tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

replace references to random 68m model w 135m smollm2 (#2570) [skip ci]

Browse Source 
* replace references to random 68m model w 135m smollm2

* use AutoTokenizer for smollm2
This commit is contained in:
Wing Lian
2025-04-28 10:08:07 -04:00
committed by GitHub
parent f1df73a798
commit 40f4ea23ab
13 changed files with 95 additions and 87 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
tests/e2e/patched/test_llama_s2_attention.py
View File

@@ -31,8 +31,8 @@ class TestLlamaShiftedSparseAttention(unittest.TestCase):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "JackFram/llama-68m",
"tokenizer_type": "LlamaTokenizer",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"tokenizer_type": "AutoTokenizer",
"sequence_len": 16384,
"sample_packing": False,
"flash_attention": True,
@@ -44,7 +44,9 @@ class TestLlamaShiftedSparseAttention(unittest.TestCase):
"lora_dropout": 0.05,
"lora_target_linear": True,
"val_set_size": 0.02,
"special_tokens": {},
"special_tokens": {
"pad_token": "<|endoftext|>",
},
"datasets": [
{
"path": "Yukang/LongAlpaca-12k",
@@ -78,14 +80,16 @@ class TestLlamaShiftedSparseAttention(unittest.TestCase):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "JackFram/llama-68m",
"tokenizer_type": "LlamaTokenizer",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"tokenizer_type": "AutoTokenizer",
"sequence_len": 16384,
"sample_packing": False,
"flash_attention": True,
"s2_attention": True,
"val_set_size": 0.02,
"special_tokens": {},
"special_tokens": {
"pad_token": "<|endoftext|>",
},
"datasets": [
{
"path": "Yukang/LongAlpaca-12k",