Update to v0.2 checkpoint
#14
by
loubnabnl
HF staff
- opened
- README.md +59 -12
- all_results.json +0 -22
- eval_results.json +0 -16
- generation_config.json +1 -0
- model.safetensors +1 -1
- onnx/model.onnx +2 -2
- onnx/model.onnx_data +1 -1
- onnx/model_bnb4.onnx +2 -2
- onnx/model_fp16.onnx +2 -2
- onnx/model_fp16.onnx_data +2 -2
- onnx/model_int8.onnx +2 -2
- onnx/model_q4.onnx +2 -2
- onnx/model_q4f16.onnx +2 -2
- onnx/model_quantized.onnx +2 -2
- onnx/model_uint8.onnx +2 -2
- runs/Jul15_23-01-52_ip-26-0-171-168/events.out.tfevents.1721084661.ip-26-0-171-168.1537695.0 +0 -3
- runs/Jul15_23-01-52_ip-26-0-171-168/events.out.tfevents.1721085183.ip-26-0-171-168.1537695.1 +0 -3
- runs/Jul15_23-02-15_ip-26-0-175-34/events.out.tfevents.1721084673.ip-26-0-175-34.1294397.0 +0 -3
- runs/Jul15_23-02-15_ip-26-0-175-34/events.out.tfevents.1721085077.ip-26-0-175-34.1294397.1 +0 -3
- train_results.json +0 -9
- trainer_state.json +0 -147
- training_args.bin +0 -3
README.md
CHANGED
@@ -1,12 +1,21 @@
|
|
1 |
---
|
2 |
-
library_name: transformers
|
3 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
language:
|
5 |
- en
|
6 |
---
|
7 |
|
8 |
-
|
9 |
-
# SmolLM-Instruct
|
10 |
|
11 |
<center>
|
12 |
<img src="https://huggingface.co/datasets/HuggingFaceTB/images/resolve/main/banner_smol.png" alt="SmolLM" width="1100" height="600">
|
@@ -15,14 +24,35 @@ language:
|
|
15 |
|
16 |
## Model Summary
|
17 |
|
18 |
-
SmolLM is a series of
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
|
20 |
-
|
21 |
-
|
|
|
|
|
|
|
|
|
22 |
|
23 |
-
|
|
|
24 |
|
25 |
-
|
|
|
|
|
26 |
```bash
|
27 |
pip install transformers
|
28 |
```
|
@@ -37,18 +67,35 @@ tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
|
37 |
# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
|
38 |
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
|
39 |
|
40 |
-
messages = [{"role": "user", "content": "
|
41 |
input_text=tokenizer.apply_chat_template(messages, tokenize=False)
|
42 |
print(input_text)
|
43 |
inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
|
44 |
-
outputs = model.generate(inputs, max_new_tokens=
|
45 |
print(tokenizer.decode(outputs[0]))
|
46 |
```
|
47 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
-
|
|
|
|
|
|
|
|
|
50 |
|
51 |
-
|
52 |
|
53 |
# Citation
|
54 |
```bash
|
|
|
1 |
---
|
|
|
2 |
license: apache-2.0
|
3 |
+
base_model: HuggingFaceTB/SmolLM-1.7B
|
4 |
+
tags:
|
5 |
+
- alignment-handbook
|
6 |
+
- trl
|
7 |
+
- sft
|
8 |
+
datasets:
|
9 |
+
- Magpie-Align/Magpie-Pro-300K-Filtered
|
10 |
+
- bigcode/self-oss-instruct-sc2-exec-filter-50k
|
11 |
+
- teknium/OpenHermes-2.5
|
12 |
+
- HuggingFaceTB/everyday-conversations-llama3.1-2k
|
13 |
+
library_name: transformers
|
14 |
language:
|
15 |
- en
|
16 |
---
|
17 |
|
18 |
+
# SmolLM-1.7B-Instruct
|
|
|
19 |
|
20 |
<center>
|
21 |
<img src="https://huggingface.co/datasets/HuggingFaceTB/images/resolve/main/banner_smol.png" alt="SmolLM" width="1100" height="600">
|
|
|
24 |
|
25 |
## Model Summary
|
26 |
|
27 |
+
SmolLM is a series of small language models available in three sizes: 135M, 360M, and 1.7B parameters.
|
28 |
+
|
29 |
+
These models are pre-trained on [SmolLM-Corpus](https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus), a curated collection of high-quality educational and synthetic data designed for training LLMs. For further details, we refer to our [blogpost](https://huggingface.co/blog/smollm).
|
30 |
+
|
31 |
+
To build SmolLM-Instruct, we finetuned the base models on publicly available datasets.
|
32 |
+
|
33 |
+
## Changelog
|
34 |
+
|
35 |
+
|
36 |
+
|Release|Description|
|
37 |
+
|-|-|
|
38 |
+
|v0.1| Initial release of SmolLM-Instruct. We finetune on the permissive subset of the [WebInstructSub](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub) dataset, combined with [StarCoder2-Self-OSS-Instruct](https://huggingface.co/datasets/bigcode/self-oss-instruct-sc2-exec-filter-50k). Then, we perform DPO (Direct Preference Optimization) for one epoch on [HelpSteer](https://huggingface.co/datasets/nvidia/HelpSteer) for the 135M and 1.7B models, and [argilla/dpo-mix-7k](https://huggingface.co/datasets/argilla/dpo-mix-7k) for the 360M model.|
|
39 |
+
|v0.2| We changed the finetuning mix to datasets more suitable for smol models. We train on a new dataset of 2k simple everyday conversations we generated by llama3.1-70B [everyday-conversations-llama3.1-2k](https://huggingface.co/datasets/HuggingFaceTB/everyday-conversations-llama3.1-2k/), [Magpie-Pro-300K-Filtered](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered), [StarCoder2-Self-OSS-Instruct](https://huggingface.co/datasets/bigcode/self-oss-instruct-sc2-exec-filter-50k), and a small subset of [OpenHermes-2.5](https://huggingface.co/datasets/teknium/OpenHermes-2.5)|
|
40 |
+
|
41 |
+
v0.2 models are better at staying on topic and responding appropriately to standard prompts, such as greetings and questions about their role as AI assistants. SmolLM-360M-Instruct (v0.2) has a 63.3% win rate over SmolLM-360M-Instruct (v0.1) on AlpacaEval. You can find the details [here](https://huggingface.co/datasets/HuggingFaceTB/alpaca_eval_details/).
|
42 |
|
43 |
+
You can load v0.1 checkpoint by specifying `revision="v0.1"` in the transformers code:
|
44 |
+
```python
|
45 |
+
model = AutoModelForCausalLM.from_pretrained("HuggingFaceTB/SmolLM-1.7B-Instruct", revision="v0.1")
|
46 |
+
```
|
47 |
+
|
48 |
+
## Usage
|
49 |
|
50 |
+
### Local Applications
|
51 |
+
⚡ For local applications, you can find optimized implementations of the model in MLC, GGUF and Transformers.js formats, in addition to fast in-browser demos in this collection: https://huggingface.co/collections/HuggingFaceTB/local-smollms-66c0f3b2a15b4eed7fb198d0
|
52 |
|
53 |
+
We noticed that 4bit quantization degrades the quality of the 135M and 360M, so we use `q016` for MLC and ONNX/Transformers.js checkpoints for the WebGPU demos. We also suggest using temperature 0.2 and top-p 0.9.
|
54 |
+
|
55 |
+
### Transformers
|
56 |
```bash
|
57 |
pip install transformers
|
58 |
```
|
|
|
67 |
# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
|
68 |
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
|
69 |
|
70 |
+
messages = [{"role": "user", "content": "What is the capital of France."}]
|
71 |
input_text=tokenizer.apply_chat_template(messages, tokenize=False)
|
72 |
print(input_text)
|
73 |
inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
|
74 |
+
outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
|
75 |
print(tokenizer.decode(outputs[0]))
|
76 |
```
|
77 |
|
78 |
+
### Chat in TRL
|
79 |
+
You can also use the TRL CLI to chat with the model from the terminal:
|
80 |
+
```bash
|
81 |
+
pip install trl
|
82 |
+
trl chat --model_name_or_path HuggingFaceTB/SmolLM-1.7B-Instruct --device cpu
|
83 |
+
```
|
84 |
+
|
85 |
+
## Limitations
|
86 |
+
|
87 |
+
Additionally, the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data, we invite users to leverage them as assistive tools rather than definitive sources of information. We find that they can handle general knowledge questions, creative writing and basic Python programming. But they are English only and may have difficulty with arithmetics, editing tasks and complex reasoning. For more details about the models' capabilities, please refer to our [blog post](https://huggingface.co/blog/smollm).
|
88 |
+
|
89 |
+
## Training parameters
|
90 |
+
We train the models using the [alignment-handbook](https://github.com/huggingface/alignment-handbook) with the datasets mentioned in the changelog, using these parameters v0.2:
|
91 |
|
92 |
+
- 1 epoch
|
93 |
+
- lr 1e-3
|
94 |
+
- cosine schedule
|
95 |
+
- warmup ratio 0.1
|
96 |
+
- global batch size 262k tokens
|
97 |
|
98 |
+
You can find the training recipe here: https://github.com/huggingface/alignment-handbook/tree/smollm/recipes/smollm
|
99 |
|
100 |
# Citation
|
101 |
```bash
|
all_results.json
DELETED
@@ -1,22 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"epoch": 0.9872958257713249,
|
3 |
-
"eval_logits/chosen": 0.9550817608833313,
|
4 |
-
"eval_logits/rejected": 0.8631864190101624,
|
5 |
-
"eval_logps/chosen": -121.98506164550781,
|
6 |
-
"eval_logps/rejected": -149.43112182617188,
|
7 |
-
"eval_loss": 0.6671856641769409,
|
8 |
-
"eval_rewards/accuracies": 0.550000011920929,
|
9 |
-
"eval_rewards/chosen": -0.0465608611702919,
|
10 |
-
"eval_rewards/margins": 0.04670196771621704,
|
11 |
-
"eval_rewards/rejected": -0.09326283633708954,
|
12 |
-
"eval_runtime": 4.6203,
|
13 |
-
"eval_samples": 464,
|
14 |
-
"eval_samples_per_second": 100.425,
|
15 |
-
"eval_steps_per_second": 3.247,
|
16 |
-
"total_flos": 0.0,
|
17 |
-
"train_loss": 0.6753217893488267,
|
18 |
-
"train_runtime": 321.7901,
|
19 |
-
"train_samples": 8806,
|
20 |
-
"train_samples_per_second": 27.366,
|
21 |
-
"train_steps_per_second": 0.211
|
22 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
eval_results.json
DELETED
@@ -1,16 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"epoch": 0.9872958257713249,
|
3 |
-
"eval_logits/chosen": 0.9550817608833313,
|
4 |
-
"eval_logits/rejected": 0.8631864190101624,
|
5 |
-
"eval_logps/chosen": -121.98506164550781,
|
6 |
-
"eval_logps/rejected": -149.43112182617188,
|
7 |
-
"eval_loss": 0.6671856641769409,
|
8 |
-
"eval_rewards/accuracies": 0.550000011920929,
|
9 |
-
"eval_rewards/chosen": -0.0465608611702919,
|
10 |
-
"eval_rewards/margins": 0.04670196771621704,
|
11 |
-
"eval_rewards/rejected": -0.09326283633708954,
|
12 |
-
"eval_runtime": 4.6203,
|
13 |
-
"eval_samples": 464,
|
14 |
-
"eval_samples_per_second": 100.425,
|
15 |
-
"eval_steps_per_second": 3.247
|
16 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
generation_config.json
CHANGED
@@ -3,5 +3,6 @@
|
|
3 |
"bos_token_id": 1,
|
4 |
"eos_token_id": 2,
|
5 |
"pad_token_id": 2,
|
|
|
6 |
"transformers_version": "4.42.3"
|
7 |
}
|
|
|
3 |
"bos_token_id": 1,
|
4 |
"eos_token_id": 2,
|
5 |
"pad_token_id": 2,
|
6 |
+
"max_new_tokens": 40,
|
7 |
"transformers_version": "4.42.3"
|
8 |
}
|
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 3422777952
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7108f45b86e4b0c9b0c8f611264b381522e093b366ec0fdf83c0e07149b0c398
|
3 |
size 3422777952
|
onnx/model.onnx
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 sha256:5e2b596dd6bd22b4337eaf3bf4161313306dfc61d12227dab61a28a1afecd95e
|
3 |
+
size 562234
|
onnx/model.onnx_data
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 6845505536
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:43c2ff1f9af866f2fc88d29d9ef6c283c2c3908908bfc45e8b7acb948a4b41bc
|
3 |
size 6845505536
|
onnx/model_bnb4.onnx
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:a632c41a9cd1bcd7116a75ce4c77c290b1e65532f001107d873c4759462a4d6b
|
3 |
+
size 1309599186
|
onnx/model_fp16.onnx
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:a38d5ebf9f8d3556de0aa6b5c7fde73bbd75b89b603f03eadc5903fb47dac346
|
3 |
+
size 578785
|
onnx/model_fp16.onnx_data
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:bfb0a58ad2b44ced92f2cb3b5c3c6554b007d210d4fd45b090d0e78156aeb292
|
3 |
+
size 3422752768
|
onnx/model_int8.onnx
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:c7bcbc362755972f15beb94e1d8a9e4d83e8ccf08e669d40c668061e4d712ac0
|
3 |
+
size 1712409483
|
onnx/model_q4.onnx
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:e80eb151f4c5c50e842bbf338a774436d0f4c3e3a81176d4bfc6760853d8a7a6
|
3 |
+
size 1410261138
|
onnx/model_q4f16.onnx
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 sha256:54ca7d257e47e478ecb9f1306399aa111305556eabac5485c470294df3273cf9
|
3 |
+
size 1108085974
|
onnx/model_quantized.onnx
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:c7bcbc362755972f15beb94e1d8a9e4d83e8ccf08e669d40c668061e4d712ac0
|
3 |
+
size 1712409483
|
onnx/model_uint8.onnx
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:b28a17162551f5165479e10819627426cdb82d3e9dfff339a2d5f651d6a6429c
|
3 |
+
size 1712409564
|
runs/Jul15_23-01-52_ip-26-0-171-168/events.out.tfevents.1721084661.ip-26-0-171-168.1537695.0
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:f9e82e0a6588823112621478b95e6fc4db5d7bb0750bbc260d45a4f6d5eaa35c
|
3 |
-
size 10294
|
|
|
|
|
|
|
|
runs/Jul15_23-01-52_ip-26-0-171-168/events.out.tfevents.1721085183.ip-26-0-171-168.1537695.1
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:27a6a863f4bf58d6242e050e5e56402f3611cf866ad775259a318e644ad89c3d
|
3 |
-
size 815
|
|
|
|
|
|
|
|
runs/Jul15_23-02-15_ip-26-0-175-34/events.out.tfevents.1721084673.ip-26-0-175-34.1294397.0
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:fee64781b1f43c757e014bacbcdc4b210d3b76ebe5fc1f4a63a28d24172c2419
|
3 |
-
size 10293
|
|
|
|
|
|
|
|
runs/Jul15_23-02-15_ip-26-0-175-34/events.out.tfevents.1721085077.ip-26-0-175-34.1294397.1
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:038d227d30387508a3dd8bb86a277ae6c97ddc940005a81c01cc74689ec17040
|
3 |
-
size 815
|
|
|
|
|
|
|
|
train_results.json
DELETED
@@ -1,9 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"epoch": 0.9872958257713249,
|
3 |
-
"total_flos": 0.0,
|
4 |
-
"train_loss": 0.6753217893488267,
|
5 |
-
"train_runtime": 321.7901,
|
6 |
-
"train_samples": 8806,
|
7 |
-
"train_samples_per_second": 27.366,
|
8 |
-
"train_steps_per_second": 0.211
|
9 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
trainer_state.json
DELETED
@@ -1,147 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"best_metric": null,
|
3 |
-
"best_model_checkpoint": null,
|
4 |
-
"epoch": 0.9872958257713249,
|
5 |
-
"eval_steps": 100,
|
6 |
-
"global_step": 68,
|
7 |
-
"is_hyper_param_search": false,
|
8 |
-
"is_local_process_zero": true,
|
9 |
-
"is_world_process_zero": true,
|
10 |
-
"log_history": [
|
11 |
-
{
|
12 |
-
"epoch": 0.014519056261343012,
|
13 |
-
"grad_norm": 2.5819620488760253,
|
14 |
-
"learning_rate": 7.142857142857143e-07,
|
15 |
-
"logits/chosen": 0.902682900428772,
|
16 |
-
"logits/rejected": 0.8661031723022461,
|
17 |
-
"logps/chosen": -125.2274169921875,
|
18 |
-
"logps/rejected": -145.89674377441406,
|
19 |
-
"loss": 0.6931,
|
20 |
-
"rewards/accuracies": 0.0,
|
21 |
-
"rewards/chosen": 0.0,
|
22 |
-
"rewards/margins": 0.0,
|
23 |
-
"rewards/rejected": 0.0,
|
24 |
-
"step": 1
|
25 |
-
},
|
26 |
-
{
|
27 |
-
"epoch": 0.14519056261343014,
|
28 |
-
"grad_norm": 3.2242486672937045,
|
29 |
-
"learning_rate": 4.970219740227693e-06,
|
30 |
-
"logits/chosen": 0.8677890300750732,
|
31 |
-
"logits/rejected": 0.88941490650177,
|
32 |
-
"logps/chosen": -120.65478515625,
|
33 |
-
"logps/rejected": -149.0773468017578,
|
34 |
-
"loss": 0.6925,
|
35 |
-
"rewards/accuracies": 0.4097222089767456,
|
36 |
-
"rewards/chosen": 0.0005513947107829154,
|
37 |
-
"rewards/margins": -0.0005504629807546735,
|
38 |
-
"rewards/rejected": 0.001101857633329928,
|
39 |
-
"step": 10
|
40 |
-
},
|
41 |
-
{
|
42 |
-
"epoch": 0.29038112522686027,
|
43 |
-
"grad_norm": 4.0763010346326904,
|
44 |
-
"learning_rate": 4.460299516441777e-06,
|
45 |
-
"logits/chosen": 1.0478929281234741,
|
46 |
-
"logits/rejected": 1.0530675649642944,
|
47 |
-
"logps/chosen": -115.20359802246094,
|
48 |
-
"logps/rejected": -142.4711456298828,
|
49 |
-
"loss": 0.6887,
|
50 |
-
"rewards/accuracies": 0.543749988079071,
|
51 |
-
"rewards/chosen": -0.007147049065679312,
|
52 |
-
"rewards/margins": 0.007046631071716547,
|
53 |
-
"rewards/rejected": -0.014193681068718433,
|
54 |
-
"step": 20
|
55 |
-
},
|
56 |
-
{
|
57 |
-
"epoch": 0.4355716878402904,
|
58 |
-
"grad_norm": 2.868995154924774,
|
59 |
-
"learning_rate": 3.441819734087963e-06,
|
60 |
-
"logits/chosen": 1.0493364334106445,
|
61 |
-
"logits/rejected": 1.0137263536453247,
|
62 |
-
"logps/chosen": -115.47831726074219,
|
63 |
-
"logps/rejected": -145.4583282470703,
|
64 |
-
"loss": 0.681,
|
65 |
-
"rewards/accuracies": 0.53125,
|
66 |
-
"rewards/chosen": -0.013122054748237133,
|
67 |
-
"rewards/margins": 0.018696457147598267,
|
68 |
-
"rewards/rejected": -0.031818509101867676,
|
69 |
-
"step": 30
|
70 |
-
},
|
71 |
-
{
|
72 |
-
"epoch": 0.5807622504537205,
|
73 |
-
"grad_norm": 2.4635583297317165,
|
74 |
-
"learning_rate": 2.1790041121336223e-06,
|
75 |
-
"logits/chosen": 1.2130175828933716,
|
76 |
-
"logits/rejected": 1.1768478155136108,
|
77 |
-
"logps/chosen": -154.19398498535156,
|
78 |
-
"logps/rejected": -181.88272094726562,
|
79 |
-
"loss": 0.6762,
|
80 |
-
"rewards/accuracies": 0.581250011920929,
|
81 |
-
"rewards/chosen": -0.04483573138713837,
|
82 |
-
"rewards/margins": 0.041850604116916656,
|
83 |
-
"rewards/rejected": -0.08668633550405502,
|
84 |
-
"step": 40
|
85 |
-
},
|
86 |
-
{
|
87 |
-
"epoch": 0.7259528130671506,
|
88 |
-
"grad_norm": 3.188683944287027,
|
89 |
-
"learning_rate": 9.994642986290797e-07,
|
90 |
-
"logits/chosen": 1.3518016338348389,
|
91 |
-
"logits/rejected": 1.2714142799377441,
|
92 |
-
"logps/chosen": -131.69985961914062,
|
93 |
-
"logps/rejected": -161.8300018310547,
|
94 |
-
"loss": 0.6622,
|
95 |
-
"rewards/accuracies": 0.6499999761581421,
|
96 |
-
"rewards/chosen": -0.011605720035731792,
|
97 |
-
"rewards/margins": 0.06830872595310211,
|
98 |
-
"rewards/rejected": -0.07991443574428558,
|
99 |
-
"step": 50
|
100 |
-
},
|
101 |
-
{
|
102 |
-
"epoch": 0.8711433756805808,
|
103 |
-
"grad_norm": 3.3445679325076476,
|
104 |
-
"learning_rate": 2.092077387824884e-07,
|
105 |
-
"logits/chosen": 1.367594838142395,
|
106 |
-
"logits/rejected": 1.2992292642593384,
|
107 |
-
"logps/chosen": -121.10174560546875,
|
108 |
-
"logps/rejected": -139.6844940185547,
|
109 |
-
"loss": 0.6612,
|
110 |
-
"rewards/accuracies": 0.612500011920929,
|
111 |
-
"rewards/chosen": -0.017885591834783554,
|
112 |
-
"rewards/margins": 0.07065734267234802,
|
113 |
-
"rewards/rejected": -0.08854293823242188,
|
114 |
-
"step": 60
|
115 |
-
},
|
116 |
-
{
|
117 |
-
"epoch": 0.9872958257713249,
|
118 |
-
"step": 68,
|
119 |
-
"total_flos": 0.0,
|
120 |
-
"train_loss": 0.6753217893488267,
|
121 |
-
"train_runtime": 321.7901,
|
122 |
-
"train_samples_per_second": 27.366,
|
123 |
-
"train_steps_per_second": 0.211
|
124 |
-
}
|
125 |
-
],
|
126 |
-
"logging_steps": 10,
|
127 |
-
"max_steps": 68,
|
128 |
-
"num_input_tokens_seen": 0,
|
129 |
-
"num_train_epochs": 1,
|
130 |
-
"save_steps": 500,
|
131 |
-
"stateful_callbacks": {
|
132 |
-
"TrainerControl": {
|
133 |
-
"args": {
|
134 |
-
"should_epoch_stop": false,
|
135 |
-
"should_evaluate": false,
|
136 |
-
"should_log": false,
|
137 |
-
"should_save": false,
|
138 |
-
"should_training_stop": false
|
139 |
-
},
|
140 |
-
"attributes": {}
|
141 |
-
}
|
142 |
-
},
|
143 |
-
"total_flos": 0.0,
|
144 |
-
"train_batch_size": 2,
|
145 |
-
"trial_name": null,
|
146 |
-
"trial_params": null
|
147 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
training_args.bin
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:c619b9db98ea959618a4839c7d36e8399c8ce4dac6b0fc49c96d5e1d3b9e1fc4
|
3 |
-
size 6520
|
|
|
|
|
|
|
|