JiyangZhang
commited on
Commit
•
f8beda1
1
Parent(s):
314dad7
Add exLong no eTest name
Browse files- adapter_config.json +5 -5
- adapter_model.bin +1 -1
- logs.txt +25 -25
adapter_config.json
CHANGED
@@ -20,13 +20,13 @@
|
|
20 |
"rank_pattern": {},
|
21 |
"revision": null,
|
22 |
"target_modules": [
|
23 |
-
"down_proj",
|
24 |
-
"k_proj",
|
25 |
"o_proj",
|
26 |
-
"
|
27 |
"gate_proj",
|
28 |
-
"
|
29 |
-
"up_proj"
|
|
|
|
|
30 |
],
|
31 |
"task_type": "CAUSAL_LM",
|
32 |
"use_dora": false,
|
|
|
20 |
"rank_pattern": {},
|
21 |
"revision": null,
|
22 |
"target_modules": [
|
|
|
|
|
23 |
"o_proj",
|
24 |
+
"k_proj",
|
25 |
"gate_proj",
|
26 |
+
"v_proj",
|
27 |
+
"up_proj",
|
28 |
+
"down_proj",
|
29 |
+
"q_proj"
|
30 |
],
|
31 |
"task_type": "CAUSAL_LM",
|
32 |
"use_dora": false,
|
adapter_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 319977674
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:783fc5d060218d483d88b91733f99c6da0383ee829f013d2c7e54399af2414c8
|
3 |
size 319977674
|
logs.txt
CHANGED
@@ -5,7 +5,7 @@ tags:
|
|
5 |
- generated_from_trainer
|
6 |
library_name: peft
|
7 |
model-index:
|
8 |
-
- name: work/10283/sarella/ls6/exlong-internal/_work/exp/conditionnestack2e-
|
9 |
results: []
|
10 |
---
|
11 |
|
@@ -23,7 +23,7 @@ base_model_config: codellama/CodeLlama-7b-Instruct-hf
|
|
23 |
bf16: true
|
24 |
dataset_prepared_path: null
|
25 |
datasets:
|
26 |
-
- path: /work/10283/sarella/ls6/exlong-internal/_work/setup/conditionnestack2e-
|
27 |
type:
|
28 |
field_input: input
|
29 |
field_instruction: instruction
|
@@ -63,7 +63,7 @@ micro_batch_size: 4
|
|
63 |
model_type: LlamaForCausalLM
|
64 |
num_epochs: 3
|
65 |
optimizer: adamw_bnb_8bit
|
66 |
-
output_dir: /work/10283/sarella/ls6/exlong-internal/_work/exp/conditionnestack2e-
|
67 |
pad_to_sequence_len: true
|
68 |
resume_from_checkpoint: null
|
69 |
sample_packing: true
|
@@ -92,11 +92,11 @@ xformers_attention: null
|
|
92 |
|
93 |
</details><br>
|
94 |
|
95 |
-
# work/10283/sarella/ls6/exlong-internal/_work/exp/conditionnestack2e-
|
96 |
|
97 |
This model is a fine-tuned version of [codellama/CodeLlama-7b-Instruct-hf](https://huggingface.co/codellama/CodeLlama-7b-Instruct-hf) on the None dataset.
|
98 |
It achieves the following results on the evaluation set:
|
99 |
-
- Loss: 0.
|
100 |
|
101 |
## Model description
|
102 |
|
@@ -130,26 +130,26 @@ The following hyperparameters were used during training:
|
|
130 |
|
131 |
| Training Loss | Epoch | Step | Validation Loss |
|
132 |
|:-------------:|:-----:|:----:|:---------------:|
|
133 |
-
| 0.
|
134 |
-
| 0.
|
135 |
-
| 0.
|
136 |
-
| 0.
|
137 |
-
| 0.
|
138 |
-
| 0.
|
139 |
-
| 0.
|
140 |
-
| 0.
|
141 |
-
| 0.
|
142 |
-
| 0.
|
143 |
-
| 0.
|
144 |
-
| 0.
|
145 |
-
| 0.
|
146 |
-
| 0.
|
147 |
-
| 0.
|
148 |
-
| 0.
|
149 |
-
| 0.
|
150 |
-
| 0.
|
151 |
-
| 0.
|
152 |
-
| 0.
|
153 |
|
154 |
|
155 |
### Framework versions
|
|
|
5 |
- generated_from_trainer
|
6 |
library_name: peft
|
7 |
model-index:
|
8 |
+
- name: work/10283/sarella/ls6/exlong-internal/_work/exp/conditionnestack2e-no-name-ft/lora-codellama-7b-123
|
9 |
results: []
|
10 |
---
|
11 |
|
|
|
23 |
bf16: true
|
24 |
dataset_prepared_path: null
|
25 |
datasets:
|
26 |
+
- path: /work/10283/sarella/ls6/exlong-internal/_work/setup/conditionnestack2e-no-name-ft/train/train/train-conditionnestack2e-no-name-ft.jsonl
|
27 |
type:
|
28 |
field_input: input
|
29 |
field_instruction: instruction
|
|
|
63 |
model_type: LlamaForCausalLM
|
64 |
num_epochs: 3
|
65 |
optimizer: adamw_bnb_8bit
|
66 |
+
output_dir: /work/10283/sarella/ls6/exlong-internal/_work/exp/conditionnestack2e-no-name-ft/lora-codellama-7b-123
|
67 |
pad_to_sequence_len: true
|
68 |
resume_from_checkpoint: null
|
69 |
sample_packing: true
|
|
|
92 |
|
93 |
</details><br>
|
94 |
|
95 |
+
# work/10283/sarella/ls6/exlong-internal/_work/exp/conditionnestack2e-no-name-ft/lora-codellama-7b-123
|
96 |
|
97 |
This model is a fine-tuned version of [codellama/CodeLlama-7b-Instruct-hf](https://huggingface.co/codellama/CodeLlama-7b-Instruct-hf) on the None dataset.
|
98 |
It achieves the following results on the evaluation set:
|
99 |
+
- Loss: 0.4931
|
100 |
|
101 |
## Model description
|
102 |
|
|
|
130 |
|
131 |
| Training Loss | Epoch | Step | Validation Loss |
|
132 |
|:-------------:|:-----:|:----:|:---------------:|
|
133 |
+
| 0.8379 | 0.01 | 1 | 1.0354 |
|
134 |
+
| 0.3779 | 0.16 | 20 | 0.4820 |
|
135 |
+
| 0.3361 | 0.31 | 40 | 0.4560 |
|
136 |
+
| 0.3153 | 0.47 | 60 | 0.4467 |
|
137 |
+
| 0.2735 | 0.63 | 80 | 0.4457 |
|
138 |
+
| 0.2437 | 0.78 | 100 | 0.4400 |
|
139 |
+
| 0.2941 | 0.94 | 120 | 0.4416 |
|
140 |
+
| 0.2153 | 1.08 | 140 | 0.4466 |
|
141 |
+
| 0.2583 | 1.23 | 160 | 0.4499 |
|
142 |
+
| 0.2026 | 1.39 | 180 | 0.4540 |
|
143 |
+
| 0.185 | 1.55 | 200 | 0.4541 |
|
144 |
+
| 0.2296 | 1.7 | 220 | 0.4604 |
|
145 |
+
| 0.2059 | 1.86 | 240 | 0.4591 |
|
146 |
+
| 0.1998 | 2.02 | 260 | 0.4626 |
|
147 |
+
| 0.1879 | 2.15 | 280 | 0.4828 |
|
148 |
+
| 0.1861 | 2.31 | 300 | 0.4944 |
|
149 |
+
| 0.1561 | 2.47 | 320 | 0.4947 |
|
150 |
+
| 0.1888 | 2.62 | 340 | 0.4939 |
|
151 |
+
| 0.1665 | 2.78 | 360 | 0.4945 |
|
152 |
+
| 0.1627 | 2.94 | 380 | 0.4931 |
|
153 |
|
154 |
|
155 |
### Framework versions
|