uncensored
Neko-Institute-of-Science commited on
Commit
4687029
1 Parent(s): 5a0db5e

Add last checkpoint of the failed training format.

Browse files

This first test used the original format from the convert tool, but it was later found out this caused broken context. It would work as expected from the initial prompt but the moment you asked it a question about anything in the past it would say something random.

checkpoint-9728-failed/adapter_config.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model_name_or_path": "models/llama-13b",
3
+ "bias": "none",
4
+ "fan_in_fan_out": false,
5
+ "inference_mode": true,
6
+ "init_lora_weights": true,
7
+ "lora_alpha": 128,
8
+ "lora_dropout": 0,
9
+ "modules_to_save": null,
10
+ "peft_type": "LORA",
11
+ "r": 64,
12
+ "target_modules": [
13
+ "q_proj",
14
+ "v_proj"
15
+ ],
16
+ "task_type": "CAUSAL_LM"
17
+ }
checkpoint-9728-failed/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fdbccce795dcac8803f9e5d01dc538261095572e7be3a416ea27f1413ee022c
3
+ size 209772877
checkpoint-9728-failed/training_parameters.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"lora_name": "VicUnLocked", "always_override": false, "save_steps": 500.0, "micro_batch_size": 4, "batch_size": 128, "epochs": 3.0, "learning_rate": "3e-4", "lr_scheduler_type": "constant_with_warmup", "lora_rank": 64, "lora_alpha": 128, "lora_dropout": 0, "cutoff_len": 2048, "dataset": "vicuna-all", "eval_dataset": "None", "format": "vicuna-format", "eval_steps": 100.0, "raw_text_file": "None", "overlap_len": 128, "newline_favor_len": 128, "higher_rank_limit": false, "warmup_steps": 100.0, "optimizer": "adamw_torch"}