Upload rl/meta_000499.json with huggingface_hub
Browse files- rl/meta_000499.json +82 -0
rl/meta_000499.json
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_config": {
|
| 3 |
+
"sequence_len": 2048,
|
| 4 |
+
"vocab_size": 65536,
|
| 5 |
+
"n_layer": 32,
|
| 6 |
+
"n_head": 16,
|
| 7 |
+
"n_kv_head": 16,
|
| 8 |
+
"n_embd": 2048,
|
| 9 |
+
"intermediate_size": null,
|
| 10 |
+
"mlp_type": "relu2",
|
| 11 |
+
"rope_theta": 10000.0,
|
| 12 |
+
"attention_bias": false,
|
| 13 |
+
"moe_num_experts": 0,
|
| 14 |
+
"moe_top_k": 1,
|
| 15 |
+
"moe_layer_start": 0,
|
| 16 |
+
"moe_layer_end": -1,
|
| 17 |
+
"moe_layer_stride": 1,
|
| 18 |
+
"moe_capacity_factor": 1.25,
|
| 19 |
+
"moe_aux_loss_coef": 0.01
|
| 20 |
+
},
|
| 21 |
+
"user_config": {
|
| 22 |
+
"run": "r1_grpo_grpo",
|
| 23 |
+
"source": "sft",
|
| 24 |
+
"ref_source": "sft",
|
| 25 |
+
"dtype": "float32",
|
| 26 |
+
"device_batch_size": 1,
|
| 27 |
+
"examples_per_step": 16,
|
| 28 |
+
"num_samples": 4,
|
| 29 |
+
"ppo_minibatch_size": 64,
|
| 30 |
+
"max_prompt_tokens": 1024,
|
| 31 |
+
"max_new_tokens": 2048,
|
| 32 |
+
"temperature": 1.0,
|
| 33 |
+
"top_k": 50,
|
| 34 |
+
"kl_coef": 0.0,
|
| 35 |
+
"kl_max_threshold": 50.0,
|
| 36 |
+
"reward_scale": 1.0,
|
| 37 |
+
"reward_mode": "dapo",
|
| 38 |
+
"group_dynamic_sampling": 0,
|
| 39 |
+
"group_dynamic_sampling_max_tries": 50,
|
| 40 |
+
"use_best_of_n": 0,
|
| 41 |
+
"active_sampling": 1,
|
| 42 |
+
"zero_grad_filtering": 1,
|
| 43 |
+
"zero_adv_eps": 1e-08,
|
| 44 |
+
"format_hint_mode": "eval",
|
| 45 |
+
"learning_rate": 1e-06,
|
| 46 |
+
"lr_schedule": "constant",
|
| 47 |
+
"temp_start": 1.0,
|
| 48 |
+
"temp_end": 1.0,
|
| 49 |
+
"temp_schedule": "none",
|
| 50 |
+
"length_penalty_mode": "dapo",
|
| 51 |
+
"length_penalty_coef": 1.0,
|
| 52 |
+
"length_penalty_target": 2048,
|
| 53 |
+
"length_penalty_floor": 0.0,
|
| 54 |
+
"clip_eps": 0.2,
|
| 55 |
+
"clip_ratio_low": 0.8,
|
| 56 |
+
"clip_ratio_high": 1.28,
|
| 57 |
+
"advantage_clip": 5.0,
|
| 58 |
+
"grpo_epochs": 4,
|
| 59 |
+
"grpo_lr_scale": 1.0,
|
| 60 |
+
"num_steps": 500,
|
| 61 |
+
"total_examples": -1,
|
| 62 |
+
"save_every": 100,
|
| 63 |
+
"eval_every": 50,
|
| 64 |
+
"eval_num_per_task": 5,
|
| 65 |
+
"eval_seed": 123,
|
| 66 |
+
"eval_temperature": 0.0,
|
| 67 |
+
"eval_top_k": 0,
|
| 68 |
+
"eval_max_new_tokens": 256,
|
| 69 |
+
"use_deepspeed": 0,
|
| 70 |
+
"deepspeed_config": "slurm/deepspeed_zero3.json",
|
| 71 |
+
"use_fsdp": 0,
|
| 72 |
+
"fsdp_min_num_params": 1000000,
|
| 73 |
+
"fsdp_cpu_offload": 0,
|
| 74 |
+
"task_mix": "dolci:1.0,gsm8k:0.45,math:0.20,mmlu_science:0.10,mbpp:0.25",
|
| 75 |
+
"dolci_dataset_id": "allenai/Dolci-Think-RL-32B",
|
| 76 |
+
"dolci_split": "train",
|
| 77 |
+
"dolci_mode": "cot",
|
| 78 |
+
"dolci_stop": -1,
|
| 79 |
+
"dolci_streaming": 0,
|
| 80 |
+
"dolci_stream_cache": ""
|
| 81 |
+
}
|
| 82 |
+
}
|