a1af6c326334613271e98dae02c548014f032ff7d4dc9f7faae3ac594b93d3e2
Browse files- README.md +1 -1
- config.json +1 -1
README.md
CHANGED
|
@@ -8,7 +8,7 @@ tags:
|
|
| 8 |
- trl
|
| 9 |
- sft
|
| 10 |
- mlx
|
| 11 |
-
base_model: HuggingFaceTB/SmolLM-135M
|
| 12 |
datasets:
|
| 13 |
- Magpie-Align/Magpie-Pro-300K-Filtered
|
| 14 |
- bigcode/self-oss-instruct-sc2-exec-filter-50k
|
|
|
|
| 8 |
- trl
|
| 9 |
- sft
|
| 10 |
- mlx
|
| 11 |
+
base_model: HuggingFaceTB/SmolLM-135M-Instruct
|
| 12 |
datasets:
|
| 13 |
- Magpie-Align/Magpie-Pro-300K-Filtered
|
| 14 |
- bigcode/self-oss-instruct-sc2-exec-filter-50k
|
config.json
CHANGED
|
@@ -18,7 +18,7 @@
|
|
| 18 |
"num_key_value_heads": 3,
|
| 19 |
"pad_token_id": 2,
|
| 20 |
"pretraining_tp": 1,
|
| 21 |
-
"
|
| 22 |
"group_size": 64,
|
| 23 |
"bits": 4
|
| 24 |
},
|
|
|
|
| 18 |
"num_key_value_heads": 3,
|
| 19 |
"pad_token_id": 2,
|
| 20 |
"pretraining_tp": 1,
|
| 21 |
+
"quantization_config": {
|
| 22 |
"group_size": 64,
|
| 23 |
"bits": 4
|
| 24 |
},
|