End of training
Browse files- README.md +137 -0
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
README.md
ADDED
@@ -0,0 +1,137 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
base_model: pszemraj/Mistral-7B-v0.3-prune6
|
4 |
+
tags:
|
5 |
+
- axolotl
|
6 |
+
- generated_from_trainer
|
7 |
+
model-index:
|
8 |
+
- name: Mistral-v0.3-6B-ii
|
9 |
+
results: []
|
10 |
+
---
|
11 |
+
|
12 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
13 |
+
should probably proofread and complete it, then remove this comment. -->
|
14 |
+
|
15 |
+
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|
16 |
+
<details><summary>See axolotl config</summary>
|
17 |
+
|
18 |
+
axolotl version: `0.4.0`
|
19 |
+
```yaml
|
20 |
+
base_model: pszemraj/Mistral-7B-v0.3-prune6
|
21 |
+
model_type: MistralForCausalLM
|
22 |
+
tokenizer_type: LlamaTokenizer
|
23 |
+
|
24 |
+
strict: false
|
25 |
+
seed: 80085
|
26 |
+
max_steps: 2000
|
27 |
+
# dataset
|
28 |
+
datasets:
|
29 |
+
- path: BEE-spoke-data/knowledge-inoc-concat-v1
|
30 |
+
name: smorgasbord-tb-quality
|
31 |
+
type: completion
|
32 |
+
field: text
|
33 |
+
val_set_size: 0.01
|
34 |
+
|
35 |
+
sequence_len: 4096
|
36 |
+
sample_packing: true
|
37 |
+
pad_to_sequence_len: false
|
38 |
+
train_on_inputs: false
|
39 |
+
group_by_length: false
|
40 |
+
|
41 |
+
# WANDB
|
42 |
+
wandb_project: llama3-pruning
|
43 |
+
wandb_entity: pszemraj
|
44 |
+
wandb_watch: gradients
|
45 |
+
wandb_name: Mistral-6B-v0.3-v0.1-ii
|
46 |
+
hub_model_id: pszemraj/Mistral-v0.3-6B-ii
|
47 |
+
hub_strategy: every_save
|
48 |
+
|
49 |
+
gradient_accumulation_steps: 16
|
50 |
+
micro_batch_size: 1
|
51 |
+
num_epochs: 1
|
52 |
+
optimizer: paged_adamw_32bit
|
53 |
+
weight_decay: 0.1
|
54 |
+
lr_scheduler: cosine
|
55 |
+
learning_rate: 2e-5
|
56 |
+
warmup_ratio: 0.1
|
57 |
+
|
58 |
+
load_in_8bit: false
|
59 |
+
load_in_4bit: false
|
60 |
+
bfloat16: true
|
61 |
+
tf32: true
|
62 |
+
|
63 |
+
flash_attention: true
|
64 |
+
torch_compile: true
|
65 |
+
torch_compile_backend: inductor
|
66 |
+
gradient_checkpointing: true
|
67 |
+
gradient_checkpointing_kwargs:
|
68 |
+
use_reentrant: false
|
69 |
+
|
70 |
+
# hyperparams for freq of evals, saving, etc
|
71 |
+
evals_per_epoch: 5
|
72 |
+
saves_per_epoch: 5
|
73 |
+
save_safetensors: true
|
74 |
+
save_total_limit: 1
|
75 |
+
output_dir: /workspace/output-axolotl/output-model-6b
|
76 |
+
logging_steps: 6
|
77 |
+
|
78 |
+
deepspeed:
|
79 |
+
|
80 |
+
special_tokens:
|
81 |
+
|
82 |
+
```
|
83 |
+
|
84 |
+
</details><br>
|
85 |
+
|
86 |
+
# Mistral-v0.3-6B-ii
|
87 |
+
|
88 |
+
This model is a fine-tuned version of [pszemraj/Mistral-7B-v0.3-prune6](https://huggingface.co/pszemraj/Mistral-7B-v0.3-prune6) on the None dataset.
|
89 |
+
It achieves the following results on the evaluation set:
|
90 |
+
- Loss: 1.2860
|
91 |
+
|
92 |
+
## Model description
|
93 |
+
|
94 |
+
More information needed
|
95 |
+
|
96 |
+
## Intended uses & limitations
|
97 |
+
|
98 |
+
More information needed
|
99 |
+
|
100 |
+
## Training and evaluation data
|
101 |
+
|
102 |
+
More information needed
|
103 |
+
|
104 |
+
## Training procedure
|
105 |
+
|
106 |
+
### Training hyperparameters
|
107 |
+
|
108 |
+
The following hyperparameters were used during training:
|
109 |
+
- learning_rate: 2e-05
|
110 |
+
- train_batch_size: 1
|
111 |
+
- eval_batch_size: 1
|
112 |
+
- seed: 80085
|
113 |
+
- gradient_accumulation_steps: 16
|
114 |
+
- total_train_batch_size: 16
|
115 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
116 |
+
- lr_scheduler_type: cosine
|
117 |
+
- lr_scheduler_warmup_steps: 200
|
118 |
+
- training_steps: 2000
|
119 |
+
|
120 |
+
### Training results
|
121 |
+
|
122 |
+
| Training Loss | Epoch | Step | Validation Loss |
|
123 |
+
|:-------------:|:------:|:----:|:---------------:|
|
124 |
+
| No log | 0.0002 | 1 | 1.5980 |
|
125 |
+
| 1.578 | 0.0955 | 400 | 1.4028 |
|
126 |
+
| 1.5828 | 0.1911 | 800 | 1.3809 |
|
127 |
+
| 1.4355 | 0.2866 | 1200 | 1.3152 |
|
128 |
+
| 1.4618 | 0.3822 | 1600 | 1.2877 |
|
129 |
+
| 1.4551 | 0.4777 | 2000 | 1.2860 |
|
130 |
+
|
131 |
+
|
132 |
+
### Framework versions
|
133 |
+
|
134 |
+
- Transformers 4.40.2
|
135 |
+
- Pytorch 2.3.0+cu118
|
136 |
+
- Datasets 2.19.1
|
137 |
+
- Tokenizers 0.19.1
|
model-00001-of-00003.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4949453792
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f6387739d7045c7530789f372298a8a81decd83c89d78b6e85d2ce23a924509c
|
3 |
size 4949453792
|
model-00002-of-00003.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4999819336
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3efbae4d9a6e86249c8121a5f6e89c79e111f0b3079a56e539899847f679644c
|
3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1929457496
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:00098a24611641eccceba7b3ed58032ff1f2cf486050f0fc3716fb0cd4fbd3fc
|
3 |
size 1929457496
|