Upload folder using huggingface_hub
Browse files- configuration_kraken.py +8 -0
- kraken_model/config.json +36 -0
- kraken_model/generation_config.json +4 -0
- kraken_model/model.safetensors +3 -0
- kraken_router/added_tokens.json +5 -0
- kraken_router/config.json +42 -0
- kraken_router/merges.txt +0 -0
- kraken_router/model.safetensors +3 -0
- kraken_router/optimizer.pt +3 -0
- kraken_router/rng_state.pth +3 -0
- kraken_router/scheduler.pt +3 -0
- kraken_router/special_tokens_map.json +14 -0
- kraken_router/tokenizer.json +0 -0
- kraken_router/tokenizer_config.json +43 -0
- kraken_router/trainer_state.json +747 -0
- kraken_router/training_args.bin +3 -0
- kraken_router/vocab.json +0 -0
- modeling_kraken.py +82 -0
- tokenizer_template_switch.py +97 -0
configuration_kraken.py
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import PretrainedConfig
|
2 |
+
|
3 |
+
class KrakenConfig(PretrainedConfig):
|
4 |
+
model_type = "kraken"
|
5 |
+
|
6 |
+
def __init__(self, config_dict=None, **kwargs):
|
7 |
+
super().__init__(**kwargs)
|
8 |
+
self.config_dict = config_dict or {}
|
kraken_model/config.json
ADDED
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"KrakenForCausalLM"
|
4 |
+
],
|
5 |
+
"config_dict": {
|
6 |
+
"class_indices": {
|
7 |
+
"LABEL_0": 0,
|
8 |
+
"LABEL_1": 1,
|
9 |
+
"LABEL_2": 2,
|
10 |
+
"LABEL_3": 3
|
11 |
+
},
|
12 |
+
"model_type": "kraken",
|
13 |
+
"models": {
|
14 |
+
"expert1": "VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct",
|
15 |
+
"expert2": "mii-community/zefiro-7b-dpo-ITA",
|
16 |
+
"expert3": "paulml/Hermes-2-Pro-French",
|
17 |
+
"expert4": "norallm/normistral-7b-warm-instruct"
|
18 |
+
},
|
19 |
+
"quantization": {
|
20 |
+
"expert1": null,
|
21 |
+
"expert2": null,
|
22 |
+
"expert3": null,
|
23 |
+
"expert4": null
|
24 |
+
},
|
25 |
+
"router": "kraken_router",
|
26 |
+
"tokenizers": {
|
27 |
+
"expert1": "VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct",
|
28 |
+
"expert2": "mii-community/zefiro-7b-dpo-ITA",
|
29 |
+
"expert3": "paulml/Hermes-2-Pro-French",
|
30 |
+
"expert4": "norallm/normistral-7b-warm-instruct"
|
31 |
+
}
|
32 |
+
},
|
33 |
+
"model_type": "kraken",
|
34 |
+
"torch_dtype": "float32",
|
35 |
+
"transformers_version": "4.41.0"
|
36 |
+
}
|
kraken_model/generation_config.json
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"transformers_version": "4.41.0"
|
4 |
+
}
|
kraken_model/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7904498ee4fe7684ea8b1e5e4b898d68bf5874b8b99a6b3a29d217f2f762cb0d
|
3 |
+
size 1856003896
|
kraken_router/added_tokens.json
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"<|endoftext|>": 151643,
|
3 |
+
"<|im_end|>": 151645,
|
4 |
+
"<|im_start|>": 151644
|
5 |
+
}
|
kraken_router/config.json
ADDED
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "Qwen/Qwen1.5-0.5B",
|
3 |
+
"architectures": [
|
4 |
+
"Qwen2ForSequenceClassification"
|
5 |
+
],
|
6 |
+
"attention_dropout": 0.0,
|
7 |
+
"bos_token_id": 151643,
|
8 |
+
"eos_token_id": 151643,
|
9 |
+
"hidden_act": "silu",
|
10 |
+
"hidden_size": 1024,
|
11 |
+
"id2label": {
|
12 |
+
"0": "LABEL_0",
|
13 |
+
"1": "LABEL_1",
|
14 |
+
"2": "LABEL_2",
|
15 |
+
"3": "LABEL_3"
|
16 |
+
},
|
17 |
+
"initializer_range": 0.02,
|
18 |
+
"intermediate_size": 2816,
|
19 |
+
"label2id": {
|
20 |
+
"LABEL_0": 0,
|
21 |
+
"LABEL_1": 1,
|
22 |
+
"LABEL_2": 2,
|
23 |
+
"LABEL_3": 3
|
24 |
+
},
|
25 |
+
"max_position_embeddings": 32768,
|
26 |
+
"max_window_layers": 21,
|
27 |
+
"model_type": "qwen2",
|
28 |
+
"num_attention_heads": 16,
|
29 |
+
"num_hidden_layers": 24,
|
30 |
+
"num_key_value_heads": 16,
|
31 |
+
"pad_token_id": 151643,
|
32 |
+
"problem_type": "single_label_classification",
|
33 |
+
"rms_norm_eps": 1e-06,
|
34 |
+
"rope_theta": 1000000.0,
|
35 |
+
"sliding_window": 32768,
|
36 |
+
"tie_word_embeddings": true,
|
37 |
+
"torch_dtype": "float32",
|
38 |
+
"transformers_version": "4.41.0",
|
39 |
+
"use_cache": true,
|
40 |
+
"use_sliding_window": false,
|
41 |
+
"vocab_size": 151936
|
42 |
+
}
|
kraken_router/merges.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
kraken_router/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a977f3e5f94dc95b07300f21e38c86fc85906081fde22526de296c32c0e686e4
|
3 |
+
size 1856000112
|
kraken_router/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:52c203605e5f0e424fd7f097eacdd937670347c777796d884eba1eaa2cc2a2f7
|
3 |
+
size 3712178682
|
kraken_router/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c062f7f375beded48b5337f5a3f3a5cb38807fa3e85dbf3e294c0ab6b627bfc2
|
3 |
+
size 14244
|
kraken_router/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:460f54983d240fe33e7a94fb335adfd576cb487f5d497e2b899d4e773a2ce84c
|
3 |
+
size 1064
|
kraken_router/special_tokens_map.json
ADDED
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"additional_special_tokens": [
|
3 |
+
"<|im_start|>",
|
4 |
+
"<|im_end|>"
|
5 |
+
],
|
6 |
+
"eos_token": {
|
7 |
+
"content": "<|endoftext|>",
|
8 |
+
"lstrip": false,
|
9 |
+
"normalized": false,
|
10 |
+
"rstrip": false,
|
11 |
+
"single_word": false
|
12 |
+
},
|
13 |
+
"pad_token": "<|endoftext|>"
|
14 |
+
}
|
kraken_router/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
kraken_router/tokenizer_config.json
ADDED
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_prefix_space": false,
|
3 |
+
"added_tokens_decoder": {
|
4 |
+
"151643": {
|
5 |
+
"content": "<|endoftext|>",
|
6 |
+
"lstrip": false,
|
7 |
+
"normalized": false,
|
8 |
+
"rstrip": false,
|
9 |
+
"single_word": false,
|
10 |
+
"special": true
|
11 |
+
},
|
12 |
+
"151644": {
|
13 |
+
"content": "<|im_start|>",
|
14 |
+
"lstrip": false,
|
15 |
+
"normalized": false,
|
16 |
+
"rstrip": false,
|
17 |
+
"single_word": false,
|
18 |
+
"special": true
|
19 |
+
},
|
20 |
+
"151645": {
|
21 |
+
"content": "<|im_end|>",
|
22 |
+
"lstrip": false,
|
23 |
+
"normalized": false,
|
24 |
+
"rstrip": false,
|
25 |
+
"single_word": false,
|
26 |
+
"special": true
|
27 |
+
}
|
28 |
+
},
|
29 |
+
"additional_special_tokens": [
|
30 |
+
"<|im_start|>",
|
31 |
+
"<|im_end|>"
|
32 |
+
],
|
33 |
+
"bos_token": null,
|
34 |
+
"chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
|
35 |
+
"clean_up_tokenization_spaces": false,
|
36 |
+
"eos_token": "<|endoftext|>",
|
37 |
+
"errors": "replace",
|
38 |
+
"model_max_length": 32768,
|
39 |
+
"pad_token": "<|endoftext|>",
|
40 |
+
"split_special_tokens": false,
|
41 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
42 |
+
"unk_token": null
|
43 |
+
}
|
kraken_router/trainer_state.json
ADDED
@@ -0,0 +1,747 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 5.283331606754377,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 51000,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.05179736869367036,
|
13 |
+
"grad_norm": 3.8771300836515366e-08,
|
14 |
+
"learning_rate": 1.9852007518018084e-05,
|
15 |
+
"loss": 0.0996,
|
16 |
+
"step": 500
|
17 |
+
},
|
18 |
+
{
|
19 |
+
"epoch": 0.10359473738734072,
|
20 |
+
"grad_norm": 9.604158321963041e-07,
|
21 |
+
"learning_rate": 1.970401503603617e-05,
|
22 |
+
"loss": 0.256,
|
23 |
+
"step": 1000
|
24 |
+
},
|
25 |
+
{
|
26 |
+
"epoch": 0.1553921060810111,
|
27 |
+
"grad_norm": 1.8179751350544393e-05,
|
28 |
+
"learning_rate": 1.9556022554054253e-05,
|
29 |
+
"loss": 0.1367,
|
30 |
+
"step": 1500
|
31 |
+
},
|
32 |
+
{
|
33 |
+
"epoch": 0.20718947477468144,
|
34 |
+
"grad_norm": 8.53914680192247e-05,
|
35 |
+
"learning_rate": 1.9408030072072343e-05,
|
36 |
+
"loss": 0.063,
|
37 |
+
"step": 2000
|
38 |
+
},
|
39 |
+
{
|
40 |
+
"epoch": 0.2589868434683518,
|
41 |
+
"grad_norm": 4.6534645662177354e-05,
|
42 |
+
"learning_rate": 1.9260037590090425e-05,
|
43 |
+
"loss": 0.0201,
|
44 |
+
"step": 2500
|
45 |
+
},
|
46 |
+
{
|
47 |
+
"epoch": 0.3107842121620222,
|
48 |
+
"grad_norm": 6.902104843220513e-08,
|
49 |
+
"learning_rate": 1.911204510810851e-05,
|
50 |
+
"loss": 0.106,
|
51 |
+
"step": 3000
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"epoch": 0.36258158085569253,
|
55 |
+
"grad_norm": 2.192794745781157e-08,
|
56 |
+
"learning_rate": 1.8964052626126594e-05,
|
57 |
+
"loss": 0.0797,
|
58 |
+
"step": 3500
|
59 |
+
},
|
60 |
+
{
|
61 |
+
"epoch": 0.4143789495493629,
|
62 |
+
"grad_norm": 4.58152558859698e-13,
|
63 |
+
"learning_rate": 1.881606014414468e-05,
|
64 |
+
"loss": 0.0264,
|
65 |
+
"step": 4000
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"epoch": 0.46617631824303324,
|
69 |
+
"grad_norm": 4.8503436119062826e-05,
|
70 |
+
"learning_rate": 1.8668067662162763e-05,
|
71 |
+
"loss": 0.0329,
|
72 |
+
"step": 4500
|
73 |
+
},
|
74 |
+
{
|
75 |
+
"epoch": 0.5179736869367036,
|
76 |
+
"grad_norm": 9.135746950050816e-05,
|
77 |
+
"learning_rate": 1.852007518018085e-05,
|
78 |
+
"loss": 0.023,
|
79 |
+
"step": 5000
|
80 |
+
},
|
81 |
+
{
|
82 |
+
"epoch": 0.569771055630374,
|
83 |
+
"grad_norm": 2.3920888381212535e-08,
|
84 |
+
"learning_rate": 1.8372082698198932e-05,
|
85 |
+
"loss": 0.044,
|
86 |
+
"step": 5500
|
87 |
+
},
|
88 |
+
{
|
89 |
+
"epoch": 0.6215684243240444,
|
90 |
+
"grad_norm": 1.1770172932301648e-05,
|
91 |
+
"learning_rate": 1.8224090216217018e-05,
|
92 |
+
"loss": 0.0097,
|
93 |
+
"step": 6000
|
94 |
+
},
|
95 |
+
{
|
96 |
+
"epoch": 0.6733657930177147,
|
97 |
+
"grad_norm": 5.798747224616818e-05,
|
98 |
+
"learning_rate": 1.8076097734235104e-05,
|
99 |
+
"loss": 0.0628,
|
100 |
+
"step": 6500
|
101 |
+
},
|
102 |
+
{
|
103 |
+
"epoch": 0.7251631617113851,
|
104 |
+
"grad_norm": 0.0055756960064172745,
|
105 |
+
"learning_rate": 1.7928105252253187e-05,
|
106 |
+
"loss": 0.0167,
|
107 |
+
"step": 7000
|
108 |
+
},
|
109 |
+
{
|
110 |
+
"epoch": 0.7769605304050554,
|
111 |
+
"grad_norm": 6.234566535567865e-05,
|
112 |
+
"learning_rate": 1.7780112770271273e-05,
|
113 |
+
"loss": 0.0342,
|
114 |
+
"step": 7500
|
115 |
+
},
|
116 |
+
{
|
117 |
+
"epoch": 0.8287578990987258,
|
118 |
+
"grad_norm": 2.065628723357804e-05,
|
119 |
+
"learning_rate": 1.7632120288289356e-05,
|
120 |
+
"loss": 0.0278,
|
121 |
+
"step": 8000
|
122 |
+
},
|
123 |
+
{
|
124 |
+
"epoch": 0.8805552677923961,
|
125 |
+
"grad_norm": 0.00034213648177683353,
|
126 |
+
"learning_rate": 1.7484127806307442e-05,
|
127 |
+
"loss": 0.0777,
|
128 |
+
"step": 8500
|
129 |
+
},
|
130 |
+
{
|
131 |
+
"epoch": 0.9323526364860665,
|
132 |
+
"grad_norm": 0.0024548424407839775,
|
133 |
+
"learning_rate": 1.7336135324325525e-05,
|
134 |
+
"loss": 0.0117,
|
135 |
+
"step": 9000
|
136 |
+
},
|
137 |
+
{
|
138 |
+
"epoch": 0.9841500051797368,
|
139 |
+
"grad_norm": 2.010272328334395e-05,
|
140 |
+
"learning_rate": 1.718814284234361e-05,
|
141 |
+
"loss": 0.0149,
|
142 |
+
"step": 9500
|
143 |
+
},
|
144 |
+
{
|
145 |
+
"epoch": 1.0359473738734073,
|
146 |
+
"grad_norm": 3.7440368032548577e-05,
|
147 |
+
"learning_rate": 1.7040150360361697e-05,
|
148 |
+
"loss": 0.0059,
|
149 |
+
"step": 10000
|
150 |
+
},
|
151 |
+
{
|
152 |
+
"epoch": 1.0877447425670776,
|
153 |
+
"grad_norm": 9.011640031530987e-06,
|
154 |
+
"learning_rate": 1.689215787837978e-05,
|
155 |
+
"loss": 0.0245,
|
156 |
+
"step": 10500
|
157 |
+
},
|
158 |
+
{
|
159 |
+
"epoch": 1.139542111260748,
|
160 |
+
"grad_norm": 3.9126422052504495e-05,
|
161 |
+
"learning_rate": 1.6744165396397866e-05,
|
162 |
+
"loss": 0.0001,
|
163 |
+
"step": 11000
|
164 |
+
},
|
165 |
+
{
|
166 |
+
"epoch": 1.1913394799544184,
|
167 |
+
"grad_norm": 8.866464668244589e-06,
|
168 |
+
"learning_rate": 1.659617291441595e-05,
|
169 |
+
"loss": 0.0,
|
170 |
+
"step": 11500
|
171 |
+
},
|
172 |
+
{
|
173 |
+
"epoch": 1.2431368486480887,
|
174 |
+
"grad_norm": 0.00013718219997826964,
|
175 |
+
"learning_rate": 1.6448180432434035e-05,
|
176 |
+
"loss": 0.0181,
|
177 |
+
"step": 12000
|
178 |
+
},
|
179 |
+
{
|
180 |
+
"epoch": 1.294934217341759,
|
181 |
+
"grad_norm": 3.808485416811891e-06,
|
182 |
+
"learning_rate": 1.6300187950452117e-05,
|
183 |
+
"loss": 0.0,
|
184 |
+
"step": 12500
|
185 |
+
},
|
186 |
+
{
|
187 |
+
"epoch": 1.3467315860354294,
|
188 |
+
"grad_norm": 7.217061011033366e-07,
|
189 |
+
"learning_rate": 1.6152195468470203e-05,
|
190 |
+
"loss": 0.0266,
|
191 |
+
"step": 13000
|
192 |
+
},
|
193 |
+
{
|
194 |
+
"epoch": 1.3985289547290998,
|
195 |
+
"grad_norm": 7.131046731956303e-05,
|
196 |
+
"learning_rate": 1.600420298648829e-05,
|
197 |
+
"loss": 0.0266,
|
198 |
+
"step": 13500
|
199 |
+
},
|
200 |
+
{
|
201 |
+
"epoch": 1.4503263234227701,
|
202 |
+
"grad_norm": 9.411406608705875e-06,
|
203 |
+
"learning_rate": 1.5856210504506372e-05,
|
204 |
+
"loss": 0.0079,
|
205 |
+
"step": 14000
|
206 |
+
},
|
207 |
+
{
|
208 |
+
"epoch": 1.5021236921164405,
|
209 |
+
"grad_norm": 7.976142660481855e-05,
|
210 |
+
"learning_rate": 1.570821802252446e-05,
|
211 |
+
"loss": 0.022,
|
212 |
+
"step": 14500
|
213 |
+
},
|
214 |
+
{
|
215 |
+
"epoch": 1.5539210608101108,
|
216 |
+
"grad_norm": 1.0580498610579525e-06,
|
217 |
+
"learning_rate": 1.556022554054254e-05,
|
218 |
+
"loss": 0.0093,
|
219 |
+
"step": 15000
|
220 |
+
},
|
221 |
+
{
|
222 |
+
"epoch": 1.6057184295037812,
|
223 |
+
"grad_norm": 6.298066182353068e-06,
|
224 |
+
"learning_rate": 1.5412233058560627e-05,
|
225 |
+
"loss": 0.0,
|
226 |
+
"step": 15500
|
227 |
+
},
|
228 |
+
{
|
229 |
+
"epoch": 1.6575157981974515,
|
230 |
+
"grad_norm": 2.102418066030065e-12,
|
231 |
+
"learning_rate": 1.526424057657871e-05,
|
232 |
+
"loss": 0.0024,
|
233 |
+
"step": 16000
|
234 |
+
},
|
235 |
+
{
|
236 |
+
"epoch": 1.709313166891122,
|
237 |
+
"grad_norm": 3.009020701938425e-06,
|
238 |
+
"learning_rate": 1.5116248094596794e-05,
|
239 |
+
"loss": 0.009,
|
240 |
+
"step": 16500
|
241 |
+
},
|
242 |
+
{
|
243 |
+
"epoch": 1.7611105355847922,
|
244 |
+
"grad_norm": 6.2723142946197186e-06,
|
245 |
+
"learning_rate": 1.4968255612614882e-05,
|
246 |
+
"loss": 0.023,
|
247 |
+
"step": 17000
|
248 |
+
},
|
249 |
+
{
|
250 |
+
"epoch": 1.8129079042784626,
|
251 |
+
"grad_norm": 5.638932634610683e-06,
|
252 |
+
"learning_rate": 1.4820263130632967e-05,
|
253 |
+
"loss": 0.0099,
|
254 |
+
"step": 17500
|
255 |
+
},
|
256 |
+
{
|
257 |
+
"epoch": 1.8647052729721332,
|
258 |
+
"grad_norm": 3.8804391806479543e-05,
|
259 |
+
"learning_rate": 1.4672270648651051e-05,
|
260 |
+
"loss": 0.0185,
|
261 |
+
"step": 18000
|
262 |
+
},
|
263 |
+
{
|
264 |
+
"epoch": 1.9165026416658035,
|
265 |
+
"grad_norm": 3.445857828410226e-06,
|
266 |
+
"learning_rate": 1.4524278166669134e-05,
|
267 |
+
"loss": 0.0,
|
268 |
+
"step": 18500
|
269 |
+
},
|
270 |
+
{
|
271 |
+
"epoch": 1.9683000103594739,
|
272 |
+
"grad_norm": 0.0029530602041631937,
|
273 |
+
"learning_rate": 1.4376285684687218e-05,
|
274 |
+
"loss": 0.0243,
|
275 |
+
"step": 19000
|
276 |
+
},
|
277 |
+
{
|
278 |
+
"epoch": 2.0200973790531442,
|
279 |
+
"grad_norm": 7.133132271519571e-07,
|
280 |
+
"learning_rate": 1.4228293202705303e-05,
|
281 |
+
"loss": 0.0189,
|
282 |
+
"step": 19500
|
283 |
+
},
|
284 |
+
{
|
285 |
+
"epoch": 2.0718947477468146,
|
286 |
+
"grad_norm": 0.00042451228364370763,
|
287 |
+
"learning_rate": 1.4080300720723387e-05,
|
288 |
+
"loss": 0.0062,
|
289 |
+
"step": 20000
|
290 |
+
},
|
291 |
+
{
|
292 |
+
"epoch": 2.123692116440485,
|
293 |
+
"grad_norm": 1.8360736930844723e-06,
|
294 |
+
"learning_rate": 1.3932308238741471e-05,
|
295 |
+
"loss": 0.0067,
|
296 |
+
"step": 20500
|
297 |
+
},
|
298 |
+
{
|
299 |
+
"epoch": 2.1754894851341553,
|
300 |
+
"grad_norm": 0.0001334488915745169,
|
301 |
+
"learning_rate": 1.378431575675956e-05,
|
302 |
+
"loss": 0.006,
|
303 |
+
"step": 21000
|
304 |
+
},
|
305 |
+
{
|
306 |
+
"epoch": 2.2272868538278257,
|
307 |
+
"grad_norm": 4.610120413417462e-06,
|
308 |
+
"learning_rate": 1.3636323274777644e-05,
|
309 |
+
"loss": 0.0061,
|
310 |
+
"step": 21500
|
311 |
+
},
|
312 |
+
{
|
313 |
+
"epoch": 2.279084222521496,
|
314 |
+
"grad_norm": 2.7200339900446124e-06,
|
315 |
+
"learning_rate": 1.3488330792795728e-05,
|
316 |
+
"loss": 0.0,
|
317 |
+
"step": 22000
|
318 |
+
},
|
319 |
+
{
|
320 |
+
"epoch": 2.3308815912151664,
|
321 |
+
"grad_norm": 3.3594403703318676e-06,
|
322 |
+
"learning_rate": 1.3340338310813813e-05,
|
323 |
+
"loss": 0.009,
|
324 |
+
"step": 22500
|
325 |
+
},
|
326 |
+
{
|
327 |
+
"epoch": 2.3826789599088367,
|
328 |
+
"grad_norm": 5.500828137883218e-06,
|
329 |
+
"learning_rate": 1.3192345828831897e-05,
|
330 |
+
"loss": 0.0083,
|
331 |
+
"step": 23000
|
332 |
+
},
|
333 |
+
{
|
334 |
+
"epoch": 2.434476328602507,
|
335 |
+
"grad_norm": 1103.414306640625,
|
336 |
+
"learning_rate": 1.304435334684998e-05,
|
337 |
+
"loss": 0.0058,
|
338 |
+
"step": 23500
|
339 |
+
},
|
340 |
+
{
|
341 |
+
"epoch": 2.4862736972961774,
|
342 |
+
"grad_norm": 1.7569537931194645e-06,
|
343 |
+
"learning_rate": 1.2896360864868064e-05,
|
344 |
+
"loss": 0.0028,
|
345 |
+
"step": 24000
|
346 |
+
},
|
347 |
+
{
|
348 |
+
"epoch": 2.5380710659898478,
|
349 |
+
"grad_norm": 9.93580897556967e-07,
|
350 |
+
"learning_rate": 1.2748368382886152e-05,
|
351 |
+
"loss": 0.0052,
|
352 |
+
"step": 24500
|
353 |
+
},
|
354 |
+
{
|
355 |
+
"epoch": 2.589868434683518,
|
356 |
+
"grad_norm": 2.71925017225616e-12,
|
357 |
+
"learning_rate": 1.2600375900904236e-05,
|
358 |
+
"loss": 0.0,
|
359 |
+
"step": 25000
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"epoch": 2.6416658033771885,
|
363 |
+
"grad_norm": 1.8420889318804257e-05,
|
364 |
+
"learning_rate": 1.245238341892232e-05,
|
365 |
+
"loss": 0.0195,
|
366 |
+
"step": 25500
|
367 |
+
},
|
368 |
+
{
|
369 |
+
"epoch": 2.693463172070859,
|
370 |
+
"grad_norm": 8.20615071006614e-07,
|
371 |
+
"learning_rate": 1.2304390936940405e-05,
|
372 |
+
"loss": 0.008,
|
373 |
+
"step": 26000
|
374 |
+
},
|
375 |
+
{
|
376 |
+
"epoch": 2.745260540764529,
|
377 |
+
"grad_norm": 6.169057451188564e-05,
|
378 |
+
"learning_rate": 1.215639845495849e-05,
|
379 |
+
"loss": 0.0005,
|
380 |
+
"step": 26500
|
381 |
+
},
|
382 |
+
{
|
383 |
+
"epoch": 2.7970579094581995,
|
384 |
+
"grad_norm": 2.9846903544239467e-06,
|
385 |
+
"learning_rate": 1.2008405972976574e-05,
|
386 |
+
"loss": 0.0037,
|
387 |
+
"step": 27000
|
388 |
+
},
|
389 |
+
{
|
390 |
+
"epoch": 2.84885527815187,
|
391 |
+
"grad_norm": 8.764583071751986e-06,
|
392 |
+
"learning_rate": 1.1860413490994659e-05,
|
393 |
+
"loss": 0.0156,
|
394 |
+
"step": 27500
|
395 |
+
},
|
396 |
+
{
|
397 |
+
"epoch": 2.9006526468455403,
|
398 |
+
"grad_norm": 5.1639810408232734e-05,
|
399 |
+
"learning_rate": 1.1712421009012743e-05,
|
400 |
+
"loss": 0.0,
|
401 |
+
"step": 28000
|
402 |
+
},
|
403 |
+
{
|
404 |
+
"epoch": 2.9524500155392106,
|
405 |
+
"grad_norm": 8.454779163002968e-06,
|
406 |
+
"learning_rate": 1.1564428527030829e-05,
|
407 |
+
"loss": 0.0,
|
408 |
+
"step": 28500
|
409 |
+
},
|
410 |
+
{
|
411 |
+
"epoch": 3.004247384232881,
|
412 |
+
"grad_norm": 1.0601724653724887e-07,
|
413 |
+
"learning_rate": 1.1416436045048913e-05,
|
414 |
+
"loss": 0.0,
|
415 |
+
"step": 29000
|
416 |
+
},
|
417 |
+
{
|
418 |
+
"epoch": 3.0560447529265513,
|
419 |
+
"grad_norm": 3.302725417597685e-06,
|
420 |
+
"learning_rate": 1.1268443563066998e-05,
|
421 |
+
"loss": 0.0,
|
422 |
+
"step": 29500
|
423 |
+
},
|
424 |
+
{
|
425 |
+
"epoch": 3.1078421216202217,
|
426 |
+
"grad_norm": 8.728113243705593e-06,
|
427 |
+
"learning_rate": 1.1120451081085082e-05,
|
428 |
+
"loss": 0.0067,
|
429 |
+
"step": 30000
|
430 |
+
},
|
431 |
+
{
|
432 |
+
"epoch": 3.159639490313892,
|
433 |
+
"grad_norm": 2.4715068320801947e-06,
|
434 |
+
"learning_rate": 1.0972458599103167e-05,
|
435 |
+
"loss": 0.0,
|
436 |
+
"step": 30500
|
437 |
+
},
|
438 |
+
{
|
439 |
+
"epoch": 3.2114368590075624,
|
440 |
+
"grad_norm": 6.171033419377636e-06,
|
441 |
+
"learning_rate": 1.0824466117121251e-05,
|
442 |
+
"loss": 0.0,
|
443 |
+
"step": 31000
|
444 |
+
},
|
445 |
+
{
|
446 |
+
"epoch": 3.2632342277012327,
|
447 |
+
"grad_norm": 2.5147855922114104e-06,
|
448 |
+
"learning_rate": 1.0676473635139336e-05,
|
449 |
+
"loss": 0.0,
|
450 |
+
"step": 31500
|
451 |
+
},
|
452 |
+
{
|
453 |
+
"epoch": 3.315031596394903,
|
454 |
+
"grad_norm": 2.676899021025747e-05,
|
455 |
+
"learning_rate": 1.052848115315742e-05,
|
456 |
+
"loss": 0.006,
|
457 |
+
"step": 32000
|
458 |
+
},
|
459 |
+
{
|
460 |
+
"epoch": 3.3668289650885734,
|
461 |
+
"grad_norm": 2.081859747704584e-05,
|
462 |
+
"learning_rate": 1.0380488671175506e-05,
|
463 |
+
"loss": 0.0028,
|
464 |
+
"step": 32500
|
465 |
+
},
|
466 |
+
{
|
467 |
+
"epoch": 3.418626333782244,
|
468 |
+
"grad_norm": 2.3868110474722926e-06,
|
469 |
+
"learning_rate": 1.023249618919359e-05,
|
470 |
+
"loss": 0.0001,
|
471 |
+
"step": 33000
|
472 |
+
},
|
473 |
+
{
|
474 |
+
"epoch": 3.470423702475914,
|
475 |
+
"grad_norm": 2.7923347261094023e-06,
|
476 |
+
"learning_rate": 1.0084503707211675e-05,
|
477 |
+
"loss": 0.0,
|
478 |
+
"step": 33500
|
479 |
+
},
|
480 |
+
{
|
481 |
+
"epoch": 3.5222210711695845,
|
482 |
+
"grad_norm": 4.678757704823511e-06,
|
483 |
+
"learning_rate": 9.93651122522976e-06,
|
484 |
+
"loss": 0.0,
|
485 |
+
"step": 34000
|
486 |
+
},
|
487 |
+
{
|
488 |
+
"epoch": 3.574018439863255,
|
489 |
+
"grad_norm": 3.305537575215567e-06,
|
490 |
+
"learning_rate": 9.788518743247844e-06,
|
491 |
+
"loss": 0.0,
|
492 |
+
"step": 34500
|
493 |
+
},
|
494 |
+
{
|
495 |
+
"epoch": 3.625815808556925,
|
496 |
+
"grad_norm": 2.4619773739686934e-06,
|
497 |
+
"learning_rate": 9.640526261265928e-06,
|
498 |
+
"loss": 0.0,
|
499 |
+
"step": 35000
|
500 |
+
},
|
501 |
+
{
|
502 |
+
"epoch": 3.6776131772505956,
|
503 |
+
"grad_norm": 2.973723951527063e-07,
|
504 |
+
"learning_rate": 9.492533779284013e-06,
|
505 |
+
"loss": 0.0,
|
506 |
+
"step": 35500
|
507 |
+
},
|
508 |
+
{
|
509 |
+
"epoch": 3.729410545944266,
|
510 |
+
"grad_norm": 5.624352183986048e-07,
|
511 |
+
"learning_rate": 9.344541297302097e-06,
|
512 |
+
"loss": 0.012,
|
513 |
+
"step": 36000
|
514 |
+
},
|
515 |
+
{
|
516 |
+
"epoch": 3.7812079146379363,
|
517 |
+
"grad_norm": 1.8933849332825048e-07,
|
518 |
+
"learning_rate": 9.196548815320182e-06,
|
519 |
+
"loss": 0.0,
|
520 |
+
"step": 36500
|
521 |
+
},
|
522 |
+
{
|
523 |
+
"epoch": 3.8330052833316066,
|
524 |
+
"grad_norm": 3.9811013266444206e-05,
|
525 |
+
"learning_rate": 9.048556333338268e-06,
|
526 |
+
"loss": 0.0,
|
527 |
+
"step": 37000
|
528 |
+
},
|
529 |
+
{
|
530 |
+
"epoch": 3.884802652025277,
|
531 |
+
"grad_norm": 2.1272378944559023e-05,
|
532 |
+
"learning_rate": 8.900563851356352e-06,
|
533 |
+
"loss": 0.0073,
|
534 |
+
"step": 37500
|
535 |
+
},
|
536 |
+
{
|
537 |
+
"epoch": 3.9366000207189473,
|
538 |
+
"grad_norm": 8.419656296609901e-07,
|
539 |
+
"learning_rate": 8.752571369374436e-06,
|
540 |
+
"loss": 0.0045,
|
541 |
+
"step": 38000
|
542 |
+
},
|
543 |
+
{
|
544 |
+
"epoch": 3.9883973894126177,
|
545 |
+
"grad_norm": 1.4807918660153518e-06,
|
546 |
+
"learning_rate": 8.604578887392521e-06,
|
547 |
+
"loss": 0.0,
|
548 |
+
"step": 38500
|
549 |
+
},
|
550 |
+
{
|
551 |
+
"epoch": 4.0401947581062885,
|
552 |
+
"grad_norm": 2.789050199680787e-07,
|
553 |
+
"learning_rate": 8.456586405410605e-06,
|
554 |
+
"loss": 0.0,
|
555 |
+
"step": 39000
|
556 |
+
},
|
557 |
+
{
|
558 |
+
"epoch": 4.091992126799958,
|
559 |
+
"grad_norm": 7.712332603659888e-07,
|
560 |
+
"learning_rate": 8.30859392342869e-06,
|
561 |
+
"loss": 0.0,
|
562 |
+
"step": 39500
|
563 |
+
},
|
564 |
+
{
|
565 |
+
"epoch": 4.143789495493629,
|
566 |
+
"grad_norm": 1.126294500863878e-05,
|
567 |
+
"learning_rate": 8.160601441446774e-06,
|
568 |
+
"loss": 0.0,
|
569 |
+
"step": 40000
|
570 |
+
},
|
571 |
+
{
|
572 |
+
"epoch": 4.195586864187299,
|
573 |
+
"grad_norm": 1.1078836905653588e-05,
|
574 |
+
"learning_rate": 8.01260895946486e-06,
|
575 |
+
"loss": 0.0,
|
576 |
+
"step": 40500
|
577 |
+
},
|
578 |
+
{
|
579 |
+
"epoch": 4.24738423288097,
|
580 |
+
"grad_norm": 4.333252491051098e-06,
|
581 |
+
"learning_rate": 7.864616477482945e-06,
|
582 |
+
"loss": 0.0,
|
583 |
+
"step": 41000
|
584 |
+
},
|
585 |
+
{
|
586 |
+
"epoch": 4.29918160157464,
|
587 |
+
"grad_norm": 7.190360065578716e-06,
|
588 |
+
"learning_rate": 7.71662399550103e-06,
|
589 |
+
"loss": 0.0,
|
590 |
+
"step": 41500
|
591 |
+
},
|
592 |
+
{
|
593 |
+
"epoch": 4.350978970268311,
|
594 |
+
"grad_norm": 6.172657776915003e-06,
|
595 |
+
"learning_rate": 7.568631513519114e-06,
|
596 |
+
"loss": 0.0,
|
597 |
+
"step": 42000
|
598 |
+
},
|
599 |
+
{
|
600 |
+
"epoch": 4.4027763389619805,
|
601 |
+
"grad_norm": 1.1028377144839396e-07,
|
602 |
+
"learning_rate": 7.420639031537199e-06,
|
603 |
+
"loss": 0.0,
|
604 |
+
"step": 42500
|
605 |
+
},
|
606 |
+
{
|
607 |
+
"epoch": 4.454573707655651,
|
608 |
+
"grad_norm": 7.63295773253958e-08,
|
609 |
+
"learning_rate": 7.272646549555283e-06,
|
610 |
+
"loss": 0.0,
|
611 |
+
"step": 43000
|
612 |
+
},
|
613 |
+
{
|
614 |
+
"epoch": 4.506371076349321,
|
615 |
+
"grad_norm": 8.308877568197204e-07,
|
616 |
+
"learning_rate": 7.124654067573368e-06,
|
617 |
+
"loss": 0.0,
|
618 |
+
"step": 43500
|
619 |
+
},
|
620 |
+
{
|
621 |
+
"epoch": 4.558168445042992,
|
622 |
+
"grad_norm": 8.788915550894671e-08,
|
623 |
+
"learning_rate": 6.976661585591452e-06,
|
624 |
+
"loss": 0.0,
|
625 |
+
"step": 44000
|
626 |
+
},
|
627 |
+
{
|
628 |
+
"epoch": 4.609965813736662,
|
629 |
+
"grad_norm": 5.980305104458239e-07,
|
630 |
+
"learning_rate": 6.828669103609537e-06,
|
631 |
+
"loss": 0.0,
|
632 |
+
"step": 44500
|
633 |
+
},
|
634 |
+
{
|
635 |
+
"epoch": 4.661763182430333,
|
636 |
+
"grad_norm": 0.00010543836833676323,
|
637 |
+
"learning_rate": 6.680676621627622e-06,
|
638 |
+
"loss": 0.0034,
|
639 |
+
"step": 45000
|
640 |
+
},
|
641 |
+
{
|
642 |
+
"epoch": 4.713560551124003,
|
643 |
+
"grad_norm": 2.6961990442941897e-05,
|
644 |
+
"learning_rate": 6.532684139645706e-06,
|
645 |
+
"loss": 0.0,
|
646 |
+
"step": 45500
|
647 |
+
},
|
648 |
+
{
|
649 |
+
"epoch": 4.765357919817673,
|
650 |
+
"grad_norm": 2.6214322133455426e-05,
|
651 |
+
"learning_rate": 6.384691657663791e-06,
|
652 |
+
"loss": 0.0006,
|
653 |
+
"step": 46000
|
654 |
+
},
|
655 |
+
{
|
656 |
+
"epoch": 4.817155288511343,
|
657 |
+
"grad_norm": 6.838554782007122e-06,
|
658 |
+
"learning_rate": 6.236699175681876e-06,
|
659 |
+
"loss": 0.0,
|
660 |
+
"step": 46500
|
661 |
+
},
|
662 |
+
{
|
663 |
+
"epoch": 4.868952657205014,
|
664 |
+
"grad_norm": 1.3388408660830464e-05,
|
665 |
+
"learning_rate": 6.08870669369996e-06,
|
666 |
+
"loss": 0.0,
|
667 |
+
"step": 47000
|
668 |
+
},
|
669 |
+
{
|
670 |
+
"epoch": 4.920750025898684,
|
671 |
+
"grad_norm": 1.1914085007447284e-06,
|
672 |
+
"learning_rate": 5.940714211718045e-06,
|
673 |
+
"loss": 0.0,
|
674 |
+
"step": 47500
|
675 |
+
},
|
676 |
+
{
|
677 |
+
"epoch": 4.972547394592355,
|
678 |
+
"grad_norm": 1.9319197235745378e-05,
|
679 |
+
"learning_rate": 5.792721729736129e-06,
|
680 |
+
"loss": 0.0,
|
681 |
+
"step": 48000
|
682 |
+
},
|
683 |
+
{
|
684 |
+
"epoch": 5.024344763286025,
|
685 |
+
"grad_norm": 7.528370815634844e-07,
|
686 |
+
"learning_rate": 5.6447292477542145e-06,
|
687 |
+
"loss": 0.0058,
|
688 |
+
"step": 48500
|
689 |
+
},
|
690 |
+
{
|
691 |
+
"epoch": 5.0761421319796955,
|
692 |
+
"grad_norm": 1.075523073268414e-06,
|
693 |
+
"learning_rate": 5.496736765772299e-06,
|
694 |
+
"loss": 0.0,
|
695 |
+
"step": 49000
|
696 |
+
},
|
697 |
+
{
|
698 |
+
"epoch": 5.1279395006733655,
|
699 |
+
"grad_norm": 4.6377437001865474e-07,
|
700 |
+
"learning_rate": 5.348744283790383e-06,
|
701 |
+
"loss": 0.0,
|
702 |
+
"step": 49500
|
703 |
+
},
|
704 |
+
{
|
705 |
+
"epoch": 5.179736869367036,
|
706 |
+
"grad_norm": 6.992227667979023e-07,
|
707 |
+
"learning_rate": 5.2007518018084694e-06,
|
708 |
+
"loss": 0.0,
|
709 |
+
"step": 50000
|
710 |
+
},
|
711 |
+
{
|
712 |
+
"epoch": 5.231534238060706,
|
713 |
+
"grad_norm": 2.332795929760323e-06,
|
714 |
+
"learning_rate": 5.052759319826553e-06,
|
715 |
+
"loss": 0.0,
|
716 |
+
"step": 50500
|
717 |
+
},
|
718 |
+
{
|
719 |
+
"epoch": 5.283331606754377,
|
720 |
+
"grad_norm": 5.32125454810739e-07,
|
721 |
+
"learning_rate": 4.9047668378446374e-06,
|
722 |
+
"loss": 0.0,
|
723 |
+
"step": 51000
|
724 |
+
}
|
725 |
+
],
|
726 |
+
"logging_steps": 500,
|
727 |
+
"max_steps": 67571,
|
728 |
+
"num_input_tokens_seen": 0,
|
729 |
+
"num_train_epochs": 7,
|
730 |
+
"save_steps": 500,
|
731 |
+
"stateful_callbacks": {
|
732 |
+
"TrainerControl": {
|
733 |
+
"args": {
|
734 |
+
"should_epoch_stop": false,
|
735 |
+
"should_evaluate": false,
|
736 |
+
"should_log": false,
|
737 |
+
"should_save": true,
|
738 |
+
"should_training_stop": false
|
739 |
+
},
|
740 |
+
"attributes": {}
|
741 |
+
}
|
742 |
+
},
|
743 |
+
"total_flos": 1.4080058755834675e+17,
|
744 |
+
"train_batch_size": 4,
|
745 |
+
"trial_name": null,
|
746 |
+
"trial_params": null
|
747 |
+
}
|
kraken_router/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:49df63d21c21735b6ffb8853fe9bbbebb0e5df9f446e7a960be8881b2f46ec03
|
3 |
+
size 5048
|
kraken_router/vocab.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
modeling_kraken.py
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from transformers import PreTrainedModel, AutoTokenizer, AutoModelForCausalLM, AutoModelForSequenceClassification, TextClassificationPipeline
|
3 |
+
from configuration_kraken import KrakenConfig
|
4 |
+
import tokenizer_template_switch
|
5 |
+
|
6 |
+
class KrakenForCausalLM(PreTrainedModel):
|
7 |
+
config_class = KrakenConfig
|
8 |
+
|
9 |
+
def __init__(self, config):
|
10 |
+
super().__init__(config)
|
11 |
+
self.tokenizers = {key: AutoTokenizer.from_pretrained(name, device_map="auto") for key, name in config.config_dict['tokenizers'].items()}
|
12 |
+
self.models = self.load_expert_models(config.config_dict['models'], config.config_dict['quantization'])
|
13 |
+
self.router_model = AutoModelForSequenceClassification.from_pretrained(config.config_dict['router'], trust_remote_code=True,device_map="auto")
|
14 |
+
self.tokenizer = AutoTokenizer.from_pretrained(config.config_dict['router'], trust_remote_code=True,device_map="auto")
|
15 |
+
self.router = TextClassificationPipeline(model=self.router_model, tokenizer=self.tokenizer)
|
16 |
+
self.models_indices = config.config_dict['class_indices']
|
17 |
+
|
18 |
+
def load_expert_models(self, models_dict, quantization_dict):
|
19 |
+
models = {}
|
20 |
+
for key, name in models_dict.items():
|
21 |
+
quantization = quantization_dict.get(key)
|
22 |
+
if quantization == "8bit":
|
23 |
+
models[key] = AutoModelForCausalLM.from_pretrained(name, trust_remote_code=True, device_map="auto", load_in_8bit=True, torch_dtype="auto")
|
24 |
+
elif quantization == "4bit":
|
25 |
+
models[key] = AutoModelForCausalLM.from_pretrained(name, trust_remote_code=True, device_map="auto", load_in_4bit=True, torch_dtype="auto")
|
26 |
+
elif quantization == "awq":
|
27 |
+
models[key] = self.load_awq_model(name)
|
28 |
+
else:
|
29 |
+
models[key] = AutoModelForCausalLM.from_pretrained(name, trust_remote_code=True, device_map="auto", torch_dtype="auto")
|
30 |
+
return models
|
31 |
+
|
32 |
+
def load_awq_model(self, name):
|
33 |
+
return AutoModelForCausalLM.from_pretrained(name, trust_remote_code=True, device_map="auto")
|
34 |
+
|
35 |
+
def tokenize_inputs(self, text, model_key):
|
36 |
+
return self.tokenizers[model_key](text, return_tensors="pt")
|
37 |
+
|
38 |
+
def determine_model(self, text):
|
39 |
+
prediction = self.router(text)[0]["label"]
|
40 |
+
model_decision_index = self.models_indices[prediction]
|
41 |
+
model_keys = ['expert1', 'expert2', 'expert3', 'expert4','expert5']
|
42 |
+
return model_keys[model_decision_index]
|
43 |
+
|
44 |
+
def expert_tokenizer(self, text):
|
45 |
+
model_key = self.determine_model(text)
|
46 |
+
return self.tokenizers[model_key]
|
47 |
+
|
48 |
+
|
49 |
+
def generate(self, input_ids, **generate_kwargs):
|
50 |
+
# Tokenize the input_ids
|
51 |
+
text = self.tokenizer.batch_decode(input_ids, skip_special_tokens=False)[0]
|
52 |
+
|
53 |
+
msgs = tokenizer_template_switch.recover_chat_messages(text, self.tokenizer)
|
54 |
+
if msgs and msgs[0]['role'] == 'system' and msgs[0]['content']=='<|im_start|>system':
|
55 |
+
# Delete the first element
|
56 |
+
msgs.pop(0)
|
57 |
+
# Check if the last element has the role 'assistant'
|
58 |
+
if msgs and msgs[-1]['role'] == 'assistant':
|
59 |
+
# Delete the last element
|
60 |
+
msgs.pop()
|
61 |
+
|
62 |
+
# Determine the model key using the existing routing logic
|
63 |
+
model_key = self.determine_model(text)
|
64 |
+
# Show the routing result
|
65 |
+
print(f"Choosing {model_key} ..")
|
66 |
+
# Retrieve the model from the dictionary
|
67 |
+
model = self.models[model_key]
|
68 |
+
|
69 |
+
mod_txt = self.tokenizers[model_key].apply_chat_template(msgs, tokenize=False, add_generation_prompt=True)
|
70 |
+
current_device = input_ids.device if isinstance(input_ids, torch.Tensor) else 'cpu'
|
71 |
+
|
72 |
+
# Tokenize accordingly to the best model
|
73 |
+
|
74 |
+
tok = self.tokenizers[model_key](mod_txt, return_tensors="pt")
|
75 |
+
tok_input_ids = tok.input_ids.to(current_device)
|
76 |
+
tok_attention_mask = tok.attention_mask.to(current_device)
|
77 |
+
|
78 |
+
# Generate text using the retrieved model
|
79 |
+
return model.generate(tok_input_ids, attention_mask=tok_attention_mask, **generate_kwargs)
|
80 |
+
|
81 |
+
|
82 |
+
|
tokenizer_template_switch.py
ADDED
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import re
|
2 |
+
from transformers import AutoTokenizer
|
3 |
+
|
4 |
+
def extract_separators(template):
|
5 |
+
"""
|
6 |
+
Extracts separators used in the tokenization template.
|
7 |
+
"""
|
8 |
+
# Adjust the regex to correctly match the specific pattern between '{{' and '+ message["content"] +'
|
9 |
+
pattern = r"\{\{\s*([^{}]+?)\s*\+ message\['content'\]"
|
10 |
+
matches = re.findall(pattern, template)
|
11 |
+
# Clean up any extra spaces and return the matches
|
12 |
+
separators = [match.strip() for match in matches]
|
13 |
+
|
14 |
+
if any("message['role']" in element for element in separators):
|
15 |
+
roles = ["system", "user", "assistant"]
|
16 |
+
separators_ = []
|
17 |
+
for role in roles:
|
18 |
+
separators_.append(separators[0].replace(" + message['role'] + ", role).replace("'",""))
|
19 |
+
return separators_
|
20 |
+
|
21 |
+
return separators
|
22 |
+
|
23 |
+
def detect_eos_token(jinja_template, tokenizer):
|
24 |
+
if "<|im_end|>" in jinja_template:
|
25 |
+
return "<|im_end|>"
|
26 |
+
if "</s>" in jinja_template:
|
27 |
+
return "</s>"
|
28 |
+
if "eos_token" in jinja_template:
|
29 |
+
return tokenizer.eos_token
|
30 |
+
if "<|endoftext|>" in jinja_template:
|
31 |
+
return tokenizer.eos_token
|
32 |
+
else:
|
33 |
+
return "<|endoftext|>"
|
34 |
+
|
35 |
+
def recover_messages(formatted_message, separators, eos_token):
|
36 |
+
"""
|
37 |
+
Recovers the original messages from the formatted message string.
|
38 |
+
"""
|
39 |
+
# Split the formatted message using the end-of-string token
|
40 |
+
split_messages = formatted_message.split(eos_token)
|
41 |
+
|
42 |
+
# Remove the last empty string if it exists due to a trailing separator
|
43 |
+
if split_messages and split_messages[-1].strip() == '':
|
44 |
+
split_messages.pop()
|
45 |
+
|
46 |
+
# Prepare the list to hold the recovered messages
|
47 |
+
recovered_messages = []
|
48 |
+
|
49 |
+
# Define roles after the first message, alternating between "user" and "assistant"
|
50 |
+
alternate_roles = ["user", "assistant"]
|
51 |
+
|
52 |
+
# Iterate over the split messages
|
53 |
+
for index, message_content in enumerate(split_messages):
|
54 |
+
# Determine the role, starting with "system" for the first message
|
55 |
+
# then alternating between "user" and "assistant" for subsequent messages
|
56 |
+
if index == 0:
|
57 |
+
role = "system"
|
58 |
+
else:
|
59 |
+
role = alternate_roles[(index - 1) % 2]
|
60 |
+
|
61 |
+
# Clean the message content by removing leading/trailing whitespace and separators
|
62 |
+
clean_content = message_content.strip()
|
63 |
+
for separator in separators:
|
64 |
+
clean_content = clean_content.replace(separator.strip("'"), '', 1).strip()
|
65 |
+
|
66 |
+
# Append the cleaned message with its role to the list
|
67 |
+
recovered_messages.append({"role": role, "content": clean_content})
|
68 |
+
|
69 |
+
return recovered_messages
|
70 |
+
|
71 |
+
def recover_chat_messages(tokenized_chat, tokenizer):
|
72 |
+
"""
|
73 |
+
Given a tokenized_chat string and a tokenizer, returns the list of message dictionaries.
|
74 |
+
"""
|
75 |
+
jinja_template = tokenizer.chat_template
|
76 |
+
separators = extract_separators(jinja_template)
|
77 |
+
eos_token = eos_token = detect_eos_token(jinja_template, tokenizer)
|
78 |
+
recovered_messages = recover_messages(tokenized_chat, separators, eos_token)
|
79 |
+
return recovered_messages
|
80 |
+
|
81 |
+
# Example usage
|
82 |
+
if __name__ == "__main__":
|
83 |
+
checkpoint = "Qwen/Qwen1.5-0.5B"
|
84 |
+
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
85 |
+
|
86 |
+
messages = [
|
87 |
+
{
|
88 |
+
"role": "system",
|
89 |
+
"content": "You are a friendly chatbot who always responds in the style of a pirate",
|
90 |
+
},
|
91 |
+
{"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
|
92 |
+
]
|
93 |
+
tokenized_chat = tokenizer.apply_chat_template(messages, tokenize=False)
|
94 |
+
print(tokenized_chat)
|
95 |
+
|
96 |
+
recovered_messages = recover_chat_messages(tokenized_chat, tokenizer)
|
97 |
+
print(recovered_messages)
|