Upload reap_args.yaml with huggingface_hub
Browse files- reap_args.yaml +76 -0
reap_args.yaml
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
cluster_args:
|
| 2 |
+
cluster_description: null
|
| 3 |
+
cluster_method: agglomerative
|
| 4 |
+
compression_ratio: 0.8
|
| 5 |
+
expert_sim: ttm
|
| 6 |
+
frequency_penalty: true
|
| 7 |
+
linkage_method: average
|
| 8 |
+
max_cluster_size: null
|
| 9 |
+
multi_layer: null
|
| 10 |
+
num_clusters: null
|
| 11 |
+
singleton_outlier_experts: false
|
| 12 |
+
singleton_super_experts: false
|
| 13 |
+
softmax_temperature: null
|
| 14 |
+
ds_args:
|
| 15 |
+
dataset_config_name: all
|
| 16 |
+
dataset_name: /workspace/PTBR-REAP-Pruning
|
| 17 |
+
dataset_test_split: test
|
| 18 |
+
shuffle: true
|
| 19 |
+
split: train
|
| 20 |
+
eval_args:
|
| 21 |
+
evalplus_tasks:
|
| 22 |
+
- mbpp
|
| 23 |
+
- humaneval
|
| 24 |
+
greedy: true
|
| 25 |
+
lm_eval_tasks:
|
| 26 |
+
- winogrande
|
| 27 |
+
- arc_challenge
|
| 28 |
+
- arc_easy
|
| 29 |
+
- boolq
|
| 30 |
+
- hellaswag
|
| 31 |
+
- mmlu
|
| 32 |
+
- openbookqa
|
| 33 |
+
- rte
|
| 34 |
+
min_p: 0.0
|
| 35 |
+
parallel_tasks: 32
|
| 36 |
+
results_dir: null
|
| 37 |
+
run_evalplus: true
|
| 38 |
+
run_livecodebench: true
|
| 39 |
+
run_lm_eval: true
|
| 40 |
+
run_math: false
|
| 41 |
+
run_wildbench: false
|
| 42 |
+
server_log_file_name: pruning-cli-0.log
|
| 43 |
+
temperature: 0.7
|
| 44 |
+
top_k: 20
|
| 45 |
+
top_p: 0.8
|
| 46 |
+
use_server: true
|
| 47 |
+
vllm_port: 8000
|
| 48 |
+
model_args:
|
| 49 |
+
model_name: /workspace/.hf_home/hub/models--Qwen--Qwen3-30B-A3B-Instruct-2507/snapshots/0d7cf23991f47feeb3a57ecb4c9cee8ea4a17bfe
|
| 50 |
+
num_experts_per_tok_override: null
|
| 51 |
+
obs_args:
|
| 52 |
+
distance_measure: cosine
|
| 53 |
+
model_max_length: 2048
|
| 54 |
+
output_file_name: observations_10_cosine-seed_42.pt
|
| 55 |
+
overwrite_observations: false
|
| 56 |
+
record_pruning_metrics_only: true
|
| 57 |
+
renormalize_router_weights: false
|
| 58 |
+
return_vllm_tokens_prompt: false
|
| 59 |
+
samples_per_category: 10
|
| 60 |
+
select_only_categories: null
|
| 61 |
+
split_by_category: false
|
| 62 |
+
truncate: false
|
| 63 |
+
prune_args:
|
| 64 |
+
n_experts_to_prune: null
|
| 65 |
+
overwrite_pruned_model: false
|
| 66 |
+
perserve_outliers: false
|
| 67 |
+
perserve_super_experts: false
|
| 68 |
+
prune_method: reap
|
| 69 |
+
reap_args:
|
| 70 |
+
debug: false
|
| 71 |
+
do_eval: false
|
| 72 |
+
plot_clusters: true
|
| 73 |
+
profile: false
|
| 74 |
+
run_observer_only: false
|
| 75 |
+
seed: 42
|
| 76 |
+
smoke_test: true
|