Instructions to use keepsloading/icml_repro_scratch with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use keepsloading/icml_repro_scratch with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="keepsloading/icml_repro_scratch")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("keepsloading/icml_repro_scratch")
model = AutoModelForCausalLM.from_pretrained("keepsloading/icml_repro_scratch", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use keepsloading/icml_repro_scratch with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "keepsloading/icml_repro_scratch"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "keepsloading/icml_repro_scratch",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/keepsloading/icml_repro_scratch

SGLang

How to use keepsloading/icml_repro_scratch with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "keepsloading/icml_repro_scratch" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "keepsloading/icml_repro_scratch",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "keepsloading/icml_repro_scratch" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "keepsloading/icml_repro_scratch",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use keepsloading/icml_repro_scratch with Docker Model Runner:
```
docker model run hf.co/keepsloading/icml_repro_scratch
```

icml_repro_scratch / outputs /eval /logs /token_t045 /ruler_b.log

keepsloading

Upload folder using huggingface_hub

6a26953 verified about 1 month ago

Raw

History Blame Contribute Delete

24.8 kB

	2026-07-18:15:05:44 WARNING [config.evaluate_config:281] --limit SHOULD ONLY BE USED FOR TESTING. REAL METRICS SHOULD NOT BE COMPUTED USING LIMIT.
	2026-07-18:15:05:47 INFO [_cli.run:376] Selected Tasks: ['niah_single_2', 'niah_multikey_1', 'niah_multikey_3', 'niah_multivalue', 'ruler_cwe', 'ruler_qa_squad']
	2026-07-18:15:05:49 INFO [evaluator:211] Setting random seed to 0 \| Setting numpy seed to 1234 \| Setting torch manual seed to 1234 \| Setting fewshot manual seed to 1234
	2026-07-18:15:05:49 INFO [evaluator:236] Initializing hf model, with arguments: {'pretrained': '/workspace/outputs/l2a_style/stage2/checkpoint-25', 'trust_remote_code': True, 'dtype': 'bfloat16', 'max_length': 16384, 'attn_implementation': 'sdpa'}
	2026-07-18:15:05:52 INFO [models.huggingface:161] Using device 'cuda:0'
	2026-07-18:15:05:52 INFO [models.huggingface:548] Model type cannot be determined. Using default model type 'causal'
	2026-07-18:15:05:52 INFO [models.huggingface:423] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
	2026-07-18:15:06:01 WARNING [api.task:856] niah_single_2: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager.
	For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme.
	2026-07-18:15:06:02 INFO [tasks.ruler.common_utils:26] Using tokenizer /workspace/outputs/l2a_style/stage2/checkpoint-25 for synthetic tasks.

	Generating synthetic samples: essay \| 8192: 0%\| \| 0/500 [00:00<?, ?it/s][A
	Generating synthetic samples: essay \| 8192: 11%\|█ \| 56/500 [00:01<00:08, 55.32it/s][A
	Generating synthetic samples: essay \| 8192: 22%\|██▏ \| 112/500 [00:02<00:07, 55.35it/s][A
	Generating synthetic samples: essay \| 8192: 34%\|███▍ \| 171/500 [00:03<00:05, 56.61it/s][A
	Generating synthetic samples: essay \| 8192: 46%\|████▌ \| 228/500 [00:04<00:04, 56.75it/s][A
	Generating synthetic samples: essay \| 8192: 57%\|█████▋ \| 285/500 [00:05<00:03, 56.56it/s][A
	Generating synthetic samples: essay \| 8192: 68%\|██████▊ \| 342/500 [00:06<00:02, 56.60it/s][A
	Generating synthetic samples: essay \| 8192: 80%\|████████ \| 400/500 [00:07<00:01, 56.94it/s][A
	Generating synthetic samples: essay \| 8192: 91%\|█████████▏\| 457/500 [00:08<00:00, 56.94it/s][A Generating synthetic samples: essay \| 8192: 100%\|██████████\| 500/500 [00:08<00:00, 56.71it/s]
	2026-07-18:15:06:11 WARNING [api.task:856] niah_multikey_1: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager.
	For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme.

	Generating synthetic samples: essay \| 8192: 0%\| \| 0/500 [00:00<?, ?it/s][A
	Generating synthetic samples: essay \| 8192: 11%\|█▏ \| 57/500 [00:01<00:07, 56.93it/s][A
	Generating synthetic samples: essay \| 8192: 23%\|██▎ \| 114/500 [00:02<00:06, 55.84it/s][A
	Generating synthetic samples: essay \| 8192: 34%\|███▍ \| 171/500 [00:03<00:05, 56.29it/s][A
	Generating synthetic samples: essay \| 8192: 46%\|████▌ \| 229/500 [00:04<00:04, 56.59it/s][A
	Generating synthetic samples: essay \| 8192: 57%\|█████▋ \| 287/500 [00:05<00:03, 56.84it/s][A
	Generating synthetic samples: essay \| 8192: 69%\|██████▉ \| 344/500 [00:06<00:02, 56.82it/s][A
	Generating synthetic samples: essay \| 8192: 80%\|████████ \| 401/500 [00:07<00:01, 56.60it/s][A
	Generating synthetic samples: essay \| 8192: 92%\|█████████▏\| 459/500 [00:08<00:00, 56.89it/s][A Generating synthetic samples: essay \| 8192: 100%\|██████████\| 500/500 [00:08<00:00, 56.54it/s]
	2026-07-18:15:06:21 WARNING [api.task:856] niah_multikey_3: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager.
	For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme.

	Generating synthetic samples: needle \| 8192: 0%\| \| 0/500 [00:00<?, ?it/s][A
	Generating synthetic samples: needle \| 8192: 17%\|█▋ \| 85/500 [00:01<00:04, 84.62it/s][A
	Generating synthetic samples: needle \| 8192: 34%\|███▍ \| 170/500 [00:02<00:03, 83.53it/s][A
	Generating synthetic samples: needle \| 8192: 51%\|█████ \| 254/500 [00:03<00:02, 83.25it/s][A
	Generating synthetic samples: needle \| 8192: 68%\|██████▊ \| 338/500 [00:04<00:01, 83.09it/s][A
	Generating synthetic samples: needle \| 8192: 84%\|████████▍ \| 422/500 [00:05<00:00, 83.00it/s][A Generating synthetic samples: needle \| 8192: 100%\|██████████\| 500/500 [00:05<00:00, 83.38it/s]
	2026-07-18:15:06:27 WARNING [api.task:856] niah_multivalue: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager.
	For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme.

	Generating synthetic samples: essay \| 8192: 0%\| \| 0/500 [00:00<?, ?it/s][A
	Generating synthetic samples: essay \| 8192: 11%\|█ \| 55/500 [00:01<00:08, 54.50it/s][A
	Generating synthetic samples: essay \| 8192: 22%\|██▏ \| 111/500 [00:02<00:07, 55.29it/s][A
	Generating synthetic samples: essay \| 8192: 34%\|███▍ \| 169/500 [00:03<00:05, 56.13it/s][A
	Generating synthetic samples: essay \| 8192: 45%\|████▌ \| 226/500 [00:04<00:04, 55.92it/s][A
	Generating synthetic samples: essay \| 8192: 56%\|█████▋ \| 282/500 [00:05<00:03, 55.66it/s][A
	Generating synthetic samples: essay \| 8192: 68%\|██████▊ \| 338/500 [00:06<00:02, 55.69it/s][A
	Generating synthetic samples: essay \| 8192: 79%\|███████▉ \| 396/500 [00:07<00:01, 56.20it/s][A
	Generating synthetic samples: essay \| 8192: 91%\|█████████ \| 454/500 [00:08<00:00, 56.70it/s][A Generating synthetic samples: essay \| 8192: 100%\|██████████\| 500/500 [00:08<00:00, 56.20it/s]
	2026-07-18:15:06:37 WARNING [api.task:856] ruler_cwe: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager.
	For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme.
	2026-07-18:15:06:37 INFO [tasks.ruler.common_utils:26] Using tokenizer /workspace/outputs/l2a_style/stage2/checkpoint-25 for synthetic tasks.

	Generating CWE Samples \| 8192: 0%\| \| 0/500 [00:00<?, ?it/s][A
	Generating CWE Samples \| 8192: 17%\|█▋ \| 83/500 [00:01<00:05, 82.24it/s][A
	Generating CWE Samples \| 8192: 33%\|███▎ \| 166/500 [00:02<00:04, 80.18it/s][A
	Generating CWE Samples \| 8192: 50%\|████▉ \| 249/500 [00:03<00:03, 81.04it/s][A
	Generating CWE Samples \| 8192: 66%\|██████▌ \| 331/500 [00:04<00:02, 81.05it/s][A
	Generating CWE Samples \| 8192: 83%\|████████▎ \| 414/500 [00:05<00:01, 81.42it/s][A
	Generating CWE Samples \| 8192: 99%\|█████████▉\| 496/500 [00:06<00:00, 81.17it/s][A Generating CWE Samples \| 8192: 100%\|██████████\| 500/500 [00:06<00:00, 81.15it/s]
	2026-07-18:15:06:43 WARNING [api.task:856] ruler_qa_squad: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager.
	For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme.

	Generating QA Samples \| 8192: 0%\| \| 0/500 [00:00<?, ?it/s][A
	Generating QA Samples \| 8192: 10%\|▉ \| 49/500 [00:01<00:09, 48.52it/s][A
	Generating QA Samples \| 8192: 20%\|█▉ \| 98/500 [00:02<00:08, 48.02it/s][A
	Generating QA Samples \| 8192: 31%\|███ \| 153/500 [00:03<00:06, 51.04it/s][A
	Generating QA Samples \| 8192: 41%\|████▏ \| 207/500 [00:04<00:05, 51.98it/s][A
	Generating QA Samples \| 8192: 52%\|█████▏ \| 261/500 [00:05<00:04, 52.55it/s][A
	Generating QA Samples \| 8192: 63%\|██████▎ \| 316/500 [00:06<00:03, 53.26it/s][A
	Generating QA Samples \| 8192: 74%\|███████▍ \| 371/500 [00:07<00:02, 53.56it/s][A
	Generating QA Samples \| 8192: 85%\|████████▌ \| 427/500 [00:08<00:01, 54.01it/s][A
	Generating QA Samples \| 8192: 96%\|█████████▋\| 482/500 [00:09<00:00, 51.37it/s][A Generating QA Samples \| 8192: 100%\|██████████\| 500/500 [00:10<00:00, 49.91it/s]
	2026-07-18:15:06:54 INFO [tasks:700] Selected tasks:
	2026-07-18:15:06:54 INFO [tasks:691] Task: ruler_qa_squad (ruler/qa_squad.yaml)
	2026-07-18:15:06:54 INFO [tasks:691] Task: ruler_cwe (ruler/cwe.yaml)
	2026-07-18:15:06:54 INFO [tasks:691] Task: niah_multivalue (ruler/niah_multivalue.yaml)
	2026-07-18:15:06:54 INFO [tasks:691] Task: niah_multikey_3 (ruler/niah_multikey_3.yaml)
	2026-07-18:15:06:54 INFO [tasks:691] Task: niah_multikey_1 (ruler/niah_multikey_1.yaml)
	2026-07-18:15:06:54 INFO [tasks:691] Task: niah_single_2 (ruler/niah_single_2.yaml)
	2026-07-18:15:06:54 INFO [evaluator:314] ruler_qa_squad: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 32, 'until': []}
	2026-07-18:15:06:54 INFO [evaluator:314] ruler_cwe: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 120, 'until': []}
	2026-07-18:15:06:54 INFO [evaluator:314] niah_multivalue: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 128, 'until': []}
	2026-07-18:15:06:54 INFO [evaluator:314] niah_multikey_3: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 128, 'until': []}
	2026-07-18:15:06:54 INFO [evaluator:314] niah_multikey_1: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 128, 'until': []}
	2026-07-18:15:06:54 INFO [evaluator:314] niah_single_2: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 128, 'until': []}
	2026-07-18:15:06:54 INFO [api.task:311] Building contexts for ruler_qa_squad on rank 0...

	0%\| \| 0/20 [00:00<?, ?it/s][A 100%\|██████████\| 20/20 [00:00<00:00, 1116.84it/s]
	2026-07-18:15:06:54 INFO [api.task:311] Building contexts for ruler_cwe on rank 0...

	0%\| \| 0/20 [00:00<?, ?it/s][A 100%\|██████████\| 20/20 [00:00<00:00, 1091.23it/s]
	2026-07-18:15:06:54 INFO [api.task:311] Building contexts for niah_multivalue on rank 0...

	0%\| \| 0/20 [00:00<?, ?it/s][A 100%\|██████████\| 20/20 [00:00<00:00, 1138.23it/s]
	2026-07-18:15:06:54 INFO [api.task:311] Building contexts for niah_multikey_3 on rank 0...

	0%\| \| 0/20 [00:00<?, ?it/s][A 100%\|██████████\| 20/20 [00:00<00:00, 1182.68it/s]
	2026-07-18:15:06:54 INFO [api.task:311] Building contexts for niah_multikey_1 on rank 0...

	0%\| \| 0/20 [00:00<?, ?it/s][A 100%\|██████████\| 20/20 [00:00<00:00, 1157.43it/s]
	2026-07-18:15:06:54 INFO [api.task:311] Building contexts for niah_single_2 on rank 0...

	0%\| \| 0/20 [00:00<?, ?it/s][A 100%\|██████████\| 20/20 [00:00<00:00, 1155.93it/s]
	2026-07-18:15:06:54 INFO [evaluator:584] Running generate_until requests

	Running generate_until requests: 0%\| \| 0/120 [00:00<?, ?it/s][A
	Running generate_until requests: 1%\| \| 1/120 [00:02<05:37, 2.84s/it][A
	Running generate_until requests: 2%\|▏ \| 2/120 [00:04<04:28, 2.28s/it][A
	Running generate_until requests: 2%\|▎ \| 3/120 [00:06<04:05, 2.10s/it][A
	Running generate_until requests: 3%\|▎ \| 4/120 [00:08<03:53, 2.02s/it][A
	Running generate_until requests: 4%\|▍ \| 5/120 [00:10<03:45, 1.97s/it][A
	Running generate_until requests: 5%\|▌ \| 6/120 [00:12<03:40, 1.93s/it][A
	Running generate_until requests: 6%\|▌ \| 7/120 [00:14<03:36, 1.91s/it][A
	Running generate_until requests: 7%\|▋ \| 8/120 [00:15<03:32, 1.90s/it][A
	Running generate_until requests: 8%\|▊ \| 9/120 [00:17<03:29, 1.89s/it][A
	Running generate_until requests: 8%\|▊ \| 10/120 [00:19<03:26, 1.88s/it][A
	Running generate_until requests: 9%\|▉ \| 11/120 [00:21<03:23, 1.87s/it][A
	Running generate_until requests: 10%\|█ \| 12/120 [00:23<03:21, 1.86s/it][A
	Running generate_until requests: 11%\|█ \| 13/120 [00:25<03:18, 1.86s/it][A
	Running generate_until requests: 12%\|█▏ \| 14/120 [00:27<03:16, 1.86s/it][A
	Running generate_until requests: 12%\|█▎ \| 15/120 [00:28<03:15, 1.86s/it][A
	Running generate_until requests: 13%\|█▎ \| 16/120 [00:30<03:13, 1.86s/it][A
	Running generate_until requests: 14%\|█▍ \| 17/120 [00:32<03:11, 1.85s/it][A
	Running generate_until requests: 15%\|█▌ \| 18/120 [00:34<03:08, 1.85s/it][A
	Running generate_until requests: 16%\|█▌ \| 19/120 [00:36<03:03, 1.81s/it][A
	Running generate_until requests: 17%\|█▋ \| 20/120 [00:37<02:58, 1.78s/it][A
	Running generate_until requests: 18%\|█▊ \| 21/120 [00:44<05:04, 3.08s/it][A
	Running generate_until requests: 18%\|█▊ \| 22/120 [00:49<06:21, 3.89s/it][A
	Running generate_until requests: 19%\|█▉ \| 23/120 [00:55<07:11, 4.45s/it][A
	Running generate_until requests: 20%\|██ \| 24/120 [01:01<07:46, 4.85s/it][A
	Running generate_until requests: 21%\|██ \| 25/120 [01:07<08:07, 5.13s/it][A
	Running generate_until requests: 22%\|██▏ \| 26/120 [01:12<08:21, 5.33s/it][A
	Running generate_until requests: 22%\|██▎ \| 27/120 [01:18<08:28, 5.47s/it][A
	Running generate_until requests: 23%\|██▎ \| 28/120 [01:24<08:31, 5.56s/it][A
	Running generate_until requests: 24%\|██▍ \| 29/120 [01:30<08:33, 5.65s/it][A
	Running generate_until requests: 25%\|██▌ \| 30/120 [01:36<08:32, 5.69s/it][A
	Running generate_until requests: 26%\|██▌ \| 31/120 [01:41<08:29, 5.72s/it][A
	Running generate_until requests: 27%\|██▋ \| 32/120 [01:47<08:25, 5.75s/it][A
	Running generate_until requests: 28%\|██▊ \| 33/120 [01:53<08:20, 5.76s/it][A
	Running generate_until requests: 28%\|██▊ \| 34/120 [01:59<08:15, 5.77s/it][A
	Running generate_until requests: 29%\|██▉ \| 35/120 [02:05<08:10, 5.77s/it][A
	Running generate_until requests: 30%\|███ \| 36/120 [02:10<08:04, 5.77s/it][A
	Running generate_until requests: 31%\|███ \| 37/120 [02:16<07:59, 5.77s/it][A
	Running generate_until requests: 32%\|███▏ \| 38/120 [02:22<07:53, 5.77s/it][A
	Running generate_until requests: 32%\|███▎ \| 39/120 [02:28<07:47, 5.77s/it][A
	Running generate_until requests: 33%\|███▎ \| 40/120 [02:34<07:42, 5.78s/it][A
	Running generate_until requests: 34%\|███▍ \| 41/120 [02:41<08:17, 6.30s/it][A
	Running generate_until requests: 35%\|███▌ \| 42/120 [02:47<08:09, 6.28s/it][A
	Running generate_until requests: 36%\|███▌ \| 43/120 [02:54<08:02, 6.27s/it][A
	Running generate_until requests: 37%\|███▋ \| 44/120 [03:00<07:56, 6.26s/it][A
	Running generate_until requests: 38%\|███▊ \| 45/120 [03:06<07:49, 6.26s/it][A
	Running generate_until requests: 38%\|███▊ \| 46/120 [03:12<07:42, 6.25s/it][A
	Running generate_until requests: 39%\|███▉ \| 47/120 [03:18<07:36, 6.25s/it][A
	Running generate_until requests: 40%\|████ \| 48/120 [03:25<07:29, 6.24s/it][A
	Running generate_until requests: 41%\|████ \| 49/120 [03:31<07:23, 6.24s/it][A
	Running generate_until requests: 42%\|████▏ \| 50/120 [03:37<07:16, 6.24s/it][A
	Running generate_until requests: 42%\|████▎ \| 51/120 [03:43<07:11, 6.26s/it][A
	Running generate_until requests: 43%\|████▎ \| 52/120 [03:50<07:06, 6.28s/it][A
	Running generate_until requests: 44%\|████▍ \| 53/120 [03:56<07:00, 6.27s/it][A
	Running generate_until requests: 45%\|████▌ \| 54/120 [04:02<06:54, 6.27s/it][A
	Running generate_until requests: 46%\|████▌ \| 55/120 [04:09<06:47, 6.28s/it][A
	Running generate_until requests: 47%\|████▋ \| 56/120 [04:15<06:41, 6.27s/it][A
	Running generate_until requests: 48%\|████▊ \| 57/120 [04:21<06:34, 6.27s/it][A
	Running generate_until requests: 48%\|████▊ \| 58/120 [04:28<06:32, 6.34s/it][A
	Running generate_until requests: 49%\|████▉ \| 59/120 [04:34<06:26, 6.34s/it][A
	Running generate_until requests: 50%\|█████ \| 60/120 [04:40<06:20, 6.34s/it][A
	Running generate_until requests: 51%\|█████ \| 61/120 [04:47<06:13, 6.33s/it][A
	Running generate_until requests: 52%\|█████▏ \| 62/120 [04:53<06:06, 6.32s/it][A
	Running generate_until requests: 52%\|█████▎ \| 63/120 [04:59<05:59, 6.30s/it][A
	Running generate_until requests: 53%\|█████▎ \| 64/120 [05:05<05:51, 6.28s/it][A
	Running generate_until requests: 54%\|█████▍ \| 65/120 [05:12<05:44, 6.26s/it][A
	Running generate_until requests: 55%\|█████▌ \| 66/120 [05:18<05:37, 6.25s/it][A
	Running generate_until requests: 56%\|█████▌ \| 67/120 [05:24<05:30, 6.24s/it][A
	Running generate_until requests: 57%\|█████▋ \| 68/120 [05:30<05:24, 6.23s/it][A
	Running generate_until requests: 57%\|█████▊ \| 69/120 [05:37<05:17, 6.23s/it][A
	Running generate_until requests: 58%\|█████▊ \| 70/120 [05:43<05:11, 6.22s/it][A
	Running generate_until requests: 59%\|█████▉ \| 71/120 [05:49<05:05, 6.23s/it][A
	Running generate_until requests: 60%\|██████ \| 72/120 [05:55<04:59, 6.24s/it][A
	Running generate_until requests: 61%\|██████ \| 73/120 [06:01<04:52, 6.22s/it][A
	Running generate_until requests: 62%\|██████▏ \| 74/120 [06:08<04:45, 6.20s/it][A
	Running generate_until requests: 62%\|██████▎ \| 75/120 [06:14<04:38, 6.19s/it][A
	Running generate_until requests: 63%\|██████▎ \| 76/120 [06:20<04:32, 6.19s/it][A
	Running generate_until requests: 64%\|██████▍ \| 77/120 [06:26<04:25, 6.18s/it][A
	Running generate_until requests: 65%\|██████▌ \| 78/120 [06:32<04:19, 6.18s/it][A
	Running generate_until requests: 66%\|██████▌ \| 79/120 [06:38<04:13, 6.17s/it][A
	Running generate_until requests: 67%\|██████▋ \| 80/120 [06:45<04:06, 6.17s/it][A
	Running generate_until requests: 68%\|██████▊ \| 81/120 [06:51<04:01, 6.18s/it][A
	Running generate_until requests: 68%\|██████▊ \| 82/120 [06:57<03:55, 6.19s/it][A
	Running generate_until requests: 69%\|██████▉ \| 83/120 [07:03<03:49, 6.20s/it][A
	Running generate_until requests: 70%\|███████ \| 84/120 [07:09<03:43, 6.21s/it][A
	Running generate_until requests: 71%\|███████ \| 85/120 [07:16<03:37, 6.21s/it][A
	Running generate_until requests: 72%\|███████▏ \| 86/120 [07:22<03:31, 6.21s/it][A
	Running generate_until requests: 72%\|███████▎ \| 87/120 [07:28<03:25, 6.22s/it][A
	Running generate_until requests: 73%\|███████▎ \| 88/120 [07:34<03:18, 6.22s/it][A
	Running generate_until requests: 74%\|███████▍ \| 89/120 [07:41<03:12, 6.22s/it][A
	Running generate_until requests: 75%\|███████▌ \| 90/120 [07:47<03:06, 6.22s/it][A
	Running generate_until requests: 76%\|███████▌ \| 91/120 [07:53<03:00, 6.22s/it][A
	Running generate_until requests: 77%\|███████▋ \| 92/120 [07:59<02:54, 6.22s/it][A
	Running generate_until requests: 78%\|███████▊ \| 93/120 [08:05<02:47, 6.22s/it][A
	Running generate_until requests: 78%\|███████▊ \| 94/120 [08:12<02:42, 6.23s/it][A
	Running generate_until requests: 79%\|███████▉ \| 95/120 [08:18<02:35, 6.22s/it][A
	Running generate_until requests: 80%\|████████ \| 96/120 [08:24<02:29, 6.22s/it][A
	Running generate_until requests: 81%\|████████ \| 97/120 [08:30<02:22, 6.22s/it][A
	Running generate_until requests: 82%\|████████▏ \| 98/120 [08:37<02:16, 6.21s/it][A
	Running generate_until requests: 82%\|████████▎ \| 99/120 [08:43<02:10, 6.22s/it][A
	Running generate_until requests: 83%\|████████▎ \| 100/120 [08:49<02:04, 6.22s/it][A
	Running generate_until requests: 84%\|████████▍ \| 101/120 [08:55<01:58, 6.21s/it][A
	Running generate_until requests: 85%\|████████▌ \| 102/120 [09:01<01:51, 6.21s/it][A
	Running generate_until requests: 86%\|████████▌ \| 103/120 [09:08<01:45, 6.21s/it][A
	Running generate_until requests: 87%\|████████▋ \| 104/120 [09:14<01:39, 6.21s/it][A
	Running generate_until requests: 88%\|████████▊ \| 105/120 [09:20<01:33, 6.21s/it][A
	Running generate_until requests: 88%\|████████▊ \| 106/120 [09:26<01:27, 6.22s/it][A
	Running generate_until requests: 89%\|████████▉ \| 107/120 [09:32<01:20, 6.21s/it][A
	Running generate_until requests: 90%\|█████████ \| 108/120 [09:39<01:14, 6.22s/it][A
	Running generate_until requests: 91%\|█████████ \| 109/120 [09:45<01:08, 6.21s/it][A
	Running generate_until requests: 92%\|█████████▏\| 110/120 [09:51<01:02, 6.21s/it][A
	Running generate_until requests: 92%\|█████████▎\| 111/120 [09:57<00:55, 6.21s/it][A
	Running generate_until requests: 93%\|█████████▎\| 112/120 [10:04<00:49, 6.21s/it][A
	Running generate_until requests: 94%\|█████████▍\| 113/120 [10:10<00:43, 6.21s/it][A
	Running generate_until requests: 95%\|█████████▌\| 114/120 [10:16<00:37, 6.21s/it][A
	Running generate_until requests: 96%\|█████████▌\| 115/120 [10:22<00:31, 6.21s/it][A
	Running generate_until requests: 97%\|█████████▋\| 116/120 [10:28<00:24, 6.21s/it][A
	Running generate_until requests: 98%\|█████████▊\| 117/120 [10:35<00:18, 6.21s/it][A
	Running generate_until requests: 98%\|█████████▊\| 118/120 [10:41<00:12, 6.21s/it][A
	Running generate_until requests: 99%\|█████████▉\| 119/120 [10:47<00:06, 6.21s/it][A
	Running generate_until requests: 100%\|██████████\| 120/120 [10:53<00:00, 6.21s/it][A Running generate_until requests: 100%\|██████████\| 120/120 [10:53<00:00, 5.45s/it]
	2026-07-18:15:17:52 INFO [loggers.evaluation_tracker:247] Saving results aggregated
	2026-07-18:15:17:52 INFO [loggers.evaluation_tracker:119] Saving per-task samples to /workspace/outputs/eval/token_t045/ruler8k_splits/b/lm_eval/__workspace__outputs__l2a_style__stage2__checkpoint-25/*.jsonl
	hf ({'pretrained': '/workspace/outputs/l2a_style/stage2/checkpoint-25', 'dtype': 'bfloat16', 'max_length': 16384, 'attn_implementation': 'sdpa'}), gen_kwargs: ({}), limit: 20.0, num_fewshot: None, batch_size: 1
	\| Tasks \|Version\|Filter\|n-shot\|Metric\| \| Value \| \|Stderr\|
	\|---------------\|------:\|------\|-----:\|-----:\|---\|------:\|---\|------\|
	\|niah_multikey_1\| 1\|none \| 0\| 4096\|↑ \|-1.0000\|± \| N/A\|
	\| \| \|none \| 0\| 8192\|↑ \| 1.0000\|± \| N/A\|
	\|niah_multikey_3\| 1\|none \| 0\| 4096\|↑ \|-1.0000\|± \| N/A\|
	\| \| \|none \| 0\| 8192\|↑ \| 0.7500\|± \| N/A\|
	\|niah_multivalue\| 1\|none \| 0\| 4096\|↑ \|-1.0000\|± \| N/A\|
	\| \| \|none \| 0\| 8192\|↑ \| 1.0000\|± \| N/A\|
	\|niah_single_2 \| 1\|none \| 0\| 4096\|↑ \|-1.0000\|± \| N/A\|
	\| \| \|none \| 0\| 8192\|↑ \| 1.0000\|± \| N/A\|
	\|ruler_cwe \| 1\|none \| 0\| 4096\|↑ \|-1.0000\|± \| N/A\|
	\| \| \|none \| 0\| 8192\|↑ \| 0.3950\|± \| N/A\|
	\|ruler_qa_squad \| 1\|none \| 0\| 4096\|↑ \|-1.0000\|± \| N/A\|
	\| \| \|none \| 0\| 8192\|↑ \| 0.4208\|± \| N/A\|