Add Lab21 LoRA r=16 adapter (Qwen2.5-3B Vietnamese Alpaca)
Browse files- README.md +35 -0
- adapter_config.json +47 -0
- adapter_model.safetensors +3 -0
README.md
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: unsloth/Qwen2.5-3B-bnb-4bit
|
| 3 |
+
library_name: peft
|
| 4 |
+
pipeline_tag: text-generation
|
| 5 |
+
tags:
|
| 6 |
+
- lora
|
| 7 |
+
- qlora
|
| 8 |
+
- vietnamese
|
| 9 |
+
- lab21
|
| 10 |
+
license: apache-2.0
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# Lab 21 — Qwen2.5-3B Vietnamese LoRA (r=16)
|
| 14 |
+
|
| 15 |
+
LoRA adapter fine-tuned on `5CD-AI/Vietnamese-alpaca-gpt4-gg-translated` (200 samples).
|
| 16 |
+
|
| 17 |
+
| Setting | Value |
|
| 18 |
+
|---------|-------|
|
| 19 |
+
| Base model | `unsloth/Qwen2.5-3B-bnb-4bit` |
|
| 20 |
+
| Rank / alpha | r=16, alpha=32 |
|
| 21 |
+
| Target modules | `q_proj`, `v_proj` |
|
| 22 |
+
| Eval perplexity | 4.55 |
|
| 23 |
+
| Student | Vũ Văn Học — 2A202600653 |
|
| 24 |
+
|
| 25 |
+
## Usage
|
| 26 |
+
|
| 27 |
+
```python
|
| 28 |
+
from peft import PeftModel
|
| 29 |
+
from unsloth import FastLanguageModel
|
| 30 |
+
|
| 31 |
+
base, tokenizer = FastLanguageModel.from_pretrained(
|
| 32 |
+
"unsloth/Qwen2.5-3B-bnb-4bit", load_in_4bit=True
|
| 33 |
+
)
|
| 34 |
+
model = PeftModel.from_pretrained(base, "2vhoc/lab21")
|
| 35 |
+
```
|
adapter_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen2ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen2.modeling_qwen2",
|
| 8 |
+
"unsloth_fixed": true
|
| 9 |
+
},
|
| 10 |
+
"base_model_name_or_path": "unsloth/Qwen2.5-3B-bnb-4bit",
|
| 11 |
+
"bias": "none",
|
| 12 |
+
"corda_config": null,
|
| 13 |
+
"ensure_weight_tying": false,
|
| 14 |
+
"eva_config": null,
|
| 15 |
+
"exclude_modules": null,
|
| 16 |
+
"fan_in_fan_out": false,
|
| 17 |
+
"inference_mode": true,
|
| 18 |
+
"init_lora_weights": true,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"peft_type": "LORA",
|
| 31 |
+
"peft_version": "0.19.1",
|
| 32 |
+
"qalora_group_size": 16,
|
| 33 |
+
"r": 16,
|
| 34 |
+
"rank_pattern": {},
|
| 35 |
+
"revision": null,
|
| 36 |
+
"target_modules": [
|
| 37 |
+
"q_proj",
|
| 38 |
+
"v_proj"
|
| 39 |
+
],
|
| 40 |
+
"target_parameters": null,
|
| 41 |
+
"task_type": "CAUSAL_LM",
|
| 42 |
+
"trainable_token_indices": null,
|
| 43 |
+
"use_bdlora": null,
|
| 44 |
+
"use_dora": false,
|
| 45 |
+
"use_qalora": false,
|
| 46 |
+
"use_rslora": false
|
| 47 |
+
}
|
adapter_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cdeb2bc03d94abaffd97cbad56ea0e6a7b0cc91177dbaf1e86dfcb4bdbbdd50e
|
| 3 |
+
size 14764792
|