math sft model (sft 20260514_093821) as group model
Browse files- chat_template.jinja +4 -87
- config.json +3 -3
- generation_config.json +0 -1
- hydra/.hydra/config.yaml +0 -108
- hydra/.hydra/hydra.yaml +0 -162
- hydra/.hydra/overrides.yaml +0 -3
- hydra/merge_peft.log +0 -23
- model.safetensors +2 -2
- tokenizer_config.json +2 -3
chat_template.jinja
CHANGED
|
@@ -1,91 +1,8 @@
|
|
| 1 |
{%- set enable_thinking = true %}
|
| 2 |
-
{%- set flab_boxed_instruction = "Solve the problem carefully. Think before answering. You must finish every response with exactly one final answer in the form \\boxed{LETTER}. Never omit the final boxed answer." %}
|
| 3 |
-
{%- if flab_boxed_instruction %}{{ '<|im_start|>system\n' + flab_boxed_instruction + '<|im_end|>\n' }}{%- endif %}{%- if tools %}
|
| 4 |
-
{{- '<|im_start|>system\n' }}
|
| 5 |
-
{%- if messages[0].role == 'system' %}
|
| 6 |
-
{{- messages[0].content + '\n\n' }}
|
| 7 |
-
{%- endif %}
|
| 8 |
-
{{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
| 9 |
-
{%- for tool in tools %}
|
| 10 |
-
{{- "\n" }}
|
| 11 |
-
{{- tool | tojson }}
|
| 12 |
-
{%- endfor %}
|
| 13 |
-
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
| 14 |
-
{%- else %}
|
| 15 |
-
{%- if messages[0].role == 'system' %}
|
| 16 |
-
{{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
|
| 17 |
-
{%- endif %}
|
| 18 |
-
{%- endif %}
|
| 19 |
-
{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
|
| 20 |
-
{%- for message in messages[::-1] %}
|
| 21 |
-
{%- set index = (messages|length - 1) - loop.index0 %}
|
| 22 |
-
{%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
|
| 23 |
-
{%- set ns.multi_step_tool = false %}
|
| 24 |
-
{%- set ns.last_query_index = index %}
|
| 25 |
-
{%- endif %}
|
| 26 |
-
{%- endfor %}
|
| 27 |
{%- for message in messages %}
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
{%- else %}
|
| 31 |
-
{%- set content = '' %}
|
| 32 |
-
{%- endif %}
|
| 33 |
-
{%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
|
| 34 |
-
{{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
|
| 35 |
-
{%- elif message.role == "assistant" %}
|
| 36 |
-
{%- set reasoning_content = '' %}
|
| 37 |
-
{%- if message.reasoning_content is string %}
|
| 38 |
-
{%- set reasoning_content = message.reasoning_content %}
|
| 39 |
-
{%- else %}
|
| 40 |
-
{%- if '</think>' in content %}
|
| 41 |
-
{%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
|
| 42 |
-
{%- set content = content.split('</think>')[-1].lstrip('\n') %}
|
| 43 |
-
{%- endif %}
|
| 44 |
-
{%- endif %}
|
| 45 |
-
{%- if loop.index0 > ns.last_query_index %}
|
| 46 |
-
{%- if loop.last or (not loop.last and reasoning_content) %}
|
| 47 |
-
{{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
|
| 48 |
-
{%- else %}
|
| 49 |
-
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 50 |
-
{%- endif %}
|
| 51 |
-
{%- else %}
|
| 52 |
-
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 53 |
-
{%- endif %}
|
| 54 |
-
{%- if message.tool_calls %}
|
| 55 |
-
{%- for tool_call in message.tool_calls %}
|
| 56 |
-
{%- if (loop.first and content) or (not loop.first) %}
|
| 57 |
-
{{- '\n' }}
|
| 58 |
-
{%- endif %}
|
| 59 |
-
{%- if tool_call.function %}
|
| 60 |
-
{%- set tool_call = tool_call.function %}
|
| 61 |
-
{%- endif %}
|
| 62 |
-
{{- '<tool_call>\n{"name": "' }}
|
| 63 |
-
{{- tool_call.name }}
|
| 64 |
-
{{- '", "arguments": ' }}
|
| 65 |
-
{%- if tool_call.arguments is string %}
|
| 66 |
-
{{- tool_call.arguments }}
|
| 67 |
-
{%- else %}
|
| 68 |
-
{{- tool_call.arguments | tojson }}
|
| 69 |
-
{%- endif %}
|
| 70 |
-
{{- '}\n</tool_call>' }}
|
| 71 |
-
{%- endfor %}
|
| 72 |
-
{%- endif %}
|
| 73 |
-
{{- '<|im_end|>\n' }}
|
| 74 |
-
{%- elif message.role == "tool" %}
|
| 75 |
-
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
|
| 76 |
-
{{- '<|im_start|>user' }}
|
| 77 |
-
{%- endif %}
|
| 78 |
-
{{- '\n<tool_response>\n' }}
|
| 79 |
-
{{- content }}
|
| 80 |
-
{{- '\n</tool_response>' }}
|
| 81 |
-
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
| 82 |
-
{{- '<|im_end|>\n' }}
|
| 83 |
-
{%- endif %}
|
| 84 |
-
{%- endif %}
|
| 85 |
{%- endfor %}
|
| 86 |
-
{%- if add_generation_prompt %}
|
| 87 |
-
|
| 88 |
-
{%- if enable_thinking is defined and enable_thinking is false %}
|
| 89 |
-
{{- '<think>\n\n</think>\n\n' }}
|
| 90 |
-
{%- endif %}
|
| 91 |
{%- endif %}
|
|
|
|
| 1 |
{%- set enable_thinking = true %}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
{%- for message in messages %}
|
| 3 |
+
<|im_start|>{{ message["role"] }}
|
| 4 |
+
{{ message["content"] }}<|im_end|>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
{%- endfor %}
|
| 6 |
+
{%- if add_generation_prompt %}<|im_start|>assistant
|
| 7 |
+
<think>
|
|
|
|
|
|
|
|
|
|
| 8 |
{%- endif %}
|
config.json
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
],
|
| 5 |
"attention_bias": false,
|
| 6 |
"attention_dropout": 0.0,
|
| 7 |
-
"bos_token_id":
|
| 8 |
"dtype": "bfloat16",
|
| 9 |
"eos_token_id": 151645,
|
| 10 |
"head_dim": 128,
|
|
@@ -48,7 +48,7 @@
|
|
| 48 |
"num_attention_heads": 16,
|
| 49 |
"num_hidden_layers": 28,
|
| 50 |
"num_key_value_heads": 8,
|
| 51 |
-
"pad_token_id":
|
| 52 |
"rms_norm_eps": 1e-06,
|
| 53 |
"rope_parameters": {
|
| 54 |
"rope_theta": 1000000,
|
|
@@ -57,7 +57,7 @@
|
|
| 57 |
"sliding_window": null,
|
| 58 |
"tie_word_embeddings": true,
|
| 59 |
"transformers_version": "5.7.0",
|
| 60 |
-
"use_cache":
|
| 61 |
"use_sliding_window": false,
|
| 62 |
"vocab_size": 151936
|
| 63 |
}
|
|
|
|
| 4 |
],
|
| 5 |
"attention_bias": false,
|
| 6 |
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": null,
|
| 8 |
"dtype": "bfloat16",
|
| 9 |
"eos_token_id": 151645,
|
| 10 |
"head_dim": 128,
|
|
|
|
| 48 |
"num_attention_heads": 16,
|
| 49 |
"num_hidden_layers": 28,
|
| 50 |
"num_key_value_heads": 8,
|
| 51 |
+
"pad_token_id": 151643,
|
| 52 |
"rms_norm_eps": 1e-06,
|
| 53 |
"rope_parameters": {
|
| 54 |
"rope_theta": 1000000,
|
|
|
|
| 57 |
"sliding_window": null,
|
| 58 |
"tie_word_embeddings": true,
|
| 59 |
"transformers_version": "5.7.0",
|
| 60 |
+
"use_cache": false,
|
| 61 |
"use_sliding_window": false,
|
| 62 |
"vocab_size": 151936
|
| 63 |
}
|
generation_config.json
CHANGED
|
@@ -1,5 +1,4 @@
|
|
| 1 |
{
|
| 2 |
-
"bos_token_id": 151643,
|
| 3 |
"do_sample": true,
|
| 4 |
"eos_token_id": [
|
| 5 |
151645,
|
|
|
|
| 1 |
{
|
|
|
|
| 2 |
"do_sample": true,
|
| 3 |
"eos_token_id": [
|
| 4 |
151645,
|
hydra/.hydra/config.yaml
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
model:
|
| 2 |
-
hf_id: Qwen/Qwen3-1.7B
|
| 3 |
-
local_path: /shared-ro/models/Qwen3-1.7B
|
| 4 |
-
dtype: bfloat16
|
| 5 |
-
attn_impl: flash_attention_2
|
| 6 |
-
max_model_len: 4096
|
| 7 |
-
data:
|
| 8 |
-
name: nemotron_math
|
| 9 |
-
loader: load_nemotron_math
|
| 10 |
-
kwargs:
|
| 11 |
-
limit: 50000
|
| 12 |
-
split: train
|
| 13 |
-
max_total_tokens: 3800
|
| 14 |
-
seed: 0
|
| 15 |
-
training:
|
| 16 |
-
recipe: sft
|
| 17 |
-
epochs: 1
|
| 18 |
-
per_device_train_batch_size: 1
|
| 19 |
-
gradient_accumulation_steps: 16
|
| 20 |
-
learning_rate: 5.0e-06
|
| 21 |
-
warmup_ratio: 0.03
|
| 22 |
-
lr_scheduler_type: cosine
|
| 23 |
-
gradient_checkpointing: true
|
| 24 |
-
bf16: true
|
| 25 |
-
max_seq_length: 4096
|
| 26 |
-
max_steps: -1
|
| 27 |
-
max_grad_norm: 1.0
|
| 28 |
-
optim: adamw_torch
|
| 29 |
-
assistant_only_loss: true
|
| 30 |
-
logging_steps: 10
|
| 31 |
-
save_steps: 500
|
| 32 |
-
save_total_limit: 2
|
| 33 |
-
report_to: wandb
|
| 34 |
-
qlora: false
|
| 35 |
-
lora_r: 16
|
| 36 |
-
lora_alpha: 32
|
| 37 |
-
lora_dropout: 0.05
|
| 38 |
-
lora_target_modules:
|
| 39 |
-
- q_proj
|
| 40 |
-
- k_proj
|
| 41 |
-
- v_proj
|
| 42 |
-
- o_proj
|
| 43 |
-
- gate_proj
|
| 44 |
-
- up_proj
|
| 45 |
-
- down_proj
|
| 46 |
-
deepspeed: configs/deepspeed_zero2.json
|
| 47 |
-
adapter_model_dir: null
|
| 48 |
-
domain:
|
| 49 |
-
name: multilingual
|
| 50 |
-
thinking_mode: true
|
| 51 |
-
benchmark: multilingual
|
| 52 |
-
languages:
|
| 53 |
-
- it
|
| 54 |
-
- es
|
| 55 |
-
- zh
|
| 56 |
-
- ru
|
| 57 |
-
- hi
|
| 58 |
-
target_format: boxed
|
| 59 |
-
max_options: 20
|
| 60 |
-
reward_fn: multilingual
|
| 61 |
-
system_instruction: Solve the problem carefully. Think before answering. You must
|
| 62 |
-
finish every response with exactly one final answer in the form \boxed{LETTER}.
|
| 63 |
-
Never omit the final boxed answer.
|
| 64 |
-
template_boxed_instruction: Solve the problem carefully. Think before answering.
|
| 65 |
-
You must finish every response with exactly one final answer in the form \boxed{LETTER}.
|
| 66 |
-
Never omit the final boxed answer.
|
| 67 |
-
post_train:
|
| 68 |
-
enabled: false
|
| 69 |
-
merge_adapter: false
|
| 70 |
-
merged_output_dir: null
|
| 71 |
-
domain_override: null
|
| 72 |
-
evals: []
|
| 73 |
-
seed: 42
|
| 74 |
-
run_id: ${now:%Y%m%d_%H%M%S}
|
| 75 |
-
output_dir: /scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3_merged
|
| 76 |
-
wandb:
|
| 77 |
-
project: flab-cs552
|
| 78 |
-
entity: cs-552-2026-flab
|
| 79 |
-
mode: null
|
| 80 |
-
tags: []
|
| 81 |
-
notes: null
|
| 82 |
-
eval:
|
| 83 |
-
model_dir: null
|
| 84 |
-
rows_source: samples
|
| 85 |
-
method: null
|
| 86 |
-
system_prompt: null
|
| 87 |
-
use_domain_system_prompt: false
|
| 88 |
-
force_domain_template: false
|
| 89 |
-
'n': 8
|
| 90 |
-
temperature: 0.7
|
| 91 |
-
top_p: 0.95
|
| 92 |
-
max_model_len: 4096
|
| 93 |
-
max_tokens: 4096
|
| 94 |
-
samples_dir: null
|
| 95 |
-
push:
|
| 96 |
-
model_dir: /scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3
|
| 97 |
-
es:
|
| 98 |
-
base_model_dir: null
|
| 99 |
-
data: null
|
| 100 |
-
hf_push:
|
| 101 |
-
enabled: false
|
| 102 |
-
org: cs-552-2026-flab
|
| 103 |
-
repo: null
|
| 104 |
-
private: false
|
| 105 |
-
vllm_validate: true
|
| 106 |
-
commit_message: null
|
| 107 |
-
validate_after_train: false
|
| 108 |
-
vllm_validate: true
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
hydra/.hydra/hydra.yaml
DELETED
|
@@ -1,162 +0,0 @@
|
|
| 1 |
-
hydra:
|
| 2 |
-
run:
|
| 3 |
-
dir: ${output_dir}/hydra
|
| 4 |
-
sweep:
|
| 5 |
-
dir: ${output_dir}/hydra_sweep
|
| 6 |
-
subdir: ${hydra.job.num}
|
| 7 |
-
launcher:
|
| 8 |
-
_target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
|
| 9 |
-
sweeper:
|
| 10 |
-
_target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
|
| 11 |
-
max_batch_size: null
|
| 12 |
-
params: null
|
| 13 |
-
help:
|
| 14 |
-
app_name: ${hydra.job.name}
|
| 15 |
-
header: '${hydra.help.app_name} is powered by Hydra.
|
| 16 |
-
|
| 17 |
-
'
|
| 18 |
-
footer: 'Powered by Hydra (https://hydra.cc)
|
| 19 |
-
|
| 20 |
-
Use --hydra-help to view Hydra specific help
|
| 21 |
-
|
| 22 |
-
'
|
| 23 |
-
template: '${hydra.help.header}
|
| 24 |
-
|
| 25 |
-
== Configuration groups ==
|
| 26 |
-
|
| 27 |
-
Compose your configuration from those groups (group=option)
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
$APP_CONFIG_GROUPS
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
== Config ==
|
| 34 |
-
|
| 35 |
-
Override anything in the config (foo.bar=value)
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
$CONFIG
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
${hydra.help.footer}
|
| 42 |
-
|
| 43 |
-
'
|
| 44 |
-
hydra_help:
|
| 45 |
-
template: 'Hydra (${hydra.runtime.version})
|
| 46 |
-
|
| 47 |
-
See https://hydra.cc for more info.
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
== Flags ==
|
| 51 |
-
|
| 52 |
-
$FLAGS_HELP
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
== Configuration groups ==
|
| 56 |
-
|
| 57 |
-
Compose your configuration from those groups (For example, append hydra/job_logging=disabled
|
| 58 |
-
to command line)
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
$HYDRA_CONFIG_GROUPS
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
Use ''--cfg hydra'' to Show the Hydra config.
|
| 65 |
-
|
| 66 |
-
'
|
| 67 |
-
hydra_help: ???
|
| 68 |
-
hydra_logging:
|
| 69 |
-
version: 1
|
| 70 |
-
formatters:
|
| 71 |
-
simple:
|
| 72 |
-
format: '[%(asctime)s][HYDRA] %(message)s'
|
| 73 |
-
handlers:
|
| 74 |
-
console:
|
| 75 |
-
class: logging.StreamHandler
|
| 76 |
-
formatter: simple
|
| 77 |
-
stream: ext://sys.stdout
|
| 78 |
-
root:
|
| 79 |
-
level: INFO
|
| 80 |
-
handlers:
|
| 81 |
-
- console
|
| 82 |
-
loggers:
|
| 83 |
-
logging_example:
|
| 84 |
-
level: DEBUG
|
| 85 |
-
disable_existing_loggers: false
|
| 86 |
-
job_logging:
|
| 87 |
-
version: 1
|
| 88 |
-
formatters:
|
| 89 |
-
simple:
|
| 90 |
-
format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
|
| 91 |
-
handlers:
|
| 92 |
-
console:
|
| 93 |
-
class: logging.StreamHandler
|
| 94 |
-
formatter: simple
|
| 95 |
-
stream: ext://sys.stdout
|
| 96 |
-
file:
|
| 97 |
-
class: logging.FileHandler
|
| 98 |
-
formatter: simple
|
| 99 |
-
filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
|
| 100 |
-
root:
|
| 101 |
-
level: INFO
|
| 102 |
-
handlers:
|
| 103 |
-
- console
|
| 104 |
-
- file
|
| 105 |
-
disable_existing_loggers: false
|
| 106 |
-
env: {}
|
| 107 |
-
mode: RUN
|
| 108 |
-
searchpath: []
|
| 109 |
-
callbacks: {}
|
| 110 |
-
output_subdir: .hydra
|
| 111 |
-
overrides:
|
| 112 |
-
hydra:
|
| 113 |
-
- hydra.mode=RUN
|
| 114 |
-
task:
|
| 115 |
-
- domain=multilingual_reasoning_thinking
|
| 116 |
-
- push.model_dir=/scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3
|
| 117 |
-
- output_dir=/scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3_merged
|
| 118 |
-
job:
|
| 119 |
-
name: merge_peft
|
| 120 |
-
chdir: false
|
| 121 |
-
override_dirname: domain=multilingual_reasoning_thinking,output_dir=/scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3_merged,push.model_dir=/scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3
|
| 122 |
-
id: ???
|
| 123 |
-
num: ???
|
| 124 |
-
config_name: base
|
| 125 |
-
env_set: {}
|
| 126 |
-
env_copy: []
|
| 127 |
-
config:
|
| 128 |
-
override_dirname:
|
| 129 |
-
kv_sep: '='
|
| 130 |
-
item_sep: ','
|
| 131 |
-
exclude_keys: []
|
| 132 |
-
runtime:
|
| 133 |
-
version: 1.3.2
|
| 134 |
-
version_base: '1.3'
|
| 135 |
-
cwd: /scratch/runs/cs552-mazniche-g07-gcall2e-c3-033328
|
| 136 |
-
config_sources:
|
| 137 |
-
- path: hydra.conf
|
| 138 |
-
schema: pkg
|
| 139 |
-
provider: hydra
|
| 140 |
-
- path: /scratch/runs/cs552-mazniche-g07-gcall2e-c3-033328/configs
|
| 141 |
-
schema: file
|
| 142 |
-
provider: main
|
| 143 |
-
- path: ''
|
| 144 |
-
schema: structured
|
| 145 |
-
provider: schema
|
| 146 |
-
output_dir: /scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3_merged/hydra
|
| 147 |
-
choices:
|
| 148 |
-
post_train: disabled
|
| 149 |
-
domain: multilingual_reasoning_thinking
|
| 150 |
-
training: sft
|
| 151 |
-
data: nemotron_math
|
| 152 |
-
model: qwen3_17b
|
| 153 |
-
hydra/env: default
|
| 154 |
-
hydra/callbacks: null
|
| 155 |
-
hydra/job_logging: default
|
| 156 |
-
hydra/hydra_logging: default
|
| 157 |
-
hydra/hydra_help: default
|
| 158 |
-
hydra/help: default
|
| 159 |
-
hydra/sweeper: basic
|
| 160 |
-
hydra/launcher: basic
|
| 161 |
-
hydra/output: default
|
| 162 |
-
verbose: false
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
hydra/.hydra/overrides.yaml
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
- domain=multilingual_reasoning_thinking
|
| 2 |
-
- push.model_dir=/scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3
|
| 3 |
-
- output_dir=/scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3_merged
|
|
|
|
|
|
|
|
|
|
|
|
hydra/merge_peft.log
DELETED
|
@@ -1,23 +0,0 @@
|
|
| 1 |
-
[2026-05-18 07:59:21,026][__main__][INFO] - Loading base model for merge ...
|
| 2 |
-
[2026-05-18 07:59:21,185][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/Qwen/Qwen3-1.7B/resolve/main/config.json "HTTP/1.1 307 Temporary Redirect"
|
| 3 |
-
[2026-05-18 07:59:21,194][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/Qwen/Qwen3-1.7B/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e/config.json "HTTP/1.1 200 OK"
|
| 4 |
-
[2026-05-18 07:59:21,322][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/Qwen/Qwen3-1.7B/resolve/main/config.json "HTTP/1.1 307 Temporary Redirect"
|
| 5 |
-
[2026-05-18 07:59:21,331][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/Qwen/Qwen3-1.7B/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e/config.json "HTTP/1.1 200 OK"
|
| 6 |
-
[2026-05-18 07:59:21,740][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/Qwen/Qwen3-1.7B/resolve/main/generation_config.json "HTTP/1.1 307 Temporary Redirect"
|
| 7 |
-
[2026-05-18 07:59:21,747][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/Qwen/Qwen3-1.7B/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e/generation_config.json "HTTP/1.1 200 OK"
|
| 8 |
-
[2026-05-18 07:59:21,751][__main__][INFO] - Loading PEFT adapter from /scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3 ...
|
| 9 |
-
[2026-05-18 07:59:24,524][__main__][INFO] - Merging adapter into base weights ...
|
| 10 |
-
[2026-05-18 07:59:32,756][root][INFO] - x86_64-linux-gnu-gcc -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O2 -Wall -fPIC -c /tmp/tmpe0op8eya/test.c -o /tmp/tmpe0op8eya/test.o
|
| 11 |
-
[2026-05-18 07:59:32,776][root][INFO] - x86_64-linux-gnu-gcc /tmp/tmpe0op8eya/test.o -laio -o /tmp/tmpe0op8eya/a.out
|
| 12 |
-
[2026-05-18 07:59:32,791][root][INFO] - x86_64-linux-gnu-gcc -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O2 -Wall -fPIC -c /tmp/tmpqaqjo4lx/test.c -o /tmp/tmpqaqjo4lx/test.o
|
| 13 |
-
[2026-05-18 07:59:32,810][root][INFO] - x86_64-linux-gnu-gcc /tmp/tmpqaqjo4lx/test.o -L/usr/local/cuda -L/usr/local/cuda/lib64 -lcufile -o /tmp/tmpqaqjo4lx/a.out
|
| 14 |
-
[2026-05-18 07:59:32,856][root][INFO] - x86_64-linux-gnu-gcc -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O2 -Wall -fPIC -c /tmp/tmp4ir03sjd/test.c -o /tmp/tmp4ir03sjd/test.o
|
| 15 |
-
[2026-05-18 07:59:32,875][root][INFO] - x86_64-linux-gnu-gcc /tmp/tmp4ir03sjd/test.o -laio -o /tmp/tmp4ir03sjd/a.out
|
| 16 |
-
[2026-05-18 07:59:40,870][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/Qwen/Qwen3-1.7B/resolve/main/config.json "HTTP/1.1 307 Temporary Redirect"
|
| 17 |
-
[2026-05-18 07:59:40,878][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/Qwen/Qwen3-1.7B/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e/config.json "HTTP/1.1 200 OK"
|
| 18 |
-
[2026-05-18 07:59:41,004][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/Qwen/Qwen3-1.7B/resolve/main/tokenizer_config.json "HTTP/1.1 307 Temporary Redirect"
|
| 19 |
-
[2026-05-18 07:59:41,013][httpx][INFO] - HTTP Request: HEAD https://huggingface.co/api/resolve-cache/models/Qwen/Qwen3-1.7B/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e/tokenizer_config.json "HTTP/1.1 200 OK"
|
| 20 |
-
[2026-05-18 07:59:41,134][httpx][INFO] - HTTP Request: GET https://huggingface.co/api/models/Qwen/Qwen3-1.7B/tree/main/additional_chat_templates?recursive=false&expand=false "HTTP/1.1 404 Not Found"
|
| 21 |
-
[2026-05-18 07:59:41,256][httpx][INFO] - HTTP Request: GET https://huggingface.co/api/models/Qwen/Qwen3-1.7B/tree/main?recursive=true&expand=false "HTTP/1.1 200 OK"
|
| 22 |
-
[2026-05-18 07:59:42,028][httpx][INFO] - HTTP Request: GET https://huggingface.co/api/models/Qwen/Qwen3-1.7B "HTTP/1.1 200 OK"
|
| 23 |
-
[2026-05-18 07:59:42,145][__main__][INFO] - Merged checkpoint saved to /scratch/checkpoints/multilingual/overnight-bestshot-20260518/gcall2e-c3_merged
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af67a286c5b5a6dac80ab6a00cbae5a4c74ccc73502c072ff47836c8d50302b8
|
| 3 |
+
size 4063515640
|
tokenizer_config.json
CHANGED
|
@@ -20,12 +20,11 @@
|
|
| 20 |
"<|image_pad|>",
|
| 21 |
"<|video_pad|>"
|
| 22 |
],
|
| 23 |
-
"is_local":
|
| 24 |
"local_files_only": false,
|
| 25 |
"model_max_length": 131072,
|
| 26 |
"pad_token": "<|endoftext|>",
|
| 27 |
"split_special_tokens": false,
|
| 28 |
"tokenizer_class": "Qwen2Tokenizer",
|
| 29 |
-
"unk_token": null
|
| 30 |
-
"chat_template": "{%- set enable_thinking = true %}\n{%- set flab_boxed_instruction = \"Solve the problem carefully. Think before answering. You must finish every response with exactly one final answer in the form \\\\boxed{LETTER}. Never omit the final boxed answer.\" %}\n{%- if flab_boxed_instruction %}{{ '<|im_start|>system\\n' + flab_boxed_instruction + '<|im_end|>\\n' }}{%- endif %}{%- if tools %}\n {{- '<|im_start|>system\\n' }}\n {%- if messages[0].role == 'system' %}\n {{- messages[0].content + '\\n\\n' }}\n {%- endif %}\n {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n {%- for tool in tools %}\n {{- \"\\n\" }}\n {{- tool | tojson }}\n {%- endfor %}\n {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n {%- if messages[0].role == 'system' %}\n {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n {%- set index = (messages|length - 1) - loop.index0 %}\n {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n {%- set ns.multi_step_tool = false %}\n {%- set ns.last_query_index = index %}\n {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n {%- if message.content is string %}\n {%- set content = message.content %}\n {%- else %}\n {%- set content = '' %}\n {%- endif %}\n {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"assistant\" %}\n {%- set reasoning_content = '' %}\n {%- if message.reasoning_content is string %}\n {%- set reasoning_content = message.reasoning_content %}\n {%- else %}\n {%- if '</think>' in content %}\n {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n {%- endif %}\n {%- endif %}\n {%- if loop.index0 > ns.last_query_index %}\n {%- if loop.last or (not loop.last and reasoning_content) %}\n {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n {%- else %}\n {{- '<|im_start|>' + message.role + '\\n' + content }}\n {%- endif %}\n {%- else %}\n {{- '<|im_start|>' + message.role + '\\n' + content }}\n {%- endif %}\n {%- if message.tool_calls %}\n {%- for tool_call in message.tool_calls %}\n {%- if (loop.first and content) or (not loop.first) %}\n {{- '\\n' }}\n {%- endif %}\n {%- if tool_call.function %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {{- '<tool_call>\\n{\"name\": \"' }}\n {{- tool_call.name }}\n {{- '\", \"arguments\": ' }}\n {%- if tool_call.arguments is string %}\n {{- tool_call.arguments }}\n {%- else %}\n {{- tool_call.arguments | tojson }}\n {%- endif %}\n {{- '}\\n</tool_call>' }}\n {%- endfor %}\n {%- endif %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n<tool_response>\\n' }}\n {{- content }}\n {{- '\\n</tool_response>' }}\n {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\\n' }}\n {%- if enable_thinking is defined and enable_thinking is false %}\n {{- '<think>\\n\\n</think>\\n\\n' }}\n {%- endif %}\n{%- endif %}"
|
| 31 |
}
|
|
|
|
| 20 |
"<|image_pad|>",
|
| 21 |
"<|video_pad|>"
|
| 22 |
],
|
| 23 |
+
"is_local": true,
|
| 24 |
"local_files_only": false,
|
| 25 |
"model_max_length": 131072,
|
| 26 |
"pad_token": "<|endoftext|>",
|
| 27 |
"split_special_tokens": false,
|
| 28 |
"tokenizer_class": "Qwen2Tokenizer",
|
| 29 |
+
"unk_token": null
|
|
|
|
| 30 |
}
|