V3.2: Stage2R iter167 weights + rp=1.005 release generation_config + eval preds (dev 28.56/17.86, devtest 29.21/17.51)
Browse files- eval/pred_haidass-v32s2r-iter167_flores_dev.genconfig.json +79 -0
- eval/pred_haidass-v32s2r-iter167_flores_dev.jsonl +0 -0
- eval/pred_haidass-v32s2r-iter167_floresplus_devtest.genconfig.json +79 -0
- eval/pred_haidass-v32s2r-iter167_floresplus_devtest.jsonl +0 -0
- eval/pred_haidass-v32s2r-iter167_rp1005_flores_dev.genconfig.json +79 -0
- eval/pred_haidass-v32s2r-iter167_rp1005_flores_dev.jsonl +0 -0
- eval/pred_haidass-v32s2r-iter167_rp1005_floresplus_devtest.genconfig.json +79 -0
- eval/pred_haidass-v32s2r-iter167_rp1005_floresplus_devtest.jsonl +0 -0
- generation_config.json +2 -0
- model-00001-of-000030.safetensors +1 -1
- model-00002-of-000030.safetensors +1 -1
- model-00003-of-000030.safetensors +1 -1
- model-00004-of-000030.safetensors +1 -1
- model-00005-of-000030.safetensors +1 -1
- model-00006-of-000030.safetensors +1 -1
- model-00007-of-000030.safetensors +1 -1
- model-00008-of-000030.safetensors +1 -1
- model-00009-of-000030.safetensors +1 -1
- model-00010-of-000030.safetensors +1 -1
- model-00011-of-000030.safetensors +1 -1
- model-00012-of-000030.safetensors +1 -1
- model-00013-of-000030.safetensors +1 -1
- model-00014-of-000030.safetensors +1 -1
- model-00015-of-000030.safetensors +1 -1
- model-00016-of-000030.safetensors +1 -1
- model-00017-of-000030.safetensors +1 -1
- model-00018-of-000030.safetensors +1 -1
- model-00019-of-000030.safetensors +1 -1
- model-00020-of-000030.safetensors +1 -1
- model-00021-of-000030.safetensors +1 -1
- model-00022-of-000030.safetensors +1 -1
- model-00023-of-000030.safetensors +1 -1
- model-00024-of-000030.safetensors +1 -1
- model-00025-of-000030.safetensors +1 -1
- model-00026-of-000030.safetensors +1 -1
- model-00027-of-000030.safetensors +1 -1
- model-00028-of-000030.safetensors +1 -1
- model-00029-of-000030.safetensors +1 -1
- model-00030-of-000030.safetensors +1 -1
eval/pred_haidass-v32s2r-iter167_flores_dev.genconfig.json
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"max_length": 20,
|
| 3 |
+
"max_new_tokens": 512,
|
| 4 |
+
"min_length": 0,
|
| 5 |
+
"min_new_tokens": 0,
|
| 6 |
+
"early_stopping": false,
|
| 7 |
+
"max_time": null,
|
| 8 |
+
"stop_strings": null,
|
| 9 |
+
"do_sample": false,
|
| 10 |
+
"num_beams": 1,
|
| 11 |
+
"use_cache": true,
|
| 12 |
+
"cache_implementation": null,
|
| 13 |
+
"cache_config": null,
|
| 14 |
+
"return_legacy_cache": null,
|
| 15 |
+
"prefill_chunk_size": null,
|
| 16 |
+
"temperature": 1.0,
|
| 17 |
+
"top_k": 50,
|
| 18 |
+
"top_p": 1.0,
|
| 19 |
+
"min_p": null,
|
| 20 |
+
"typical_p": 1.0,
|
| 21 |
+
"epsilon_cutoff": 0.0,
|
| 22 |
+
"eta_cutoff": 0.0,
|
| 23 |
+
"repetition_penalty": 1.0,
|
| 24 |
+
"encoder_repetition_penalty": 1.0,
|
| 25 |
+
"length_penalty": 1.0,
|
| 26 |
+
"no_repeat_ngram_size": 0,
|
| 27 |
+
"bad_words_ids": null,
|
| 28 |
+
"renormalize_logits": false,
|
| 29 |
+
"forced_bos_token_id": null,
|
| 30 |
+
"forced_eos_token_id": null,
|
| 31 |
+
"remove_invalid_values": false,
|
| 32 |
+
"exponential_decay_length_penalty": null,
|
| 33 |
+
"suppress_tokens": null,
|
| 34 |
+
"begin_suppress_tokens": null,
|
| 35 |
+
"sequence_bias": null,
|
| 36 |
+
"token_healing": false,
|
| 37 |
+
"guidance_scale": null,
|
| 38 |
+
"watermarking_config": null,
|
| 39 |
+
"num_return_sequences": 1,
|
| 40 |
+
"output_attentions": false,
|
| 41 |
+
"output_hidden_states": false,
|
| 42 |
+
"output_scores": false,
|
| 43 |
+
"output_logits": null,
|
| 44 |
+
"return_dict_in_generate": false,
|
| 45 |
+
"pad_token_id": 5,
|
| 46 |
+
"bos_token_id": 3,
|
| 47 |
+
"eos_token_id": 5,
|
| 48 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 49 |
+
"decoder_start_token_id": null,
|
| 50 |
+
"is_assistant": false,
|
| 51 |
+
"num_assistant_tokens": 20,
|
| 52 |
+
"num_assistant_tokens_schedule": "constant",
|
| 53 |
+
"assistant_confidence_threshold": 0.4,
|
| 54 |
+
"prompt_lookup_num_tokens": null,
|
| 55 |
+
"max_matching_ngram_size": null,
|
| 56 |
+
"assistant_early_exit": null,
|
| 57 |
+
"assistant_lookbehind": 10,
|
| 58 |
+
"target_lookbehind": 10,
|
| 59 |
+
"disable_compile": false,
|
| 60 |
+
"low_memory": null,
|
| 61 |
+
"penalty_alpha": null,
|
| 62 |
+
"dola_layers": null,
|
| 63 |
+
"diversity_penalty": 0.0,
|
| 64 |
+
"num_beam_groups": 1,
|
| 65 |
+
"constraints": null,
|
| 66 |
+
"force_words_ids": null,
|
| 67 |
+
"_from_model_config": true,
|
| 68 |
+
"transformers_version": "4.57.1",
|
| 69 |
+
"_meta": {
|
| 70 |
+
"model": "/mnt/models/CODE/mzy/v32_ckpts/hf_s2r_iter167",
|
| 71 |
+
"greedy": true,
|
| 72 |
+
"sampling_params_inert": "do_sample=False, temperature/top_p/top_k 不参与生成",
|
| 73 |
+
"prompt_mode": "chat_template",
|
| 74 |
+
"tokenizer_pad": "<|im_end|>",
|
| 75 |
+
"tokenizer_eos": "<|im_end|>",
|
| 76 |
+
"transformers": "4.57.1",
|
| 77 |
+
"torch": "2.7.1+cpu"
|
| 78 |
+
}
|
| 79 |
+
}
|
eval/pred_haidass-v32s2r-iter167_flores_dev.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
eval/pred_haidass-v32s2r-iter167_floresplus_devtest.genconfig.json
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"max_length": 20,
|
| 3 |
+
"max_new_tokens": 512,
|
| 4 |
+
"min_length": 0,
|
| 5 |
+
"min_new_tokens": 0,
|
| 6 |
+
"early_stopping": false,
|
| 7 |
+
"max_time": null,
|
| 8 |
+
"stop_strings": null,
|
| 9 |
+
"do_sample": false,
|
| 10 |
+
"num_beams": 1,
|
| 11 |
+
"use_cache": true,
|
| 12 |
+
"cache_implementation": null,
|
| 13 |
+
"cache_config": null,
|
| 14 |
+
"return_legacy_cache": null,
|
| 15 |
+
"prefill_chunk_size": null,
|
| 16 |
+
"temperature": 1.0,
|
| 17 |
+
"top_k": 50,
|
| 18 |
+
"top_p": 1.0,
|
| 19 |
+
"min_p": null,
|
| 20 |
+
"typical_p": 1.0,
|
| 21 |
+
"epsilon_cutoff": 0.0,
|
| 22 |
+
"eta_cutoff": 0.0,
|
| 23 |
+
"repetition_penalty": 1.0,
|
| 24 |
+
"encoder_repetition_penalty": 1.0,
|
| 25 |
+
"length_penalty": 1.0,
|
| 26 |
+
"no_repeat_ngram_size": 0,
|
| 27 |
+
"bad_words_ids": null,
|
| 28 |
+
"renormalize_logits": false,
|
| 29 |
+
"forced_bos_token_id": null,
|
| 30 |
+
"forced_eos_token_id": null,
|
| 31 |
+
"remove_invalid_values": false,
|
| 32 |
+
"exponential_decay_length_penalty": null,
|
| 33 |
+
"suppress_tokens": null,
|
| 34 |
+
"begin_suppress_tokens": null,
|
| 35 |
+
"sequence_bias": null,
|
| 36 |
+
"token_healing": false,
|
| 37 |
+
"guidance_scale": null,
|
| 38 |
+
"watermarking_config": null,
|
| 39 |
+
"num_return_sequences": 1,
|
| 40 |
+
"output_attentions": false,
|
| 41 |
+
"output_hidden_states": false,
|
| 42 |
+
"output_scores": false,
|
| 43 |
+
"output_logits": null,
|
| 44 |
+
"return_dict_in_generate": false,
|
| 45 |
+
"pad_token_id": 5,
|
| 46 |
+
"bos_token_id": 3,
|
| 47 |
+
"eos_token_id": 5,
|
| 48 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 49 |
+
"decoder_start_token_id": null,
|
| 50 |
+
"is_assistant": false,
|
| 51 |
+
"num_assistant_tokens": 20,
|
| 52 |
+
"num_assistant_tokens_schedule": "constant",
|
| 53 |
+
"assistant_confidence_threshold": 0.4,
|
| 54 |
+
"prompt_lookup_num_tokens": null,
|
| 55 |
+
"max_matching_ngram_size": null,
|
| 56 |
+
"assistant_early_exit": null,
|
| 57 |
+
"assistant_lookbehind": 10,
|
| 58 |
+
"target_lookbehind": 10,
|
| 59 |
+
"disable_compile": false,
|
| 60 |
+
"low_memory": null,
|
| 61 |
+
"penalty_alpha": null,
|
| 62 |
+
"dola_layers": null,
|
| 63 |
+
"diversity_penalty": 0.0,
|
| 64 |
+
"num_beam_groups": 1,
|
| 65 |
+
"constraints": null,
|
| 66 |
+
"force_words_ids": null,
|
| 67 |
+
"_from_model_config": true,
|
| 68 |
+
"transformers_version": "4.57.1",
|
| 69 |
+
"_meta": {
|
| 70 |
+
"model": "/mnt/models/CODE/mzy/v32_ckpts/hf_s2r_iter167",
|
| 71 |
+
"greedy": true,
|
| 72 |
+
"sampling_params_inert": "do_sample=False, temperature/top_p/top_k 不参与生成",
|
| 73 |
+
"prompt_mode": "chat_template",
|
| 74 |
+
"tokenizer_pad": "<|im_end|>",
|
| 75 |
+
"tokenizer_eos": "<|im_end|>",
|
| 76 |
+
"transformers": "4.57.1",
|
| 77 |
+
"torch": "2.7.1+cpu"
|
| 78 |
+
}
|
| 79 |
+
}
|
eval/pred_haidass-v32s2r-iter167_floresplus_devtest.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
eval/pred_haidass-v32s2r-iter167_rp1005_flores_dev.genconfig.json
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"max_length": 20,
|
| 3 |
+
"max_new_tokens": 512,
|
| 4 |
+
"min_length": 0,
|
| 5 |
+
"min_new_tokens": 0,
|
| 6 |
+
"early_stopping": false,
|
| 7 |
+
"max_time": null,
|
| 8 |
+
"stop_strings": null,
|
| 9 |
+
"do_sample": false,
|
| 10 |
+
"num_beams": 1,
|
| 11 |
+
"use_cache": true,
|
| 12 |
+
"cache_implementation": null,
|
| 13 |
+
"cache_config": null,
|
| 14 |
+
"return_legacy_cache": null,
|
| 15 |
+
"prefill_chunk_size": null,
|
| 16 |
+
"temperature": 1.0,
|
| 17 |
+
"top_k": 50,
|
| 18 |
+
"top_p": 1.0,
|
| 19 |
+
"min_p": null,
|
| 20 |
+
"typical_p": 1.0,
|
| 21 |
+
"epsilon_cutoff": 0.0,
|
| 22 |
+
"eta_cutoff": 0.0,
|
| 23 |
+
"repetition_penalty": 1.005,
|
| 24 |
+
"encoder_repetition_penalty": 1.0,
|
| 25 |
+
"length_penalty": 1.0,
|
| 26 |
+
"no_repeat_ngram_size": 0,
|
| 27 |
+
"bad_words_ids": null,
|
| 28 |
+
"renormalize_logits": false,
|
| 29 |
+
"forced_bos_token_id": null,
|
| 30 |
+
"forced_eos_token_id": null,
|
| 31 |
+
"remove_invalid_values": false,
|
| 32 |
+
"exponential_decay_length_penalty": null,
|
| 33 |
+
"suppress_tokens": null,
|
| 34 |
+
"begin_suppress_tokens": null,
|
| 35 |
+
"sequence_bias": null,
|
| 36 |
+
"token_healing": false,
|
| 37 |
+
"guidance_scale": null,
|
| 38 |
+
"watermarking_config": null,
|
| 39 |
+
"num_return_sequences": 1,
|
| 40 |
+
"output_attentions": false,
|
| 41 |
+
"output_hidden_states": false,
|
| 42 |
+
"output_scores": false,
|
| 43 |
+
"output_logits": null,
|
| 44 |
+
"return_dict_in_generate": false,
|
| 45 |
+
"pad_token_id": 5,
|
| 46 |
+
"bos_token_id": 3,
|
| 47 |
+
"eos_token_id": 5,
|
| 48 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 49 |
+
"decoder_start_token_id": null,
|
| 50 |
+
"is_assistant": false,
|
| 51 |
+
"num_assistant_tokens": 20,
|
| 52 |
+
"num_assistant_tokens_schedule": "constant",
|
| 53 |
+
"assistant_confidence_threshold": 0.4,
|
| 54 |
+
"prompt_lookup_num_tokens": null,
|
| 55 |
+
"max_matching_ngram_size": null,
|
| 56 |
+
"assistant_early_exit": null,
|
| 57 |
+
"assistant_lookbehind": 10,
|
| 58 |
+
"target_lookbehind": 10,
|
| 59 |
+
"disable_compile": false,
|
| 60 |
+
"low_memory": null,
|
| 61 |
+
"penalty_alpha": null,
|
| 62 |
+
"dola_layers": null,
|
| 63 |
+
"diversity_penalty": 0.0,
|
| 64 |
+
"num_beam_groups": 1,
|
| 65 |
+
"constraints": null,
|
| 66 |
+
"force_words_ids": null,
|
| 67 |
+
"_from_model_config": true,
|
| 68 |
+
"transformers_version": "4.57.1",
|
| 69 |
+
"_meta": {
|
| 70 |
+
"model": "/mnt/models/CODE/mzy/v32_ckpts/hf_s2r_iter167",
|
| 71 |
+
"greedy": true,
|
| 72 |
+
"sampling_params_inert": "do_sample=False, temperature/top_p/top_k 不参与生成",
|
| 73 |
+
"prompt_mode": "chat_template",
|
| 74 |
+
"tokenizer_pad": "<|im_end|>",
|
| 75 |
+
"tokenizer_eos": "<|im_end|>",
|
| 76 |
+
"transformers": "4.57.1",
|
| 77 |
+
"torch": "2.7.1+cpu"
|
| 78 |
+
}
|
| 79 |
+
}
|
eval/pred_haidass-v32s2r-iter167_rp1005_flores_dev.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
eval/pred_haidass-v32s2r-iter167_rp1005_floresplus_devtest.genconfig.json
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"max_length": 20,
|
| 3 |
+
"max_new_tokens": 512,
|
| 4 |
+
"min_length": 0,
|
| 5 |
+
"min_new_tokens": 0,
|
| 6 |
+
"early_stopping": false,
|
| 7 |
+
"max_time": null,
|
| 8 |
+
"stop_strings": null,
|
| 9 |
+
"do_sample": false,
|
| 10 |
+
"num_beams": 1,
|
| 11 |
+
"use_cache": true,
|
| 12 |
+
"cache_implementation": null,
|
| 13 |
+
"cache_config": null,
|
| 14 |
+
"return_legacy_cache": null,
|
| 15 |
+
"prefill_chunk_size": null,
|
| 16 |
+
"temperature": 1.0,
|
| 17 |
+
"top_k": 50,
|
| 18 |
+
"top_p": 1.0,
|
| 19 |
+
"min_p": null,
|
| 20 |
+
"typical_p": 1.0,
|
| 21 |
+
"epsilon_cutoff": 0.0,
|
| 22 |
+
"eta_cutoff": 0.0,
|
| 23 |
+
"repetition_penalty": 1.005,
|
| 24 |
+
"encoder_repetition_penalty": 1.0,
|
| 25 |
+
"length_penalty": 1.0,
|
| 26 |
+
"no_repeat_ngram_size": 0,
|
| 27 |
+
"bad_words_ids": null,
|
| 28 |
+
"renormalize_logits": false,
|
| 29 |
+
"forced_bos_token_id": null,
|
| 30 |
+
"forced_eos_token_id": null,
|
| 31 |
+
"remove_invalid_values": false,
|
| 32 |
+
"exponential_decay_length_penalty": null,
|
| 33 |
+
"suppress_tokens": null,
|
| 34 |
+
"begin_suppress_tokens": null,
|
| 35 |
+
"sequence_bias": null,
|
| 36 |
+
"token_healing": false,
|
| 37 |
+
"guidance_scale": null,
|
| 38 |
+
"watermarking_config": null,
|
| 39 |
+
"num_return_sequences": 1,
|
| 40 |
+
"output_attentions": false,
|
| 41 |
+
"output_hidden_states": false,
|
| 42 |
+
"output_scores": false,
|
| 43 |
+
"output_logits": null,
|
| 44 |
+
"return_dict_in_generate": false,
|
| 45 |
+
"pad_token_id": 5,
|
| 46 |
+
"bos_token_id": 3,
|
| 47 |
+
"eos_token_id": 5,
|
| 48 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 49 |
+
"decoder_start_token_id": null,
|
| 50 |
+
"is_assistant": false,
|
| 51 |
+
"num_assistant_tokens": 20,
|
| 52 |
+
"num_assistant_tokens_schedule": "constant",
|
| 53 |
+
"assistant_confidence_threshold": 0.4,
|
| 54 |
+
"prompt_lookup_num_tokens": null,
|
| 55 |
+
"max_matching_ngram_size": null,
|
| 56 |
+
"assistant_early_exit": null,
|
| 57 |
+
"assistant_lookbehind": 10,
|
| 58 |
+
"target_lookbehind": 10,
|
| 59 |
+
"disable_compile": false,
|
| 60 |
+
"low_memory": null,
|
| 61 |
+
"penalty_alpha": null,
|
| 62 |
+
"dola_layers": null,
|
| 63 |
+
"diversity_penalty": 0.0,
|
| 64 |
+
"num_beam_groups": 1,
|
| 65 |
+
"constraints": null,
|
| 66 |
+
"force_words_ids": null,
|
| 67 |
+
"_from_model_config": true,
|
| 68 |
+
"transformers_version": "4.57.1",
|
| 69 |
+
"_meta": {
|
| 70 |
+
"model": "/mnt/models/CODE/mzy/v32_ckpts/hf_s2r_iter167",
|
| 71 |
+
"greedy": true,
|
| 72 |
+
"sampling_params_inert": "do_sample=False, temperature/top_p/top_k 不参与生成",
|
| 73 |
+
"prompt_mode": "chat_template",
|
| 74 |
+
"tokenizer_pad": "<|im_end|>",
|
| 75 |
+
"tokenizer_eos": "<|im_end|>",
|
| 76 |
+
"transformers": "4.57.1",
|
| 77 |
+
"torch": "2.7.1+cpu"
|
| 78 |
+
}
|
| 79 |
+
}
|
eval/pred_haidass-v32s2r-iter167_rp1005_floresplus_devtest.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
generation_config.json
CHANGED
|
@@ -3,5 +3,7 @@
|
|
| 3 |
"eos_token_id": 5,
|
| 4 |
"pad_token_id": 5,
|
| 5 |
"do_sample": false,
|
|
|
|
|
|
|
| 6 |
"max_new_tokens": 256
|
| 7 |
}
|
|
|
|
| 3 |
"eos_token_id": 5,
|
| 4 |
"pad_token_id": 5,
|
| 5 |
"do_sample": false,
|
| 6 |
+
"num_beams": 1,
|
| 7 |
+
"repetition_penalty": 1.005,
|
| 8 |
"max_new_tokens": 256
|
| 9 |
}
|
model-00001-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 80809784
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:23e41f18c08ec27142276cdd9deee0a3d033bb44385ed3d12ed5c741c2e3fb85
|
| 3 |
size 80809784
|
model-00002-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d111e49d8ea52aa691ea5c58e2e0d4fabae9fc309bf7f910e478777203785216
|
| 3 |
size 7081656
|
model-00003-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02f1d97fc5a5bad1968ce31610ef74866765723e9ccbf8e0c354eb3d55b08df5
|
| 3 |
size 7081656
|
model-00004-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd71ddcb5ebd8e6abdb9636a84831e5fb73226df65d8dd285d03b5546b9f569e
|
| 3 |
size 7081656
|
model-00005-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96b20abeae990238cb8de3720ed96fa8fd735a60ff61e80a5286b4d23e3bce75
|
| 3 |
size 7081656
|
model-00006-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9f1c54a94120f888dc594ba67b22faf35545941c52e89ac66e1f666ac645098
|
| 3 |
size 7081656
|
model-00007-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d5f7d032bdc54b4d793c77abeb5e8318fbfc6824e6bb66576d06f044f019330
|
| 3 |
size 7081656
|
model-00008-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ae5a24c266cc60ee54f3348e90d87f7485cf89409505ea8fdca992c38a5be54c
|
| 3 |
size 7081656
|
model-00009-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f717bc50a122ffa3323a4a758ad0c47216728f8f42e0d80d61bc073418bf881b
|
| 3 |
size 7081656
|
model-00010-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3867db52d9081cfa820be9c9dee0d1a713ebbba54c107e245fd9f92a6a0b4002
|
| 3 |
size 7081656
|
model-00011-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48e7fbb0d66677939bc4f38fb7371590f2afd99657c72f9d82c9378967880c16
|
| 3 |
size 7081672
|
model-00012-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:21cabccce2e08fe1bbcb9ad885cd1d77ccbcf5002706bb8ea1d5dda035cdc8f8
|
| 3 |
size 7081672
|
model-00013-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08b269b8e0678db78d0fce084d3668526b6fde7941e5ae290c99b6f97119a5ba
|
| 3 |
size 7081672
|
model-00014-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f78917e69de48f82ee2f15ceb330f1f76cafc92e617319698b0d40ac6c54016
|
| 3 |
size 7081672
|
model-00015-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68cb17fb02e3d7427e0f326ab343caf3102368fad546e08a2552ef8d04ea5e5c
|
| 3 |
size 7081672
|
model-00016-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a1220b785becf8892df1ed2ab496acf76e2d30c1d4c55b50a11b972fd57f381
|
| 3 |
size 7081672
|
model-00017-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df90f9bb9ac44f81f9a1df8557d3de48c114319a966ee100c0adadb20f6dba14
|
| 3 |
size 7081672
|
model-00018-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6729705472242dd06f1eb267bf0b223db565e7b5089f2857ecc54c713ab7bfc4
|
| 3 |
size 7081672
|
model-00019-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a27053123fc5119b217828d715bb333c2bd128dd325dc89cd4df9500aa267df9
|
| 3 |
size 7081672
|
model-00020-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0111f786f9d39badf7ab5eebdb306e03d10d0413f8cc3853ba0730357711c145
|
| 3 |
size 7081672
|
model-00021-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a93c1aaa5bc102296d9e329872ad73d53b95fc58d2fa86cdc7c00833697a923
|
| 3 |
size 7081672
|
model-00022-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0cde4769843d2ad174aa637d50048826b4372599304818464fa962ccc4c366bd
|
| 3 |
size 7081672
|
model-00023-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa127219e79c7a312ef99734952edc1554d5517357809ff2e7fed658e659fa1a
|
| 3 |
size 7081672
|
model-00024-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cf77305173d76679cc60e67502d8ae7e2d7246dc1a854defb03e97c56988b83f
|
| 3 |
size 7081672
|
model-00025-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:061a5cc84830d48e1214500d006a96c9d3514a9b1a45c998f51b164f5dce47ac
|
| 3 |
size 7081672
|
model-00026-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5a90d40b6619948835bda7990474d2bef545aa09a91d3a025957db61566fb53d
|
| 3 |
size 7081672
|
model-00027-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8fc52a00e99908a01bcf2ae9ad99ed3fc4d11a6e7fdfab0825749af116a5d07e
|
| 3 |
size 7081672
|
model-00028-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:be885dd88fadd5b05804e118e441266782930bd210666d6e116d8f18fed6bfde
|
| 3 |
size 7081672
|
model-00029-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7081672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3308cf966e15b21d0193c9b818cce51a6c11b4d98ca991dc7d797f632695a8e1
|
| 3 |
size 7081672
|
model-00030-of-000030.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7082904
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4f7117a0abe217bddb8ed21a0ea6e0650b6a654da26aead04bf4961f7402b1b
|
| 3 |
size 7082904
|