Upload folder using huggingface_hub
Browse files- config.json +32 -0
- generation_config.json +7 -0
- model.safetensors +3 -0
- optimizer.pt +3 -0
- rng_state.pth +3 -0
- scheduler.pt +3 -0
- special_tokens_map.json +30 -0
- tokenizer.json +0 -0
- tokenizer_config.json +43 -0
- trainer_state.json +739 -0
- training_args.bin +3 -0
config.json
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "pretrain_anneal/checkpoint-16912",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"LlamaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"attention_bias": false,
|
| 7 |
+
"attention_dropout": 0.0,
|
| 8 |
+
"bos_token_id": 1,
|
| 9 |
+
"eos_token_id": 2,
|
| 10 |
+
"head_dim": 96,
|
| 11 |
+
"hidden_act": "silu",
|
| 12 |
+
"hidden_size": 1536,
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 4096,
|
| 15 |
+
"max_position_embeddings": 8192,
|
| 16 |
+
"mlp_bias": false,
|
| 17 |
+
"model_type": "llama",
|
| 18 |
+
"num_attention_heads": 16,
|
| 19 |
+
"num_hidden_layers": 16,
|
| 20 |
+
"num_key_value_heads": 8,
|
| 21 |
+
"pad_token_id": 0,
|
| 22 |
+
"pretraining_tp": 1,
|
| 23 |
+
"rms_norm_eps": 1e-05,
|
| 24 |
+
"rope_scaling": null,
|
| 25 |
+
"rope_theta": 100000,
|
| 26 |
+
"sliding_window": null,
|
| 27 |
+
"tie_word_embeddings": false,
|
| 28 |
+
"torch_dtype": "bfloat16",
|
| 29 |
+
"transformers_version": "4.44.2",
|
| 30 |
+
"use_cache": true,
|
| 31 |
+
"vocab_size": 32000
|
| 32 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.44.2"
|
| 7 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2f543445fee4a911cf2f9559e152c15257f621dc69316a6fded9b4422f7270d
|
| 3 |
+
size 3519556840
|
optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0c9995ff07baed3c0275b29bb5a07431bbc176b099582b7b5056065b277f538
|
| 3 |
+
size 3324071674
|
rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:956c0353592a48f9d490d3289718276a14d3df1265878ef47b3b09958cb9577f
|
| 3 |
+
size 14244
|
scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0f0d9bac4f62dd1b5b2d75cd9a9eec01eca37b01cf938c058f2db3b929bbe13
|
| 3 |
+
size 1064
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<s>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": false,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "</s>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": false,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": {
|
| 17 |
+
"content": "</s>",
|
| 18 |
+
"lstrip": false,
|
| 19 |
+
"normalized": false,
|
| 20 |
+
"rstrip": false,
|
| 21 |
+
"single_word": false
|
| 22 |
+
},
|
| 23 |
+
"unk_token": {
|
| 24 |
+
"content": "<unk>",
|
| 25 |
+
"lstrip": false,
|
| 26 |
+
"normalized": false,
|
| 27 |
+
"rstrip": false,
|
| 28 |
+
"single_word": false
|
| 29 |
+
}
|
| 30 |
+
}
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": false,
|
| 3 |
+
"add_eos_token": false,
|
| 4 |
+
"add_prefix_space": null,
|
| 5 |
+
"added_tokens_decoder": {
|
| 6 |
+
"0": {
|
| 7 |
+
"content": "<unk>",
|
| 8 |
+
"lstrip": false,
|
| 9 |
+
"normalized": false,
|
| 10 |
+
"rstrip": false,
|
| 11 |
+
"single_word": false,
|
| 12 |
+
"special": true
|
| 13 |
+
},
|
| 14 |
+
"1": {
|
| 15 |
+
"content": "<s>",
|
| 16 |
+
"lstrip": false,
|
| 17 |
+
"normalized": false,
|
| 18 |
+
"rstrip": false,
|
| 19 |
+
"single_word": false,
|
| 20 |
+
"special": true
|
| 21 |
+
},
|
| 22 |
+
"2": {
|
| 23 |
+
"content": "</s>",
|
| 24 |
+
"lstrip": false,
|
| 25 |
+
"normalized": false,
|
| 26 |
+
"rstrip": false,
|
| 27 |
+
"single_word": false,
|
| 28 |
+
"special": true
|
| 29 |
+
}
|
| 30 |
+
},
|
| 31 |
+
"additional_special_tokens": [],
|
| 32 |
+
"bos_token": "<s>",
|
| 33 |
+
"clean_up_tokenization_spaces": false,
|
| 34 |
+
"eos_token": "</s>",
|
| 35 |
+
"legacy": true,
|
| 36 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 37 |
+
"pad_token": "</s>",
|
| 38 |
+
"sp_model_kwargs": {},
|
| 39 |
+
"spaces_between_special_tokens": false,
|
| 40 |
+
"tokenizer_class": "LlamaTokenizer",
|
| 41 |
+
"unk_token": "<unk>",
|
| 42 |
+
"use_default_system_prompt": false
|
| 43 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,739 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 10.0,
|
| 5 |
+
"eval_steps": 2000,
|
| 6 |
+
"global_step": 80,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"crossentropy": 2.487246513366699,
|
| 13 |
+
"epoch": 0.125,
|
| 14 |
+
"grad_norm": 0.03208793327212334,
|
| 15 |
+
"learning_rate": 0.001,
|
| 16 |
+
"loss": 2.4872,
|
| 17 |
+
"step": 1
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"crossentropy": 2.379896402359009,
|
| 21 |
+
"epoch": 0.25,
|
| 22 |
+
"grad_norm": 0.03239322826266289,
|
| 23 |
+
"learning_rate": 0.002,
|
| 24 |
+
"loss": 2.3799,
|
| 25 |
+
"step": 2
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"crossentropy": 2.5010085105895996,
|
| 29 |
+
"epoch": 0.375,
|
| 30 |
+
"grad_norm": 0.03320762887597084,
|
| 31 |
+
"learning_rate": 0.003,
|
| 32 |
+
"loss": 2.501,
|
| 33 |
+
"step": 3
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"crossentropy": 2.688127040863037,
|
| 37 |
+
"epoch": 0.5,
|
| 38 |
+
"grad_norm": 0.03224330395460129,
|
| 39 |
+
"learning_rate": 0.004,
|
| 40 |
+
"loss": 2.6881,
|
| 41 |
+
"step": 4
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"crossentropy": 2.4009199142456055,
|
| 45 |
+
"epoch": 0.625,
|
| 46 |
+
"grad_norm": 0.029966533184051514,
|
| 47 |
+
"learning_rate": 0.005,
|
| 48 |
+
"loss": 2.4009,
|
| 49 |
+
"step": 5
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"crossentropy": 2.474385976791382,
|
| 53 |
+
"epoch": 0.75,
|
| 54 |
+
"grad_norm": 0.030283687636256218,
|
| 55 |
+
"learning_rate": 0.006,
|
| 56 |
+
"loss": 2.4744,
|
| 57 |
+
"step": 6
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"crossentropy": 2.5291025638580322,
|
| 61 |
+
"epoch": 0.875,
|
| 62 |
+
"grad_norm": 0.03127186745405197,
|
| 63 |
+
"learning_rate": 0.006999999999999999,
|
| 64 |
+
"loss": 2.5291,
|
| 65 |
+
"step": 7
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"crossentropy": 2.3309171199798584,
|
| 69 |
+
"epoch": 1.0,
|
| 70 |
+
"grad_norm": 0.030661335214972496,
|
| 71 |
+
"learning_rate": 0.008,
|
| 72 |
+
"loss": 2.3309,
|
| 73 |
+
"step": 8
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"crossentropy": 2.4654123783111572,
|
| 77 |
+
"epoch": 1.125,
|
| 78 |
+
"grad_norm": 0.029608242213726044,
|
| 79 |
+
"learning_rate": 0.009000000000000001,
|
| 80 |
+
"loss": 2.4654,
|
| 81 |
+
"step": 9
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"crossentropy": 2.468688488006592,
|
| 85 |
+
"epoch": 1.25,
|
| 86 |
+
"grad_norm": 0.03335599973797798,
|
| 87 |
+
"learning_rate": 0.01,
|
| 88 |
+
"loss": 2.4687,
|
| 89 |
+
"step": 10
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"crossentropy": 2.3773512840270996,
|
| 93 |
+
"epoch": 1.375,
|
| 94 |
+
"grad_norm": 0.031801655888557434,
|
| 95 |
+
"learning_rate": 0.009994965332706574,
|
| 96 |
+
"loss": 2.3774,
|
| 97 |
+
"step": 11
|
| 98 |
+
},
|
| 99 |
+
{
|
| 100 |
+
"crossentropy": 2.5056920051574707,
|
| 101 |
+
"epoch": 1.5,
|
| 102 |
+
"grad_norm": 0.03860826417803764,
|
| 103 |
+
"learning_rate": 0.009979871469976196,
|
| 104 |
+
"loss": 2.5057,
|
| 105 |
+
"step": 12
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"crossentropy": 2.4569733142852783,
|
| 109 |
+
"epoch": 1.625,
|
| 110 |
+
"grad_norm": 0.03568004071712494,
|
| 111 |
+
"learning_rate": 0.009954748808839673,
|
| 112 |
+
"loss": 2.457,
|
| 113 |
+
"step": 13
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"crossentropy": 2.4127414226531982,
|
| 117 |
+
"epoch": 1.75,
|
| 118 |
+
"grad_norm": 0.03557576611638069,
|
| 119 |
+
"learning_rate": 0.009919647942993149,
|
| 120 |
+
"loss": 2.4127,
|
| 121 |
+
"step": 14
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"crossentropy": 2.4089269638061523,
|
| 125 |
+
"epoch": 1.875,
|
| 126 |
+
"grad_norm": 0.0321757011115551,
|
| 127 |
+
"grad_norm_var": 5.793923908080815e-06,
|
| 128 |
+
"learning_rate": 0.009874639560909117,
|
| 129 |
+
"loss": 2.4089,
|
| 130 |
+
"step": 15
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"crossentropy": 2.385178804397583,
|
| 134 |
+
"epoch": 2.0,
|
| 135 |
+
"grad_norm": 0.033555347472429276,
|
| 136 |
+
"grad_norm_var": 5.852987523136916e-06,
|
| 137 |
+
"learning_rate": 0.009819814303479266,
|
| 138 |
+
"loss": 2.3852,
|
| 139 |
+
"step": 16
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"crossentropy": 2.4342517852783203,
|
| 143 |
+
"epoch": 2.125,
|
| 144 |
+
"grad_norm": 0.034557584673166275,
|
| 145 |
+
"grad_norm_var": 6.127065953599033e-06,
|
| 146 |
+
"learning_rate": 0.009755282581475769,
|
| 147 |
+
"loss": 2.4343,
|
| 148 |
+
"step": 17
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"crossentropy": 2.4237754344940186,
|
| 152 |
+
"epoch": 2.25,
|
| 153 |
+
"grad_norm": 0.03371824324131012,
|
| 154 |
+
"grad_norm_var": 6.168768579761655e-06,
|
| 155 |
+
"learning_rate": 0.009681174353198686,
|
| 156 |
+
"loss": 2.4238,
|
| 157 |
+
"step": 18
|
| 158 |
+
},
|
| 159 |
+
{
|
| 160 |
+
"crossentropy": 2.4231510162353516,
|
| 161 |
+
"epoch": 2.375,
|
| 162 |
+
"grad_norm": 0.02907504141330719,
|
| 163 |
+
"grad_norm_var": 6.557002911025779e-06,
|
| 164 |
+
"learning_rate": 0.009597638862757255,
|
| 165 |
+
"loss": 2.4232,
|
| 166 |
+
"step": 19
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"crossentropy": 2.3850157260894775,
|
| 170 |
+
"epoch": 2.5,
|
| 171 |
+
"grad_norm": 0.03278205543756485,
|
| 172 |
+
"grad_norm_var": 6.122522196004565e-06,
|
| 173 |
+
"learning_rate": 0.009504844339512096,
|
| 174 |
+
"loss": 2.385,
|
| 175 |
+
"step": 20
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"crossentropy": 2.3287932872772217,
|
| 179 |
+
"epoch": 2.625,
|
| 180 |
+
"grad_norm": 0.03301858901977539,
|
| 181 |
+
"grad_norm_var": 5.930476905840337e-06,
|
| 182 |
+
"learning_rate": 0.00940297765928369,
|
| 183 |
+
"loss": 2.3288,
|
| 184 |
+
"step": 21
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"crossentropy": 2.3058865070343018,
|
| 188 |
+
"epoch": 2.75,
|
| 189 |
+
"grad_norm": 0.03293571248650551,
|
| 190 |
+
"grad_norm_var": 5.5371734660507875e-06,
|
| 191 |
+
"learning_rate": 0.009292243968009331,
|
| 192 |
+
"loss": 2.3059,
|
| 193 |
+
"step": 22
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"crossentropy": 2.2683637142181396,
|
| 197 |
+
"epoch": 2.875,
|
| 198 |
+
"grad_norm": 0.03541896864771843,
|
| 199 |
+
"grad_norm_var": 5.362674881745298e-06,
|
| 200 |
+
"learning_rate": 0.009172866268606514,
|
| 201 |
+
"loss": 2.2684,
|
| 202 |
+
"step": 23
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"crossentropy": 2.411990165710449,
|
| 206 |
+
"epoch": 3.0,
|
| 207 |
+
"grad_norm": 0.03502194955945015,
|
| 208 |
+
"grad_norm_var": 4.411311488799798e-06,
|
| 209 |
+
"learning_rate": 0.009045084971874737,
|
| 210 |
+
"loss": 2.412,
|
| 211 |
+
"step": 24
|
| 212 |
+
},
|
| 213 |
+
{
|
| 214 |
+
"crossentropy": 2.2515363693237305,
|
| 215 |
+
"epoch": 3.125,
|
| 216 |
+
"grad_norm": 0.03379856050014496,
|
| 217 |
+
"grad_norm_var": 4.1744788750110175e-06,
|
| 218 |
+
"learning_rate": 0.008909157412340149,
|
| 219 |
+
"loss": 2.2515,
|
| 220 |
+
"step": 25
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"crossentropy": 2.3486626148223877,
|
| 224 |
+
"epoch": 3.25,
|
| 225 |
+
"grad_norm": 0.034475792199373245,
|
| 226 |
+
"grad_norm_var": 2.7198637048790144e-06,
|
| 227 |
+
"learning_rate": 0.008765357330018056,
|
| 228 |
+
"loss": 2.3487,
|
| 229 |
+
"step": 26
|
| 230 |
+
},
|
| 231 |
+
{
|
| 232 |
+
"crossentropy": 2.379669666290283,
|
| 233 |
+
"epoch": 3.375,
|
| 234 |
+
"grad_norm": 0.03319563344120979,
|
| 235 |
+
"grad_norm_var": 2.4738877727155267e-06,
|
| 236 |
+
"learning_rate": 0.008613974319136958,
|
| 237 |
+
"loss": 2.3797,
|
| 238 |
+
"step": 27
|
| 239 |
+
},
|
| 240 |
+
{
|
| 241 |
+
"crossentropy": 2.3640384674072266,
|
| 242 |
+
"epoch": 3.5,
|
| 243 |
+
"grad_norm": 0.03426358476281166,
|
| 244 |
+
"grad_norm_var": 2.2389126654258237e-06,
|
| 245 |
+
"learning_rate": 0.008455313244934324,
|
| 246 |
+
"loss": 2.364,
|
| 247 |
+
"step": 28
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"crossentropy": 2.324143409729004,
|
| 251 |
+
"epoch": 3.625,
|
| 252 |
+
"grad_norm": 0.0358288437128067,
|
| 253 |
+
"grad_norm_var": 2.410602035355508e-06,
|
| 254 |
+
"learning_rate": 0.008289693629698563,
|
| 255 |
+
"loss": 2.3241,
|
| 256 |
+
"step": 29
|
| 257 |
+
},
|
| 258 |
+
{
|
| 259 |
+
"crossentropy": 2.159423828125,
|
| 260 |
+
"epoch": 3.75,
|
| 261 |
+
"grad_norm": 0.031966786831617355,
|
| 262 |
+
"grad_norm_var": 2.612506091834166e-06,
|
| 263 |
+
"learning_rate": 0.008117449009293669,
|
| 264 |
+
"loss": 2.1594,
|
| 265 |
+
"step": 30
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"crossentropy": 2.2669661045074463,
|
| 269 |
+
"epoch": 3.875,
|
| 270 |
+
"grad_norm": 0.0339609794318676,
|
| 271 |
+
"grad_norm_var": 2.61687730451927e-06,
|
| 272 |
+
"learning_rate": 0.007938926261462366,
|
| 273 |
+
"loss": 2.267,
|
| 274 |
+
"step": 31
|
| 275 |
+
},
|
| 276 |
+
{
|
| 277 |
+
"crossentropy": 2.482945680618286,
|
| 278 |
+
"epoch": 4.0,
|
| 279 |
+
"grad_norm": 0.03357086703181267,
|
| 280 |
+
"grad_norm_var": 2.5635888162410063e-06,
|
| 281 |
+
"learning_rate": 0.007754484907260513,
|
| 282 |
+
"loss": 2.4829,
|
| 283 |
+
"step": 32
|
| 284 |
+
},
|
| 285 |
+
{
|
| 286 |
+
"crossentropy": 2.1757168769836426,
|
| 287 |
+
"epoch": 4.125,
|
| 288 |
+
"grad_norm": 0.032582979649305344,
|
| 289 |
+
"grad_norm_var": 1.2068945133852464e-06,
|
| 290 |
+
"learning_rate": 0.007564496387029531,
|
| 291 |
+
"loss": 2.1757,
|
| 292 |
+
"step": 33
|
| 293 |
+
},
|
| 294 |
+
{
|
| 295 |
+
"crossentropy": 2.4300379753112793,
|
| 296 |
+
"epoch": 4.25,
|
| 297 |
+
"grad_norm": 0.03353552147746086,
|
| 298 |
+
"grad_norm_var": 1.1362555840309259e-06,
|
| 299 |
+
"learning_rate": 0.007369343312364994,
|
| 300 |
+
"loss": 2.43,
|
| 301 |
+
"step": 34
|
| 302 |
+
},
|
| 303 |
+
{
|
| 304 |
+
"crossentropy": 2.1233766078948975,
|
| 305 |
+
"epoch": 4.375,
|
| 306 |
+
"grad_norm": 0.034830041229724884,
|
| 307 |
+
"grad_norm_var": 1.1319644129707703e-06,
|
| 308 |
+
"learning_rate": 0.007169418695587791,
|
| 309 |
+
"loss": 2.1234,
|
| 310 |
+
"step": 35
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"crossentropy": 2.2619526386260986,
|
| 314 |
+
"epoch": 4.5,
|
| 315 |
+
"grad_norm": 0.033126723021268845,
|
| 316 |
+
"grad_norm_var": 1.1071727437842251e-06,
|
| 317 |
+
"learning_rate": 0.006965125158269619,
|
| 318 |
+
"loss": 2.262,
|
| 319 |
+
"step": 36
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"crossentropy": 2.3065433502197266,
|
| 323 |
+
"epoch": 4.625,
|
| 324 |
+
"grad_norm": 0.0348593033850193,
|
| 325 |
+
"grad_norm_var": 1.0216560744425243e-06,
|
| 326 |
+
"learning_rate": 0.0067568741204067145,
|
| 327 |
+
"loss": 2.3065,
|
| 328 |
+
"step": 37
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"crossentropy": 2.2902674674987793,
|
| 332 |
+
"epoch": 4.75,
|
| 333 |
+
"grad_norm": 0.0360955074429512,
|
| 334 |
+
"grad_norm_var": 1.2434575549111667e-06,
|
| 335 |
+
"learning_rate": 0.006545084971874737,
|
| 336 |
+
"loss": 2.2903,
|
| 337 |
+
"step": 38
|
| 338 |
+
},
|
| 339 |
+
{
|
| 340 |
+
"crossentropy": 2.2587778568267822,
|
| 341 |
+
"epoch": 4.875,
|
| 342 |
+
"grad_norm": 0.03467360511422157,
|
| 343 |
+
"grad_norm_var": 1.2055615432713293e-06,
|
| 344 |
+
"learning_rate": 0.006330184227833375,
|
| 345 |
+
"loss": 2.2588,
|
| 346 |
+
"step": 39
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"crossentropy": 2.3978285789489746,
|
| 350 |
+
"epoch": 5.0,
|
| 351 |
+
"grad_norm": 0.035668086260557175,
|
| 352 |
+
"grad_norm_var": 1.3685657271088221e-06,
|
| 353 |
+
"learning_rate": 0.006112604669781572,
|
| 354 |
+
"loss": 2.3978,
|
| 355 |
+
"step": 40
|
| 356 |
+
},
|
| 357 |
+
{
|
| 358 |
+
"crossentropy": 2.3564870357513428,
|
| 359 |
+
"epoch": 5.125,
|
| 360 |
+
"grad_norm": 0.0373300276696682,
|
| 361 |
+
"grad_norm_var": 2.0190065310690096e-06,
|
| 362 |
+
"learning_rate": 0.005892784473993183,
|
| 363 |
+
"loss": 2.3565,
|
| 364 |
+
"step": 41
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"crossentropy": 2.202275514602661,
|
| 368 |
+
"epoch": 5.25,
|
| 369 |
+
"grad_norm": 0.03768543526530266,
|
| 370 |
+
"grad_norm_var": 2.6563097811015932e-06,
|
| 371 |
+
"learning_rate": 0.0056711663290882775,
|
| 372 |
+
"loss": 2.2023,
|
| 373 |
+
"step": 42
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"crossentropy": 2.192146062850952,
|
| 377 |
+
"epoch": 5.375,
|
| 378 |
+
"grad_norm": 0.03134535253047943,
|
| 379 |
+
"grad_norm_var": 3.228640330968311e-06,
|
| 380 |
+
"learning_rate": 0.005448196544517168,
|
| 381 |
+
"loss": 2.1921,
|
| 382 |
+
"step": 43
|
| 383 |
+
},
|
| 384 |
+
{
|
| 385 |
+
"crossentropy": 2.2519092559814453,
|
| 386 |
+
"epoch": 5.5,
|
| 387 |
+
"grad_norm": 0.032612044364213943,
|
| 388 |
+
"grad_norm_var": 3.044945465432008e-06,
|
| 389 |
+
"learning_rate": 0.005224324151752576,
|
| 390 |
+
"loss": 2.2519,
|
| 391 |
+
"step": 44
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"crossentropy": 2.3610141277313232,
|
| 395 |
+
"epoch": 5.625,
|
| 396 |
+
"grad_norm": 0.03532750904560089,
|
| 397 |
+
"grad_norm_var": 3.073519780112897e-06,
|
| 398 |
+
"learning_rate": 0.005,
|
| 399 |
+
"loss": 2.361,
|
| 400 |
+
"step": 45
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"crossentropy": 2.2574493885040283,
|
| 404 |
+
"epoch": 5.75,
|
| 405 |
+
"grad_norm": 0.03538508713245392,
|
| 406 |
+
"grad_norm_var": 3.047191916038562e-06,
|
| 407 |
+
"learning_rate": 0.004775675848247427,
|
| 408 |
+
"loss": 2.2574,
|
| 409 |
+
"step": 46
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"crossentropy": 2.170254707336426,
|
| 413 |
+
"epoch": 5.875,
|
| 414 |
+
"grad_norm": 0.03475534915924072,
|
| 415 |
+
"grad_norm_var": 2.9654755954788835e-06,
|
| 416 |
+
"learning_rate": 0.004551803455482833,
|
| 417 |
+
"loss": 2.1703,
|
| 418 |
+
"step": 47
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"crossentropy": 2.2004218101501465,
|
| 422 |
+
"epoch": 6.0,
|
| 423 |
+
"grad_norm": 0.0375908762216568,
|
| 424 |
+
"grad_norm_var": 3.107626395358089e-06,
|
| 425 |
+
"learning_rate": 0.004328833670911724,
|
| 426 |
+
"loss": 2.2004,
|
| 427 |
+
"step": 48
|
| 428 |
+
},
|
| 429 |
+
{
|
| 430 |
+
"crossentropy": 2.162071943283081,
|
| 431 |
+
"epoch": 6.125,
|
| 432 |
+
"grad_norm": 0.031826432794332504,
|
| 433 |
+
"grad_norm_var": 4.072774386597716e-06,
|
| 434 |
+
"learning_rate": 0.004107215526006817,
|
| 435 |
+
"loss": 2.1621,
|
| 436 |
+
"step": 49
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"crossentropy": 2.144589424133301,
|
| 440 |
+
"epoch": 6.25,
|
| 441 |
+
"grad_norm": 0.03268062323331833,
|
| 442 |
+
"grad_norm_var": 4.202360731680779e-06,
|
| 443 |
+
"learning_rate": 0.003887395330218428,
|
| 444 |
+
"loss": 2.1446,
|
| 445 |
+
"step": 50
|
| 446 |
+
},
|
| 447 |
+
{
|
| 448 |
+
"crossentropy": 2.202075481414795,
|
| 449 |
+
"epoch": 6.375,
|
| 450 |
+
"grad_norm": 0.03384561836719513,
|
| 451 |
+
"grad_norm_var": 4.294842472950198e-06,
|
| 452 |
+
"learning_rate": 0.003669815772166625,
|
| 453 |
+
"loss": 2.2021,
|
| 454 |
+
"step": 51
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"crossentropy": 2.3169946670532227,
|
| 458 |
+
"epoch": 6.5,
|
| 459 |
+
"grad_norm": 0.03661832585930824,
|
| 460 |
+
"grad_norm_var": 4.387942230220839e-06,
|
| 461 |
+
"learning_rate": 0.003454915028125263,
|
| 462 |
+
"loss": 2.317,
|
| 463 |
+
"step": 52
|
| 464 |
+
},
|
| 465 |
+
{
|
| 466 |
+
"crossentropy": 2.191814422607422,
|
| 467 |
+
"epoch": 6.625,
|
| 468 |
+
"grad_norm": 0.03522748872637749,
|
| 469 |
+
"grad_norm_var": 4.380226970383122e-06,
|
| 470 |
+
"learning_rate": 0.003243125879593286,
|
| 471 |
+
"loss": 2.1918,
|
| 472 |
+
"step": 53
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"crossentropy": 2.196298360824585,
|
| 476 |
+
"epoch": 6.75,
|
| 477 |
+
"grad_norm": 0.03524046018719673,
|
| 478 |
+
"grad_norm_var": 4.3576876356847075e-06,
|
| 479 |
+
"learning_rate": 0.0030348748417303823,
|
| 480 |
+
"loss": 2.1963,
|
| 481 |
+
"step": 54
|
| 482 |
+
},
|
| 483 |
+
{
|
| 484 |
+
"crossentropy": 2.2684872150421143,
|
| 485 |
+
"epoch": 6.875,
|
| 486 |
+
"grad_norm": 0.03640512377023697,
|
| 487 |
+
"grad_norm_var": 4.452811959714381e-06,
|
| 488 |
+
"learning_rate": 0.00283058130441221,
|
| 489 |
+
"loss": 2.2685,
|
| 490 |
+
"step": 55
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"crossentropy": 2.3180689811706543,
|
| 494 |
+
"epoch": 7.0,
|
| 495 |
+
"grad_norm": 0.03479791432619095,
|
| 496 |
+
"grad_norm_var": 4.097831569832335e-06,
|
| 497 |
+
"learning_rate": 0.002630656687635007,
|
| 498 |
+
"loss": 2.3181,
|
| 499 |
+
"step": 56
|
| 500 |
+
},
|
| 501 |
+
{
|
| 502 |
+
"crossentropy": 2.201280117034912,
|
| 503 |
+
"epoch": 7.125,
|
| 504 |
+
"grad_norm": 0.03417252376675606,
|
| 505 |
+
"grad_norm_var": 2.7744502044233416e-06,
|
| 506 |
+
"learning_rate": 0.00243550361297047,
|
| 507 |
+
"loss": 2.2013,
|
| 508 |
+
"step": 57
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"crossentropy": 2.3339715003967285,
|
| 512 |
+
"epoch": 7.25,
|
| 513 |
+
"grad_norm": 0.033313509076833725,
|
| 514 |
+
"grad_norm_var": 2.5884423837709467e-06,
|
| 515 |
+
"learning_rate": 0.002245515092739488,
|
| 516 |
+
"loss": 2.334,
|
| 517 |
+
"step": 58
|
| 518 |
+
},
|
| 519 |
+
{
|
| 520 |
+
"crossentropy": 2.1603586673736572,
|
| 521 |
+
"epoch": 7.375,
|
| 522 |
+
"grad_norm": 0.03365428000688553,
|
| 523 |
+
"grad_norm_var": 2.684439790351147e-06,
|
| 524 |
+
"learning_rate": 0.0020610737385376348,
|
| 525 |
+
"loss": 2.1604,
|
| 526 |
+
"step": 59
|
| 527 |
+
},
|
| 528 |
+
{
|
| 529 |
+
"crossentropy": 2.2999627590179443,
|
| 530 |
+
"epoch": 7.5,
|
| 531 |
+
"grad_norm": 0.034641556441783905,
|
| 532 |
+
"grad_norm_var": 2.6678187146791006e-06,
|
| 533 |
+
"learning_rate": 0.0018825509907063327,
|
| 534 |
+
"loss": 2.3,
|
| 535 |
+
"step": 60
|
| 536 |
+
},
|
| 537 |
+
{
|
| 538 |
+
"crossentropy": 2.1747264862060547,
|
| 539 |
+
"epoch": 7.625,
|
| 540 |
+
"grad_norm": 0.03396380692720413,
|
| 541 |
+
"grad_norm_var": 2.7140570016346197e-06,
|
| 542 |
+
"learning_rate": 0.001710306370301437,
|
| 543 |
+
"loss": 2.1747,
|
| 544 |
+
"step": 61
|
| 545 |
+
},
|
| 546 |
+
{
|
| 547 |
+
"crossentropy": 2.120774269104004,
|
| 548 |
+
"epoch": 7.75,
|
| 549 |
+
"grad_norm": 0.03624221310019493,
|
| 550 |
+
"grad_norm_var": 2.3210148057607623e-06,
|
| 551 |
+
"learning_rate": 0.0015446867550656768,
|
| 552 |
+
"loss": 2.1208,
|
| 553 |
+
"step": 62
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"crossentropy": 2.2385036945343018,
|
| 557 |
+
"epoch": 7.875,
|
| 558 |
+
"grad_norm": 0.034913428127765656,
|
| 559 |
+
"grad_norm_var": 1.732991609125894e-06,
|
| 560 |
+
"learning_rate": 0.0013860256808630427,
|
| 561 |
+
"loss": 2.2385,
|
| 562 |
+
"step": 63
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"crossentropy": 2.1571521759033203,
|
| 566 |
+
"epoch": 8.0,
|
| 567 |
+
"grad_norm": 0.03277244418859482,
|
| 568 |
+
"grad_norm_var": 1.4490052253732697e-06,
|
| 569 |
+
"learning_rate": 0.0012346426699819458,
|
| 570 |
+
"loss": 2.1572,
|
| 571 |
+
"step": 64
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"crossentropy": 2.2847859859466553,
|
| 575 |
+
"epoch": 8.125,
|
| 576 |
+
"grad_norm": 0.033221788704395294,
|
| 577 |
+
"grad_norm_var": 1.512194472239048e-06,
|
| 578 |
+
"learning_rate": 0.001090842587659851,
|
| 579 |
+
"loss": 2.2848,
|
| 580 |
+
"step": 65
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"crossentropy": 2.253898859024048,
|
| 584 |
+
"epoch": 8.25,
|
| 585 |
+
"grad_norm": 0.03283696994185448,
|
| 586 |
+
"grad_norm_var": 1.361639072858939e-06,
|
| 587 |
+
"learning_rate": 0.0009549150281252633,
|
| 588 |
+
"loss": 2.2539,
|
| 589 |
+
"step": 66
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"crossentropy": 2.1402039527893066,
|
| 593 |
+
"epoch": 8.375,
|
| 594 |
+
"grad_norm": 0.03309963271021843,
|
| 595 |
+
"grad_norm_var": 1.3845661239702636e-06,
|
| 596 |
+
"learning_rate": 0.0008271337313934868,
|
| 597 |
+
"loss": 2.1402,
|
| 598 |
+
"step": 67
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"crossentropy": 2.1381993293762207,
|
| 602 |
+
"epoch": 8.5,
|
| 603 |
+
"grad_norm": 0.03431132435798645,
|
| 604 |
+
"grad_norm_var": 1.3068839073976725e-06,
|
| 605 |
+
"learning_rate": 0.0007077560319906695,
|
| 606 |
+
"loss": 2.1382,
|
| 607 |
+
"step": 68
|
| 608 |
+
},
|
| 609 |
+
{
|
| 610 |
+
"crossentropy": 2.1610183715820312,
|
| 611 |
+
"epoch": 8.625,
|
| 612 |
+
"grad_norm": 0.03382508084177971,
|
| 613 |
+
"grad_norm_var": 9.367598844583584e-07,
|
| 614 |
+
"learning_rate": 0.00059702234071631,
|
| 615 |
+
"loss": 2.161,
|
| 616 |
+
"step": 69
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"crossentropy": 2.2194838523864746,
|
| 620 |
+
"epoch": 8.75,
|
| 621 |
+
"grad_norm": 0.03652471676468849,
|
| 622 |
+
"grad_norm_var": 1.3163803696068672e-06,
|
| 623 |
+
"learning_rate": 0.0004951556604879049,
|
| 624 |
+
"loss": 2.2195,
|
| 625 |
+
"step": 70
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"crossentropy": 2.179126262664795,
|
| 629 |
+
"epoch": 8.875,
|
| 630 |
+
"grad_norm": 0.03386425971984863,
|
| 631 |
+
"grad_norm_var": 1.2798076699575718e-06,
|
| 632 |
+
"learning_rate": 0.0004023611372427471,
|
| 633 |
+
"loss": 2.1791,
|
| 634 |
+
"step": 71
|
| 635 |
+
},
|
| 636 |
+
{
|
| 637 |
+
"crossentropy": 2.2535359859466553,
|
| 638 |
+
"epoch": 9.0,
|
| 639 |
+
"grad_norm": 0.032655857503414154,
|
| 640 |
+
"grad_norm_var": 1.3903296123838708e-06,
|
| 641 |
+
"learning_rate": 0.00031882564680131396,
|
| 642 |
+
"loss": 2.2535,
|
| 643 |
+
"step": 72
|
| 644 |
+
},
|
| 645 |
+
{
|
| 646 |
+
"crossentropy": 2.083571434020996,
|
| 647 |
+
"epoch": 9.125,
|
| 648 |
+
"grad_norm": 0.03222977742552757,
|
| 649 |
+
"grad_norm_var": 1.6381791075122573e-06,
|
| 650 |
+
"learning_rate": 0.00024471741852423234,
|
| 651 |
+
"loss": 2.0836,
|
| 652 |
+
"step": 73
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"crossentropy": 2.1447954177856445,
|
| 656 |
+
"epoch": 9.25,
|
| 657 |
+
"grad_norm": 0.03370295464992523,
|
| 658 |
+
"grad_norm_var": 1.585818962655111e-06,
|
| 659 |
+
"learning_rate": 0.0001801856965207338,
|
| 660 |
+
"loss": 2.1448,
|
| 661 |
+
"step": 74
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"crossentropy": 2.3297011852264404,
|
| 665 |
+
"epoch": 9.375,
|
| 666 |
+
"grad_norm": 0.03342713788151741,
|
| 667 |
+
"grad_norm_var": 1.5866984901570397e-06,
|
| 668 |
+
"learning_rate": 0.0001253604390908819,
|
| 669 |
+
"loss": 2.3297,
|
| 670 |
+
"step": 75
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"crossentropy": 2.184936761856079,
|
| 674 |
+
"epoch": 9.5,
|
| 675 |
+
"grad_norm": 0.03379293903708458,
|
| 676 |
+
"grad_norm_var": 1.128480817529492e-06,
|
| 677 |
+
"learning_rate": 8.035205700685166e-05,
|
| 678 |
+
"loss": 2.1849,
|
| 679 |
+
"step": 76
|
| 680 |
+
},
|
| 681 |
+
{
|
| 682 |
+
"crossentropy": 2.244959592819214,
|
| 683 |
+
"epoch": 9.625,
|
| 684 |
+
"grad_norm": 0.03535815700888634,
|
| 685 |
+
"grad_norm_var": 1.2224064569112125e-06,
|
| 686 |
+
"learning_rate": 4.52511911603265e-05,
|
| 687 |
+
"loss": 2.245,
|
| 688 |
+
"step": 77
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"crossentropy": 2.145529270172119,
|
| 692 |
+
"epoch": 9.75,
|
| 693 |
+
"grad_norm": 0.03338780626654625,
|
| 694 |
+
"grad_norm_var": 1.180987359381859e-06,
|
| 695 |
+
"learning_rate": 2.012853002380466e-05,
|
| 696 |
+
"loss": 2.1455,
|
| 697 |
+
"step": 78
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"crossentropy": 2.282687187194824,
|
| 701 |
+
"epoch": 9.875,
|
| 702 |
+
"grad_norm": 0.03313542157411575,
|
| 703 |
+
"grad_norm_var": 1.1489689212500093e-06,
|
| 704 |
+
"learning_rate": 5.034667293427053e-06,
|
| 705 |
+
"loss": 2.2827,
|
| 706 |
+
"step": 79
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"crossentropy": 2.1934750080108643,
|
| 710 |
+
"epoch": 10.0,
|
| 711 |
+
"grad_norm": 0.03277941048145294,
|
| 712 |
+
"grad_norm_var": 1.18509241838338e-06,
|
| 713 |
+
"learning_rate": 0.0,
|
| 714 |
+
"loss": 2.1935,
|
| 715 |
+
"step": 80
|
| 716 |
+
}
|
| 717 |
+
],
|
| 718 |
+
"logging_steps": 1,
|
| 719 |
+
"max_steps": 80,
|
| 720 |
+
"num_input_tokens_seen": 0,
|
| 721 |
+
"num_train_epochs": 10,
|
| 722 |
+
"save_steps": 2000,
|
| 723 |
+
"stateful_callbacks": {
|
| 724 |
+
"TrainerControl": {
|
| 725 |
+
"args": {
|
| 726 |
+
"should_epoch_stop": false,
|
| 727 |
+
"should_evaluate": false,
|
| 728 |
+
"should_log": false,
|
| 729 |
+
"should_save": true,
|
| 730 |
+
"should_training_stop": true
|
| 731 |
+
},
|
| 732 |
+
"attributes": {}
|
| 733 |
+
}
|
| 734 |
+
},
|
| 735 |
+
"total_flos": 1.384645494177792e+16,
|
| 736 |
+
"train_batch_size": 16,
|
| 737 |
+
"trial_name": null,
|
| 738 |
+
"trial_params": null
|
| 739 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36525e45296e89abb811d54f8d48c25f84509486e0812ee034d7029aea3952d0
|
| 3 |
+
size 5304
|