Init
Browse files- .gitattributes +1 -0
- README.md +65 -0
- adapter_config.json +26 -0
- adapter_model.bin +3 -0
.gitattributes
CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
+
adapter_model.bin filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
@@ -1,3 +1,68 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: llama2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
library_name: transformers
|
5 |
+
library_name: peft
|
6 |
+
pipeline_tag: text-generation
|
7 |
+
tags:
|
8 |
+
- Mistral
|
9 |
license: llama2
|
10 |
+
model-index:
|
11 |
+
- name: SpeechlessCoder
|
12 |
+
results:
|
13 |
+
- task:
|
14 |
+
type: text-generation
|
15 |
+
dataset:
|
16 |
+
type: openai_humaneval
|
17 |
+
name: HumanEval
|
18 |
+
metrics:
|
19 |
+
- name: pass@1
|
20 |
+
type: pass@1
|
21 |
+
value: 0.0
|
22 |
+
verified: false
|
23 |
---
|
24 |
+
|
25 |
+
**This is a test.**
|
26 |
+
|
27 |
+
|
28 |
+
This LoRA model is extracted from the efficient parameter fine-tuned model, and now it needs to be verified whether this LoRA model can achieve comparable performance with the original model.
|
29 |
+
|
30 |
+
The final goal is to create a toolkit that can simultaneously load multiple LoRA modules, and automatically switch to the appropriate combination of LoRA modules based on user queries to generate the best answer.
|
31 |
+
|
32 |
+
|
33 |
+
## lm-evaluation-harness
|
34 |
+
|
35 |
+
[Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
36 |
+
|
37 |
+
| Metric | Mistral-7B-OpenOrca | Mistral-7B-OpenOrca-lora|
|
38 |
+
| --- | --- |--- |
|
39 |
+
| ARC | 64.08 | |
|
40 |
+
| HellaSwag | 83.99 | |
|
41 |
+
| MMLU | 62.24 | |
|
42 |
+
| TruthfulQA | 53.05 | |
|
43 |
+
| Average | 65.84 | |
|
44 |
+
|
45 |
+
## HumanEval
|
46 |
+
|
47 |
+
| Metric | Mistral-7B-OpenOrca | Mistral-7B-OpenOrca-lora|
|
48 |
+
| --- | --- | --- |
|
49 |
+
| humaneval-python | 35.976 | |
|
50 |
+
|
51 |
+
|
52 |
+
## Training procedure
|
53 |
+
|
54 |
+
The following `bitsandbytes` quantization config was used during training:
|
55 |
+
- quant_method: bitsandbytes
|
56 |
+
- load_in_8bit: False
|
57 |
+
- load_in_4bit: True
|
58 |
+
- llm_int8_threshold: 6.0
|
59 |
+
- llm_int8_skip_modules: None
|
60 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
61 |
+
- llm_int8_has_fp16_weight: False
|
62 |
+
- bnb_4bit_quant_type: nf4
|
63 |
+
- bnb_4bit_use_double_quant: True
|
64 |
+
- bnb_4bit_compute_dtype: bfloat16
|
65 |
+
### Framework versions
|
66 |
+
|
67 |
+
|
68 |
+
- PEFT 0.5.0
|
adapter_config.json
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"auto_mapping": null,
|
3 |
+
"base_model_name_or_path": "/opt/local/llm_models/huggingface.co/mistralai/Mistral-7B-v0.1",
|
4 |
+
"bias": "none",
|
5 |
+
"fan_in_fan_out": false,
|
6 |
+
"inference_mode": true,
|
7 |
+
"init_lora_weights": true,
|
8 |
+
"layers_pattern": null,
|
9 |
+
"layers_to_transform": null,
|
10 |
+
"lora_alpha": 16,
|
11 |
+
"lora_dropout": 0.05,
|
12 |
+
"modules_to_save": null,
|
13 |
+
"peft_type": "LORA",
|
14 |
+
"r": 64,
|
15 |
+
"revision": null,
|
16 |
+
"target_modules": [
|
17 |
+
"down_proj",
|
18 |
+
"k_proj",
|
19 |
+
"q_proj",
|
20 |
+
"up_proj",
|
21 |
+
"gate_proj",
|
22 |
+
"o_proj",
|
23 |
+
"v_proj"
|
24 |
+
],
|
25 |
+
"task_type": "CAUSAL_LM"
|
26 |
+
}
|
adapter_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2b9ef2dba141352f3648b41cd42192e8d3ebac489b80252f04794b2cd0654bf1
|
3 |
+
size 335699405
|