Training in progress, step 500, checkpoint
Browse files- checkpoint-500/config.json +195 -0
- checkpoint-500/generation_config.json +4 -0
- checkpoint-500/model.safetensors +3 -0
- checkpoint-500/optimizer.pt +3 -0
- checkpoint-500/rng_state_0.pth +3 -0
- checkpoint-500/rng_state_1.pth +3 -0
- checkpoint-500/rng_state_2.pth +3 -0
- checkpoint-500/rng_state_3.pth +3 -0
- checkpoint-500/scheduler.pt +3 -0
- checkpoint-500/trainer_state.json +379 -0
- checkpoint-500/training_args.bin +3 -0
checkpoint-500/config.json
ADDED
|
@@ -0,0 +1,195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SrcProberForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"asm_encoder_config": {
|
| 6 |
+
"_name_or_path": "",
|
| 7 |
+
"add_cross_attention": false,
|
| 8 |
+
"architectures": [
|
| 9 |
+
"LongelmForMaskedLMWithEdgePrediction"
|
| 10 |
+
],
|
| 11 |
+
"attention_probs_dropout_prob": 0.1,
|
| 12 |
+
"bad_words_ids": null,
|
| 13 |
+
"begin_suppress_tokens": null,
|
| 14 |
+
"block_size": 8,
|
| 15 |
+
"bos_token_id": 1,
|
| 16 |
+
"chunk_size_feed_forward": 0,
|
| 17 |
+
"classifier_dropout": null,
|
| 18 |
+
"cross_attention_hidden_size": null,
|
| 19 |
+
"decoder_start_token_id": null,
|
| 20 |
+
"diversity_penalty": 0.0,
|
| 21 |
+
"do_sample": false,
|
| 22 |
+
"early_stopping": false,
|
| 23 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 24 |
+
"eos_token_id": 2,
|
| 25 |
+
"ep_add_linear_projection": true,
|
| 26 |
+
"exponential_decay_length_penalty": null,
|
| 27 |
+
"finetuning_task": null,
|
| 28 |
+
"forced_bos_token_id": null,
|
| 29 |
+
"forced_eos_token_id": null,
|
| 30 |
+
"global_memory_size": 1,
|
| 31 |
+
"hidden_act": "gelu",
|
| 32 |
+
"hidden_dropout_prob": 0.1,
|
| 33 |
+
"hidden_size": 768,
|
| 34 |
+
"id2label": {
|
| 35 |
+
"0": "LABEL_0",
|
| 36 |
+
"1": "LABEL_1"
|
| 37 |
+
},
|
| 38 |
+
"initializer_range": 0.02,
|
| 39 |
+
"intermediate_size": 3072,
|
| 40 |
+
"is_decoder": false,
|
| 41 |
+
"is_encoder_decoder": false,
|
| 42 |
+
"label2id": {
|
| 43 |
+
"LABEL_0": 0,
|
| 44 |
+
"LABEL_1": 1
|
| 45 |
+
},
|
| 46 |
+
"layer_norm_eps": 1e-12,
|
| 47 |
+
"length_penalty": 1.0,
|
| 48 |
+
"max_blocks": 200,
|
| 49 |
+
"max_length": 20,
|
| 50 |
+
"max_position_embeddings": 1805,
|
| 51 |
+
"max_relative_position_embeddings": 8,
|
| 52 |
+
"min_length": 0,
|
| 53 |
+
"model_type": "longelm",
|
| 54 |
+
"no_repeat_ngram_size": 0,
|
| 55 |
+
"node_size": 1,
|
| 56 |
+
"num_attention_heads": 12,
|
| 57 |
+
"num_beam_groups": 1,
|
| 58 |
+
"num_beams": 1,
|
| 59 |
+
"num_hidden_layers": 12,
|
| 60 |
+
"num_return_sequences": 1,
|
| 61 |
+
"output_attentions": false,
|
| 62 |
+
"output_hidden_states": false,
|
| 63 |
+
"output_scores": false,
|
| 64 |
+
"pad_token_id": 3,
|
| 65 |
+
"position_embedding_type": "mixed",
|
| 66 |
+
"prefix": null,
|
| 67 |
+
"problem_type": null,
|
| 68 |
+
"pruned_heads": {},
|
| 69 |
+
"remove_invalid_values": false,
|
| 70 |
+
"repetition_penalty": 1.0,
|
| 71 |
+
"return_dict": true,
|
| 72 |
+
"return_dict_in_generate": false,
|
| 73 |
+
"sep_token_id": null,
|
| 74 |
+
"suppress_tokens": null,
|
| 75 |
+
"task_specific_params": null,
|
| 76 |
+
"temperature": 1.0,
|
| 77 |
+
"tf_legacy_loss": false,
|
| 78 |
+
"tie_encoder_decoder": false,
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"tokenizer_class": null,
|
| 81 |
+
"top_k": 50,
|
| 82 |
+
"top_p": 1.0,
|
| 83 |
+
"torch_dtype": "float32",
|
| 84 |
+
"torchscript": false,
|
| 85 |
+
"type_vocab_size": 2,
|
| 86 |
+
"typical_p": 1.0,
|
| 87 |
+
"use_bfloat16": false,
|
| 88 |
+
"use_cache": true,
|
| 89 |
+
"vocab_size": 50265
|
| 90 |
+
},
|
| 91 |
+
"asm_feature_select_strategy": "nodes",
|
| 92 |
+
"asm_token_index": 32016,
|
| 93 |
+
"ignore_index": -100,
|
| 94 |
+
"model_type": "rad-prober",
|
| 95 |
+
"pad_token_id": 2,
|
| 96 |
+
"projector_hidden_act": "gelu",
|
| 97 |
+
"src_lm_config": {
|
| 98 |
+
"_name_or_path": "codellama/CodeLlama-13b-hf",
|
| 99 |
+
"add_cross_attention": false,
|
| 100 |
+
"architectures": [
|
| 101 |
+
"LlamaForCausalLM"
|
| 102 |
+
],
|
| 103 |
+
"attention_bias": false,
|
| 104 |
+
"attention_dropout": 0.0,
|
| 105 |
+
"bad_words_ids": null,
|
| 106 |
+
"begin_suppress_tokens": null,
|
| 107 |
+
"bos_token_id": 1,
|
| 108 |
+
"chunk_size_feed_forward": 0,
|
| 109 |
+
"cross_attention_hidden_size": null,
|
| 110 |
+
"decoder_start_token_id": null,
|
| 111 |
+
"diversity_penalty": 0.0,
|
| 112 |
+
"do_sample": false,
|
| 113 |
+
"early_stopping": false,
|
| 114 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 115 |
+
"eos_token_id": 2,
|
| 116 |
+
"exponential_decay_length_penalty": null,
|
| 117 |
+
"finetuning_task": null,
|
| 118 |
+
"forced_bos_token_id": null,
|
| 119 |
+
"forced_eos_token_id": null,
|
| 120 |
+
"hidden_act": "silu",
|
| 121 |
+
"hidden_size": 5120,
|
| 122 |
+
"id2label": {
|
| 123 |
+
"0": "LABEL_0",
|
| 124 |
+
"1": "LABEL_1"
|
| 125 |
+
},
|
| 126 |
+
"initializer_range": 0.02,
|
| 127 |
+
"intermediate_size": 13824,
|
| 128 |
+
"is_decoder": false,
|
| 129 |
+
"is_encoder_decoder": false,
|
| 130 |
+
"label2id": {
|
| 131 |
+
"LABEL_0": 0,
|
| 132 |
+
"LABEL_1": 1
|
| 133 |
+
},
|
| 134 |
+
"length_penalty": 1.0,
|
| 135 |
+
"max_length": 20,
|
| 136 |
+
"max_position_embeddings": 16384,
|
| 137 |
+
"min_length": 0,
|
| 138 |
+
"model_type": "llama",
|
| 139 |
+
"no_repeat_ngram_size": 0,
|
| 140 |
+
"num_attention_heads": 40,
|
| 141 |
+
"num_beam_groups": 1,
|
| 142 |
+
"num_beams": 1,
|
| 143 |
+
"num_hidden_layers": 40,
|
| 144 |
+
"num_key_value_heads": 40,
|
| 145 |
+
"num_return_sequences": 1,
|
| 146 |
+
"output_attentions": false,
|
| 147 |
+
"output_hidden_states": false,
|
| 148 |
+
"output_scores": false,
|
| 149 |
+
"pad_token_id": null,
|
| 150 |
+
"prefix": null,
|
| 151 |
+
"pretraining_tp": 1,
|
| 152 |
+
"problem_type": null,
|
| 153 |
+
"pruned_heads": {},
|
| 154 |
+
"quantization_config": {
|
| 155 |
+
"_load_in_4bit": false,
|
| 156 |
+
"_load_in_8bit": true,
|
| 157 |
+
"bnb_4bit_compute_dtype": "float32",
|
| 158 |
+
"bnb_4bit_quant_storage": "uint8",
|
| 159 |
+
"bnb_4bit_quant_type": "fp4",
|
| 160 |
+
"bnb_4bit_use_double_quant": false,
|
| 161 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
| 162 |
+
"llm_int8_has_fp16_weight": false,
|
| 163 |
+
"llm_int8_skip_modules": null,
|
| 164 |
+
"llm_int8_threshold": 6.0,
|
| 165 |
+
"load_in_4bit": false,
|
| 166 |
+
"load_in_8bit": true,
|
| 167 |
+
"quant_method": "bitsandbytes"
|
| 168 |
+
},
|
| 169 |
+
"remove_invalid_values": false,
|
| 170 |
+
"repetition_penalty": 1.0,
|
| 171 |
+
"return_dict": true,
|
| 172 |
+
"return_dict_in_generate": false,
|
| 173 |
+
"rms_norm_eps": 1e-05,
|
| 174 |
+
"rope_scaling": null,
|
| 175 |
+
"rope_theta": 1000000,
|
| 176 |
+
"sep_token_id": null,
|
| 177 |
+
"suppress_tokens": null,
|
| 178 |
+
"task_specific_params": null,
|
| 179 |
+
"temperature": 1.0,
|
| 180 |
+
"tf_legacy_loss": false,
|
| 181 |
+
"tie_encoder_decoder": false,
|
| 182 |
+
"tie_word_embeddings": false,
|
| 183 |
+
"tokenizer_class": null,
|
| 184 |
+
"top_k": 50,
|
| 185 |
+
"top_p": 1.0,
|
| 186 |
+
"torch_dtype": "bfloat16",
|
| 187 |
+
"torchscript": false,
|
| 188 |
+
"typical_p": 1.0,
|
| 189 |
+
"use_bfloat16": false,
|
| 190 |
+
"use_cache": true,
|
| 191 |
+
"vocab_size": 32017
|
| 192 |
+
},
|
| 193 |
+
"torch_dtype": "float32",
|
| 194 |
+
"transformers_version": "4.39.3"
|
| 195 |
+
}
|
checkpoint-500/generation_config.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"transformers_version": "4.39.3"
|
| 4 |
+
}
|
checkpoint-500/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:97a2954d1ea984e824ced1d24c53c58376f92b1fd6666127ab4f82d5fd7d0f78
|
| 3 |
+
size 620848256
|
checkpoint-500/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:05540d6d8ea733912db63bc75f69249f41504587cd0d3d8ba03eb3796b2f20d4
|
| 3 |
+
size 297976223
|
checkpoint-500/rng_state_0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38c59aafd474ee5fd0faf37d860b2545cccc62f101ecec021456558c324efe4f
|
| 3 |
+
size 14960
|
checkpoint-500/rng_state_1.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c7050b7c336ab263ea32a6317ab92a0e128f688355724aafc8b18448bd5984d9
|
| 3 |
+
size 14960
|
checkpoint-500/rng_state_2.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a70f7d9acdbb2cfd3abec5bcd0e43728c98dfe209686a95dbe46d62be22d7ff0
|
| 3 |
+
size 14960
|
checkpoint-500/rng_state_3.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a80f9f2812da80cd0e89c494b6dcc7e3e6e8b0ffb56bbb3d47301828e667edb6
|
| 3 |
+
size 14960
|
checkpoint-500/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fce5f75097acf44fa3921c99b5dc314a613a3e0ceec415094bdb8cf993a059ef
|
| 3 |
+
size 1064
|
checkpoint-500/trainer_state.json
ADDED
|
@@ -0,0 +1,379 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.1208313194780087,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 500,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0,
|
| 13 |
+
"grad_norm": NaN,
|
| 14 |
+
"learning_rate": 2.0000000000000002e-07,
|
| 15 |
+
"loss": 2.0414,
|
| 16 |
+
"step": 10
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.0,
|
| 20 |
+
"grad_norm": 53.515953063964844,
|
| 21 |
+
"learning_rate": 6.5e-07,
|
| 22 |
+
"loss": 1.8355,
|
| 23 |
+
"step": 20
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.01,
|
| 27 |
+
"grad_norm": 27.920907974243164,
|
| 28 |
+
"learning_rate": 1.15e-06,
|
| 29 |
+
"loss": 1.5172,
|
| 30 |
+
"step": 30
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.01,
|
| 34 |
+
"grad_norm": 20.47172737121582,
|
| 35 |
+
"learning_rate": 1.65e-06,
|
| 36 |
+
"loss": 1.2654,
|
| 37 |
+
"step": 40
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.01,
|
| 41 |
+
"grad_norm": 8.903529167175293,
|
| 42 |
+
"learning_rate": 2.1499999999999997e-06,
|
| 43 |
+
"loss": 1.1095,
|
| 44 |
+
"step": 50
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.01,
|
| 48 |
+
"grad_norm": 5.902841567993164,
|
| 49 |
+
"learning_rate": 2.65e-06,
|
| 50 |
+
"loss": 1.0017,
|
| 51 |
+
"step": 60
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.02,
|
| 55 |
+
"grad_norm": 2.9340908527374268,
|
| 56 |
+
"learning_rate": 3.1500000000000003e-06,
|
| 57 |
+
"loss": 0.8846,
|
| 58 |
+
"step": 70
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.02,
|
| 62 |
+
"grad_norm": 8.701367378234863,
|
| 63 |
+
"learning_rate": 3.6499999999999998e-06,
|
| 64 |
+
"loss": 0.8572,
|
| 65 |
+
"step": 80
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.02,
|
| 69 |
+
"grad_norm": 1.3205362558364868,
|
| 70 |
+
"learning_rate": 4.15e-06,
|
| 71 |
+
"loss": 0.839,
|
| 72 |
+
"step": 90
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.02,
|
| 76 |
+
"grad_norm": 3.0334558486938477,
|
| 77 |
+
"learning_rate": 4.65e-06,
|
| 78 |
+
"loss": 0.7838,
|
| 79 |
+
"step": 100
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.03,
|
| 83 |
+
"grad_norm": 1.8086520433425903,
|
| 84 |
+
"learning_rate": 5.15e-06,
|
| 85 |
+
"loss": 0.7782,
|
| 86 |
+
"step": 110
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.03,
|
| 90 |
+
"grad_norm": 29.50135040283203,
|
| 91 |
+
"learning_rate": 5.65e-06,
|
| 92 |
+
"loss": 0.7883,
|
| 93 |
+
"step": 120
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.03,
|
| 97 |
+
"grad_norm": 3.2708683013916016,
|
| 98 |
+
"learning_rate": 6.15e-06,
|
| 99 |
+
"loss": 0.7961,
|
| 100 |
+
"step": 130
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.03,
|
| 104 |
+
"grad_norm": 33.43790817260742,
|
| 105 |
+
"learning_rate": 6.650000000000001e-06,
|
| 106 |
+
"loss": 0.7626,
|
| 107 |
+
"step": 140
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.04,
|
| 111 |
+
"grad_norm": 1.2334959506988525,
|
| 112 |
+
"learning_rate": 7.15e-06,
|
| 113 |
+
"loss": 0.7732,
|
| 114 |
+
"step": 150
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.04,
|
| 118 |
+
"grad_norm": 1.7489686012268066,
|
| 119 |
+
"learning_rate": 7.65e-06,
|
| 120 |
+
"loss": 0.7715,
|
| 121 |
+
"step": 160
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.04,
|
| 125 |
+
"grad_norm": 3.358823776245117,
|
| 126 |
+
"learning_rate": 8.15e-06,
|
| 127 |
+
"loss": 0.7778,
|
| 128 |
+
"step": 170
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.04,
|
| 132 |
+
"grad_norm": 4.348245620727539,
|
| 133 |
+
"learning_rate": 8.65e-06,
|
| 134 |
+
"loss": 0.7619,
|
| 135 |
+
"step": 180
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.05,
|
| 139 |
+
"grad_norm": 0.9023125767707825,
|
| 140 |
+
"learning_rate": 9.15e-06,
|
| 141 |
+
"loss": 0.7746,
|
| 142 |
+
"step": 190
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"epoch": 0.05,
|
| 146 |
+
"grad_norm": 0.9158114790916443,
|
| 147 |
+
"learning_rate": 9.65e-06,
|
| 148 |
+
"loss": 0.7592,
|
| 149 |
+
"step": 200
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.05,
|
| 153 |
+
"grad_norm": 0.6256226301193237,
|
| 154 |
+
"learning_rate": 1.0150000000000001e-05,
|
| 155 |
+
"loss": 0.7787,
|
| 156 |
+
"step": 210
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"epoch": 0.05,
|
| 160 |
+
"grad_norm": 0.3805778920650482,
|
| 161 |
+
"learning_rate": 1.065e-05,
|
| 162 |
+
"loss": 0.7811,
|
| 163 |
+
"step": 220
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.06,
|
| 167 |
+
"grad_norm": 0.6968041062355042,
|
| 168 |
+
"learning_rate": 1.115e-05,
|
| 169 |
+
"loss": 0.73,
|
| 170 |
+
"step": 230
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.06,
|
| 174 |
+
"grad_norm": 0.9035410284996033,
|
| 175 |
+
"learning_rate": 1.1650000000000002e-05,
|
| 176 |
+
"loss": 0.7518,
|
| 177 |
+
"step": 240
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 0.06,
|
| 181 |
+
"grad_norm": 1.6727488040924072,
|
| 182 |
+
"learning_rate": 1.215e-05,
|
| 183 |
+
"loss": 0.7805,
|
| 184 |
+
"step": 250
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"epoch": 0.06,
|
| 188 |
+
"grad_norm": 0.4153461158275604,
|
| 189 |
+
"learning_rate": 1.2650000000000001e-05,
|
| 190 |
+
"loss": 0.7706,
|
| 191 |
+
"step": 260
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.07,
|
| 195 |
+
"grad_norm": 1.4846651554107666,
|
| 196 |
+
"learning_rate": 1.3150000000000001e-05,
|
| 197 |
+
"loss": 0.7492,
|
| 198 |
+
"step": 270
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 0.07,
|
| 202 |
+
"grad_norm": 2.6631388664245605,
|
| 203 |
+
"learning_rate": 1.3650000000000001e-05,
|
| 204 |
+
"loss": 0.7681,
|
| 205 |
+
"step": 280
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.07,
|
| 209 |
+
"grad_norm": 0.6325013041496277,
|
| 210 |
+
"learning_rate": 1.415e-05,
|
| 211 |
+
"loss": 0.7753,
|
| 212 |
+
"step": 290
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.07,
|
| 216 |
+
"grad_norm": 0.877907395362854,
|
| 217 |
+
"learning_rate": 1.465e-05,
|
| 218 |
+
"loss": 0.7188,
|
| 219 |
+
"step": 300
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.07,
|
| 223 |
+
"grad_norm": 0.37142279744148254,
|
| 224 |
+
"learning_rate": 1.515e-05,
|
| 225 |
+
"loss": 0.7204,
|
| 226 |
+
"step": 310
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.08,
|
| 230 |
+
"grad_norm": 0.7337246537208557,
|
| 231 |
+
"learning_rate": 1.565e-05,
|
| 232 |
+
"loss": 0.778,
|
| 233 |
+
"step": 320
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.08,
|
| 237 |
+
"grad_norm": 0.5847220420837402,
|
| 238 |
+
"learning_rate": 1.6150000000000003e-05,
|
| 239 |
+
"loss": 0.7288,
|
| 240 |
+
"step": 330
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 0.08,
|
| 244 |
+
"grad_norm": 3.745180606842041,
|
| 245 |
+
"learning_rate": 1.665e-05,
|
| 246 |
+
"loss": 0.7531,
|
| 247 |
+
"step": 340
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"epoch": 0.08,
|
| 251 |
+
"grad_norm": 0.357301265001297,
|
| 252 |
+
"learning_rate": 1.7150000000000004e-05,
|
| 253 |
+
"loss": 0.7448,
|
| 254 |
+
"step": 350
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 0.09,
|
| 258 |
+
"grad_norm": 0.9032486081123352,
|
| 259 |
+
"learning_rate": 1.765e-05,
|
| 260 |
+
"loss": 0.7651,
|
| 261 |
+
"step": 360
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"epoch": 0.09,
|
| 265 |
+
"grad_norm": 0.2864232361316681,
|
| 266 |
+
"learning_rate": 1.815e-05,
|
| 267 |
+
"loss": 0.7193,
|
| 268 |
+
"step": 370
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 0.09,
|
| 272 |
+
"grad_norm": 1.8560261726379395,
|
| 273 |
+
"learning_rate": 1.865e-05,
|
| 274 |
+
"loss": 0.7421,
|
| 275 |
+
"step": 380
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 0.09,
|
| 279 |
+
"grad_norm": 0.2971792221069336,
|
| 280 |
+
"learning_rate": 1.915e-05,
|
| 281 |
+
"loss": 0.7171,
|
| 282 |
+
"step": 390
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.1,
|
| 286 |
+
"grad_norm": 0.40850627422332764,
|
| 287 |
+
"learning_rate": 1.9650000000000003e-05,
|
| 288 |
+
"loss": 0.7459,
|
| 289 |
+
"step": 400
|
| 290 |
+
},
|
| 291 |
+
{
|
| 292 |
+
"epoch": 0.1,
|
| 293 |
+
"grad_norm": 0.3934139311313629,
|
| 294 |
+
"learning_rate": 2.0150000000000002e-05,
|
| 295 |
+
"loss": 0.7205,
|
| 296 |
+
"step": 410
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 0.1,
|
| 300 |
+
"grad_norm": 1.8674131631851196,
|
| 301 |
+
"learning_rate": 2.065e-05,
|
| 302 |
+
"loss": 0.752,
|
| 303 |
+
"step": 420
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"epoch": 0.1,
|
| 307 |
+
"grad_norm": 0.36707818508148193,
|
| 308 |
+
"learning_rate": 2.115e-05,
|
| 309 |
+
"loss": 0.7494,
|
| 310 |
+
"step": 430
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.11,
|
| 314 |
+
"grad_norm": 0.5044310092926025,
|
| 315 |
+
"learning_rate": 2.165e-05,
|
| 316 |
+
"loss": 0.7595,
|
| 317 |
+
"step": 440
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"epoch": 0.11,
|
| 321 |
+
"grad_norm": 1.9025150537490845,
|
| 322 |
+
"learning_rate": 2.215e-05,
|
| 323 |
+
"loss": 0.7379,
|
| 324 |
+
"step": 450
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.11,
|
| 328 |
+
"grad_norm": 0.5004140138626099,
|
| 329 |
+
"learning_rate": 2.265e-05,
|
| 330 |
+
"loss": 0.7886,
|
| 331 |
+
"step": 460
|
| 332 |
+
},
|
| 333 |
+
{
|
| 334 |
+
"epoch": 0.11,
|
| 335 |
+
"grad_norm": 3.544482707977295,
|
| 336 |
+
"learning_rate": 2.3150000000000004e-05,
|
| 337 |
+
"loss": 0.7259,
|
| 338 |
+
"step": 470
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 0.12,
|
| 342 |
+
"grad_norm": 0.33023321628570557,
|
| 343 |
+
"learning_rate": 2.365e-05,
|
| 344 |
+
"loss": 0.7333,
|
| 345 |
+
"step": 480
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.12,
|
| 349 |
+
"grad_norm": 0.3548080325126648,
|
| 350 |
+
"learning_rate": 2.415e-05,
|
| 351 |
+
"loss": 0.7527,
|
| 352 |
+
"step": 490
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.12,
|
| 356 |
+
"grad_norm": 0.42450150847435,
|
| 357 |
+
"learning_rate": 2.465e-05,
|
| 358 |
+
"loss": 0.7443,
|
| 359 |
+
"step": 500
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 0.12,
|
| 363 |
+
"eval_loss": 0.7428915500640869,
|
| 364 |
+
"eval_runtime": 68.3143,
|
| 365 |
+
"eval_samples_per_second": 29.276,
|
| 366 |
+
"eval_steps_per_second": 0.922,
|
| 367 |
+
"step": 500
|
| 368 |
+
}
|
| 369 |
+
],
|
| 370 |
+
"logging_steps": 10,
|
| 371 |
+
"max_steps": 20690,
|
| 372 |
+
"num_input_tokens_seen": 0,
|
| 373 |
+
"num_train_epochs": 5,
|
| 374 |
+
"save_steps": 500,
|
| 375 |
+
"total_flos": 1.0807388003190702e+18,
|
| 376 |
+
"train_batch_size": 4,
|
| 377 |
+
"trial_name": null,
|
| 378 |
+
"trial_params": null
|
| 379 |
+
}
|
checkpoint-500/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd6a31d2da7b4686068d511137b843e541147626d5878a410233f3ef91ae5dc6
|
| 3 |
+
size 5048
|