH;R;CS]
+[cH;R;NN]
+[cH;R;NO]
+[cH;R;NS]
+[n;R;BrCN]
+[n;R;CCC]
+[n;R;CCF]
+[n;R;CCN]
+[n;R;CCO]
+[n;R;CCP]
+[n;R;CCS]
+[n;R;CCSi]
+[n;R;CC]
+[n;R;CNN]
+[n;R;CNO]
+[n;R;CNS]
+[n;R;CN]
+[n;R;CO]
+[n;R;CS]
+[n;R;NN]
+[n;R;NO]
+[n;R;NS]
+[nH;R;CC]
+[nH;R;CN]
+[nH;R;CO]
+[nH;R;CS]
+[nH;R;NN]
+[o;R;CC]
+[o;R;CN]
+[o;R;NN]
+[p;R;CC]
+[s;R;CC]
+[s;R;CN]
+[s;R;CS]
+[s;R;NN]
+\

run-0/checkpoint-6/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a4b766f685049ba65fd5771463826b62c7df1d1a4a20f21847a78bc102475f5
+oid sha256:fb3a452baf441fd314458146c1d514b27cc94ebdd83cd7c9527fae9563ffdd0e
 size 137970488

run-0/checkpoint-6/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4f7de65616c87c0ee68a03f0d459e947b5c387dc5ff20f3dae43556e7c0d5da9
+oid sha256:2dfe1071a486d8e0ee3b2a295de09ddcd9cdda43fc96fe25a6c1b589576f8fe1
 size 276010426

run-0/checkpoint-6/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b6953f1a7495c9748bd2eb1e64e84fdc5306c73e91d27ba1e270697965707d7f
+oid sha256:cf09e5d42068af78718f389b253ca2eb25c713b626107d2bb3a00c3751c8bf18
 size 14244

run-0/checkpoint-6/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f16a410c016fcc4ab94d35b8fa2bcc70b6ee14abbc3ccfa55e4912befe64fa3
+oid sha256:42dd2d64f1cf85ac9b5e2307b25ef62cb00c888b1773a8bc919c264d5281730c
 size 1064

run-0/checkpoint-6/trainer_state.json

@@ -2,67 +2,21 @@
   "best_metric": null,
   "best_model_checkpoint": null,
   "epoch": 0.3,
-  "eval_steps": 2,
+  "eval_steps": 10,
   "global_step": 6,
   "is_hyper_param_search": true,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
-  "log_history": [
-    {
-      "epoch": 0.1,
-      "grad_norm": 815587.25,
-      "learning_rate": 0.0013426765493225812,
-      "loss": 6.8501,
-      "step": 2
-    },
-    {
-      "epoch": 0.1,
-      "eval_loss": 6.029593467712402,
-      "eval_runtime": 16.6784,
-      "eval_samples_per_second": 599.579,
-      "eval_steps_per_second": 1.139,
-      "step": 2
-    },
-    {
-      "epoch": 0.2,
-      "grad_norm": 213595.953125,
-      "learning_rate": 0.0013426765493225812,
-      "loss": 5.4602,
-      "step": 4
-    },
-    {
-      "epoch": 0.2,
-      "eval_loss": 4.330280303955078,
-      "eval_runtime": 16.5988,
-      "eval_samples_per_second": 602.452,
-      "eval_steps_per_second": 1.145,
-      "step": 4
-    },
-    {
-      "epoch": 0.3,
-      "grad_norm": 120017.78125,
-      "learning_rate": 0.0013426765493225812,
-      "loss": 4.1818,
-      "step": 6
-    },
-    {
-      "epoch": 0.3,
-      "eval_loss": 3.8199374675750732,
-      "eval_runtime": 15.967,
-      "eval_samples_per_second": 626.292,
-      "eval_steps_per_second": 1.19,
-      "step": 6
-    }
-  ],
-  "logging_steps": 2,
+  "log_history": [],
+  "logging_steps": 10,
   "max_steps": 20,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 9223372036854775807,
   "save_steps": 6,
-  "total_flos": 50190024038400.0,
+  "total_flos": 0,
   "train_batch_size": 544,
   "trial_name": null,
   "trial_params": {
-    "learning_rate": 0.0013426765493225812
+    "learning_rate": 0.00011654557407940855
   }
 }

run-0/checkpoint-6/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4430642746794e99e831575cb552a62fbb8c3702764acd26a805d45e8f4798d7
+oid sha256:3c0adb69a864270aa8734e61821561a06ca3d35d3cd81446c72576c4c967a811
 size 4920

runs/Mar29_23-30-55_b169022313d8/events.out.tfevents.1711755057.b169022313d8.400.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a77fdf2bc957fe11f303ecf0a7a08d910ff1ffc3ccc6e61271fc3a3c360465bc
+size 5032

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4430642746794e99e831575cb552a62fbb8c3702764acd26a805d45e8f4798d7
+oid sha256:3c0adb69a864270aa8734e61821561a06ca3d35d3cd81446c72576c4c967a811
 size 4920