danielhanchen
commited on
Commit
•
c67e2ff
1
Parent(s):
33c7b76
Upload config
Browse files- README.md +1 -2
- config.json +10 -5
README.md
CHANGED
@@ -1,15 +1,14 @@
|
|
1 |
---
|
2 |
language:
|
3 |
- en
|
4 |
-
license: apache-2.0
|
5 |
library_name: transformers
|
|
|
6 |
tags:
|
7 |
- unsloth
|
8 |
- transformers
|
9 |
- llama
|
10 |
- llama2
|
11 |
- llama-2
|
12 |
-
|
13 |
---
|
14 |
|
15 |
# Finetune Mistral, Gemma, Llama 2-5x faster with 70% less memory via Unsloth!
|
|
|
1 |
---
|
2 |
language:
|
3 |
- en
|
|
|
4 |
library_name: transformers
|
5 |
+
license: apache-2.0
|
6 |
tags:
|
7 |
- unsloth
|
8 |
- transformers
|
9 |
- llama
|
10 |
- llama2
|
11 |
- llama-2
|
|
|
12 |
---
|
13 |
|
14 |
# Finetune Mistral, Gemma, Llama 2-5x faster with 70% less memory via Unsloth!
|
config.json
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "
|
3 |
"architectures": [
|
4 |
"LlamaForCausalLM"
|
5 |
],
|
@@ -12,14 +12,18 @@
|
|
12 |
"initializer_range": 0.02,
|
13 |
"intermediate_size": 11008,
|
14 |
"max_position_embeddings": 4096,
|
|
|
15 |
"model_type": "llama",
|
16 |
"num_attention_heads": 32,
|
17 |
"num_hidden_layers": 32,
|
18 |
"num_key_value_heads": 32,
|
19 |
-
"pad_token_id":
|
20 |
"pretraining_tp": 1,
|
21 |
"quantization_config": {
|
22 |
-
"
|
|
|
|
|
|
|
23 |
"bnb_4bit_quant_type": "nf4",
|
24 |
"bnb_4bit_use_double_quant": true,
|
25 |
"llm_int8_enable_fp32_cpu_offload": false,
|
@@ -34,8 +38,9 @@
|
|
34 |
"rope_scaling": null,
|
35 |
"rope_theta": 10000.0,
|
36 |
"tie_word_embeddings": false,
|
37 |
-
"torch_dtype": "
|
38 |
-
"transformers_version": "4.
|
|
|
39 |
"use_cache": true,
|
40 |
"vocab_size": 32000
|
41 |
}
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "unsloth/llama-2-7b",
|
3 |
"architectures": [
|
4 |
"LlamaForCausalLM"
|
5 |
],
|
|
|
12 |
"initializer_range": 0.02,
|
13 |
"intermediate_size": 11008,
|
14 |
"max_position_embeddings": 4096,
|
15 |
+
"mlp_bias": false,
|
16 |
"model_type": "llama",
|
17 |
"num_attention_heads": 32,
|
18 |
"num_hidden_layers": 32,
|
19 |
"num_key_value_heads": 32,
|
20 |
+
"pad_token_id": 0,
|
21 |
"pretraining_tp": 1,
|
22 |
"quantization_config": {
|
23 |
+
"_load_in_4bit": true,
|
24 |
+
"_load_in_8bit": false,
|
25 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
26 |
+
"bnb_4bit_quant_storage": "uint8",
|
27 |
"bnb_4bit_quant_type": "nf4",
|
28 |
"bnb_4bit_use_double_quant": true,
|
29 |
"llm_int8_enable_fp32_cpu_offload": false,
|
|
|
38 |
"rope_scaling": null,
|
39 |
"rope_theta": 10000.0,
|
40 |
"tie_word_embeddings": false,
|
41 |
+
"torch_dtype": "bfloat16",
|
42 |
+
"transformers_version": "4.44.0",
|
43 |
+
"unsloth_version": "2024.8",
|
44 |
"use_cache": true,
|
45 |
"vocab_size": 32000
|
46 |
}
|