Delta-Vector commited on
Commit
e403bdb
·
verified ·
1 Parent(s): 9f987fd

Update rp.yml

Browse files
Files changed (1) hide show
  1. rp.yml +1 -110
rp.yml CHANGED
@@ -1,110 +1 @@
1
- base_model: NewEden/32B-inst
2
- model_type: AutoModelForCausalLM
3
- tokenizer_type: AutoTokenizer
4
-
5
- hub_model_id: NewEden/32b-rp
6
- hub_strategy: "all_checkpoints"
7
- push_dataset_to_hub:
8
- hf_use_auth_token: true
9
-
10
- plugins:
11
- - axolotl.integrations.liger.LigerPlugin
12
- - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
13
- liger_rope: true
14
- liger_rms_norm: true
15
- liger_layer_norm: true
16
- liger_glu_activation: true
17
- liger_fused_linear_cross_entropy: false
18
- cut_cross_entropy: true
19
-
20
- load_in_8bit: false
21
- load_in_4bit: false
22
- strict: false
23
-
24
- datasets:
25
- - path: NewEden/RP-logs-V2-Experimental-prefixed
26
- type: dan-chat-advanced
27
- - path: NewEden/Creative_Writing-Complexity
28
- type: dan-chat-advanced
29
- - path: NewEden/Discord-Filtered
30
- type: dan-chat-advanced
31
- - path: NewEden/DeepseekRP-Filtered
32
- type: dan-chat-advanced
33
- - path: NewEden/Storium-Prefixed-Clean
34
- type: dan-chat-advanced
35
- - path: NewEden/Basket-Weaving-Filtered
36
- type: dan-chat-advanced
37
- - path: NewEden/LIMARP-Complexity
38
- type: dan-chat-advancedd
39
- - path: NewEden/Misc-Data-Sharegpt-Prefixed
40
- type: dan-chat-advanced
41
- - path: NewEden/BlueSky-10K-Complexity
42
- type: dan-chat-advanced
43
- - path: NewEden/OpenCAI-ShareGPT
44
- type: dan-chat-advanced
45
- - path: NewEden/Basket-Weaving-Filtered
46
- type: dan-chat-advanced
47
- - path: PocketDoc/Dans-Personamaxx-VN
48
- type: dan-chat-advanced
49
- - path: PocketDoc/Dans-Kinomaxx-VanillaBackrooms
50
- type: dan-chat-advanced
51
- dataset_prepared_path: prepared_data
52
- val_set_size: 0.02
53
- output_dir: ./qwq-inst
54
-
55
- sequence_len: 32768
56
- sample_packing: true
57
- pad_to_sequence_len: true
58
-
59
- # adapter: lora
60
- # lora_model_dir:
61
- # lora_r: 128
62
- # lora_alpha: 16
63
- # lora_dropout: 0.05
64
- # lora_target_modules:
65
- # - gate_proj
66
- # - down_proj
67
- # - up_proj
68
- # - q_proj
69
- # - v_proj
70
- # - k_proj
71
- # - o_proj
72
-
73
- wandb_project: qwq
74
- wandb_entity:
75
- wandb_watch:
76
- wandb_name: rp-attempt-01
77
- wandb_log_model:
78
-
79
- gradient_accumulation_steps: 2
80
- micro_batch_size: 2
81
- num_epochs: 4
82
- optimizer: adamw_bnb_8bit
83
- lr_scheduler: cosine
84
- learning_rate: 3.17e-5
85
-
86
- train_on_inputs: false
87
- group_by_length: false
88
- bf16: auto
89
- fp16:
90
- tf32: false
91
-
92
- gradient_checkpointing: true
93
- early_stopping_patience:
94
- resume_from_checkpoint:
95
- local_rank:
96
- logging_steps: 1
97
- xformers_attention:
98
- flash_attention: true
99
-
100
- warmup_steps: 40
101
- evals_per_epoch: 4
102
- eval_table_size: 128
103
- eval_max_new_tokens:
104
- saves_per_epoch: 2
105
- debug:
106
- deepspeed: deepspeed_configs/zero3_bf16.json
107
- weight_decay: 0.02
108
- fsdp:
109
- fsdp_config:
110
- special_tokens:
 
1
+ https://huggingface.co/datasets/Delta-Vector/Configs/blob/main/mag.yml