|
--- |
|
tags: |
|
- sentence-transformers |
|
- sentence-similarity |
|
- feature-extraction |
|
- generated_from_trainer |
|
- dataset_size:892 |
|
- loss:MatryoshkaLoss |
|
- loss:MultipleNegativesRankingLoss |
|
base_model: Snowflake/snowflake-arctic-embed-l |
|
widget: |
|
- source_sentence: 'content=''1. What is the relationship between casting cooling |
|
rate and the formation of the Mg17A12 phase in the context of dynamic recrystallization |
|
(DRX)?\n2. How does the ANN model contribute to predicting hardness variations |
|
in AZ80 cast-forged I-beams under different casting and forging conditions?'' |
|
additional_kwargs={''refusal'': None} response_metadata={''token_usage'': {''completion_tokens'': |
|
61, ''prompt_tokens'': 752, ''total_tokens'': 813, ''completion_tokens_details'': |
|
{''accepted_prediction_tokens'': 0, ''audio_tokens'': 0, ''reasoning_tokens'': |
|
0, ''rejected_prediction_tokens'': 0}, ''prompt_tokens_details'': {''audio_tokens'': |
|
0, ''cached_tokens'': 0}}, ''model_name'': ''gpt-4o-mini-2024-07-18'', ''system_fingerprint'': |
|
''fp_7fcd609668'', ''finish_reason'': ''stop'', ''logprobs'': None} id=''run-acea7371-1790-4bcc-a817-7ecc5812a890-0'' |
|
usage_metadata={''input_tokens'': 752, ''output_tokens'': 61, ''total_tokens'': |
|
813, ''input_token_details'': {''audio'': 0, ''cache_read'': 0}, ''output_token_details'': |
|
{''audio'': 0, ''reasoning'': 0}}' |
|
sentences: |
|
- 'denoises the data [82, 83, 84, 85]. During the forward process, the noise is |
|
added incrementally through numerous steps in a Markovian process: |
|
|
|
|
|
𝑞(𝑥𝑡|𝑥𝑡−1) ≔ 𝒩(𝑥𝑡; √1 − 𝛽𝑡𝑥𝑡−1, 𝛽𝑡𝑰) (3) |
|
|
|
|
|
Where 𝑥0 is a data sampled from the real data distribution 𝑞(𝑥) (i.e., 𝑥0 ∼ 𝑞(𝑥)). |
|
𝛽t ∈ (0,1) is the variance schedule (𝛽0 is small, and 𝛽T is large), and 𝑰 is the |
|
identity matrix. By using the notation 𝛼t ∶= 1 − 𝛽t and 𝛼t ∶= ∏st=0𝛼s, the noised |
|
data at any arbitrary step can be found: |
|
|
|
|
|
𝑞 (𝑥t|𝑥t−1) |
|
≔ 𝒩(𝑥t; √̅𝛼t𝑥0, (1 − 𝛼t)𝑰) ̅ (4) |
|
|
|
|
|
With reparameterization, 𝑥 can be written as: |
|
|
|
|
|
𝑥t = √̅𝛼t𝑥0 + √1 − 𝛼t𝝐 ̅ (5) |
|
|
|
|
|
Where 𝜖 ~ 𝒩(𝜖|0, 𝑰). |
|
|
|
|
|
During the backward process, the goal is to denoise images iteratively to get |
|
back to a less noisy version: |
|
|
|
|
|
𝑝 (𝑥t−1|𝑥t) |
|
≔ 𝒩(𝑥t−1; 𝜇t(𝑥t, 𝑡), 𝜎t𝑰) ̃ (6) |
|
|
|
|
|
The reverse conditional shown in equation 6 does not have a closed form, but the |
|
following approximation is derived for that [82]: |
|
|
|
|
|
̃𝜇t(𝑥t, 𝑡) ≈ 1 𝛼t(𝑥t |
|
− √1−𝛼t 1−𝛼t𝜖t) (7) |
|
|
|
|
|
Where 𝜖t is the noise introduced in the step 𝑡. The main idea is to learn this |
|
distribution, so we can denoise images by the reverse conditional. In order to |
|
do this, we train a neural network 𝜖𝜃(𝑥t, 𝑡) to approximate it: |
|
|
|
|
|
𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝐿(𝜃) = ‖𝜖t− 𝜖𝜃(𝑥t, 𝑡)‖2 2 (8) |
|
|
|
|
|
In this work, a U-net architecture, which is first introduced by Ronneberger et |
|
al. [86] for image segmentation. Due to its symmetric down-sampling and up-sampling, |
|
this architecture is suitable for going from a noisy image to its equivalent denoised |
|
version. The encoding part consists of five blocks of max-pooling followed by |
|
several convolutional layers, group normalization, and GELU activation function. |
|
The self-attention mechanism with four heads is applied after each down-sampling |
|
layer. The decoding part is a somewhat mirrored structure of the encoding. It |
|
consists of five blocks, each one of them has bi-linear up-sampling followed by |
|
several convolutional layers, group normalization, and GELU activation function. |
|
The only difference is that after each up-sampling layer the data is concatenated |
|
with skip connections corresponding to channels with the same resolution from |
|
the encoding part. The model is conditioned on the time step by the Transformer |
|
positional embedding. |
|
|
|
|
|
38' |
|
- '# Figure 39 |
|
|
|
|
|
SEM images of the different morphologies of the intermetallic compounds present |
|
in the cast-forged I-beams produced in the range of casting and forging conditions |
|
investigated in this study. a) Eutectic Mg17Al12, b) Lamellar Mg17Al12, c) Globular |
|
and Continuous Mg17Al12, and d) Intergranular Mg17Al12 and AlMn5. The casting |
|
cooling condition, forging temperature, and the location of each sample is depicted |
|
at the top right corner of each image. |
|
|
|
|
|
The microstructures of the I-beams exhibit significant variation due to different |
|
casting and forging conditions, as well as the evolution of Mg17Al12 phase. Figure |
|
40 and Figure 41 show the microstructures of the I-beams obtained from the web |
|
region and short flange, respectively, under different casting cooling rates and |
|
forging temperatures. Notably, the characteristics of DRX, including the extent |
|
of DRX, its distribution, and DRX grain size vary among the investigated samples. |
|
For the samples forged at 350°C, both the web region and short flange display |
|
a completely coverage of DRXed grains, indicating extensive recrystallization |
|
(see Figure 40a-b and Figure 41a-b). Conversely, a bimodal microstructure is observed |
|
in the remaining samples. At the web region of the samples forged at 250 and 300°C, |
|
the deformation and subsequent DRX has resulted in a distinctive “necklace” structure, |
|
characterized by initial elongated grains surrounded by DRXed grains (see Figure |
|
40c-f). At the short flange of the samples forged at 250 and 300°C, DRX is limited |
|
to the vicinity of the eutectic Mg17Al12 and the globular particles formed around |
|
them (see Figure 41c-f).' |
|
- '# Cooling Rate |
|
|
|
|
|
# Forging |
|
|
|
|
|
|Temperature|Measured|Predicted|Measured|Predicted|Measured|Predicted| | | |
|
|
|
|---|---|---|---|---|---|---|---|---| |
|
|
|
|250|47.58 + 2.70|46.31 + 1.69|47.50 + 2.59|47.56 + 2.34|43.73 + 3.29|44.01 + |
|
1.90|43.32 + 2.73|42.77 + 2.23| |
|
|
|
|NRMSE:|0.0501|0.0416| |0.0503| |0.0555| | | |
|
|
|
|300|40.63 + 2.25|39.93 + 1.42|40.48 + 2.19|41.82 + 1.90|38.80 + 2.63|39.27 + |
|
1.57|37.89 + 2.82|38.24 + 1.76| |
|
|
|
|NRMSE:|0.0468|0.0571| |0.0597| |0.0688| | | |
|
|
|
|350|35.46 + 3.12|33.89 + 1.49|34.75 + 2.69|35.49 + 1.78|34.16 + 2.16|32.94 + |
|
1.05|32.57 + 2.24|33.42 + 1.20| |
|
|
|
|NRMSE:|0.0933|0.0891| |0.0734| |0.0826| | | |
|
|
|
|
|
Figure 55: The predicted and actual contour maps and average HR30T hardness values |
|
with standard deviation for different combinations of the casting cooling rate |
|
and the forging temperature in the one-condition-out-scenario. NRMSE values for |
|
each prediction are also provided. |
|
|
|
|
|
The effect of the casting cooling rate is also related to the occurrence of DRX. |
|
Higher cooling rates result in finer α-Mg dendritic structure and prevent the |
|
formation of the lamellar Mg17A12 phase [127]. This can result in a more uniform |
|
distribution of the Mg17A12 intermetallic particles, which promote dynamic recrystallization |
|
by particle stimulated nucleation (PSN) mechanism [36]. Consequently, the samples |
|
forged from the casts with higher cooling rates can have more DRXed regions, resulting |
|
in slightly higher hardness. |
|
|
|
|
|
The ANN model accurately captures these hardness variations which are controlled |
|
by the evolution of several different microstructural features, even in prediction |
|
of completely unseen conditions. This model can be used in the range of studied |
|
casting and forging conditions to produce reliable hardness distribution contour |
|
maps and predict the average hardness of the AZ80 cast-forged I-beam. |
|
|
|
|
|
This model can be extended by including the effective local strain at each measurement |
|
point instead of the x and y coordinates of the measurement point, so that the |
|
model can be generalized for any arbitrary forged geometry. In addition, the casting |
|
cooling rate parameter can be replaced by a microstructure feature, such as secondary |
|
dendrite arm spacing (SDAS), to generalize the model for varied casting methods. |
|
On the other hand, the benefit of the current model is that it needs no further |
|
measurement or observation and produces hardness distribution contour maps from |
|
known processing parameters.' |
|
- source_sentence: 'content=''1. What does a lower relative CRSS ratio indicate in |
|
terms of activation with reference to the basal slip?\n2. How does the initial |
|
texture affect material flow during deformation in AZ31 alloy?'' additional_kwargs={''refusal'': |
|
None} response_metadata={''token_usage'': {''completion_tokens'': 40, ''prompt_tokens'': |
|
844, ''total_tokens'': 884, ''completion_tokens_details'': {''accepted_prediction_tokens'': |
|
0, ''audio_tokens'': 0, ''reasoning_tokens'': 0, ''rejected_prediction_tokens'': |
|
0}, ''prompt_tokens_details'': {''audio_tokens'': 0, ''cached_tokens'': 0}}, ''model_name'': |
|
''gpt-4o-mini-2024-07-18'', ''system_fingerprint'': ''fp_709714d124'', ''finish_reason'': |
|
''stop'', ''logprobs'': None} id=''run-7c6ab37c-114d-4658-9cfd-da50507b31bf-0'' |
|
usage_metadata={''input_tokens'': 844, ''output_tokens'': 40, ''total_tokens'': |
|
884, ''input_token_details'': {''audio'': 0, ''cache_read'': 0}, ''output_token_details'': |
|
{''audio'': 0, ''reasoning'': 0}}' |
|
sentences: |
|
- '# 2.3.2 CRSS in the present case |
|
|
|
|
|
(b) relative CRSS ratio (lower values indicate an ease in activation with reference |
|
to the basal slip) [63] |
|
|
|
|
|
# Figures |
|
|
|
|
|
# Figure 2-12 |
|
|
|
|
|
Effect of starting grain size on DRX fraction and DRX grain size in an extruded |
|
AZ31 alloy deformed in uniaxial compression at 300 °C, 0.01 s-1 to a strain of |
|
0.5. Adapted from [66] |
|
|
|
|
|
# Figure 2-13 |
|
|
|
|
|
Effect of initial texture on material flow (a) symmetric distribution of the starting |
|
texture results in a symmetric material flow during deformation, (b) asymmetric |
|
distribution of starting texture with respect to compression direction (CD) results |
|
in a preferred material flow along ND. Deformation was done using symmetrical |
|
cylindrical samples (with cylindrical axis along the CD). CD: Compression direction, |
|
ND: Normal direction, TD: Transverse direction, ED: Extrusion direction. Adapted |
|
from [42] |
|
|
|
|
|
# Figure 2-14 |
|
|
|
|
|
Effect of starting texture on (a) DRX fraction, (b) DRX grain size. AZ31 alloy |
|
deformed in plane strain compression, in temperature range of 300 °C– 400 °C, |
|
and strain rate range of 0.001 s-1 – 1 s-1. Adapted from [77] |
|
|
|
|
|
# Figure 2-15 |
|
|
|
|
|
Effect of Al alloying in polycrystalline Mg on the CRSS for various deformation |
|
modes at: (a) Room temperature, (b) 200 °C. Twinning refer to {1012} tensile twinning. |
|
X-axis shows concentration of Al in square root of wt.% [82] |
|
|
|
|
|
# Figure 2-16 |
|
|
|
|
|
Equilibrium phase fraction map of AZ80 alloy [93] |
|
|
|
|
|
# Figure 2-17 |
|
|
|
|
|
Different Mg17Al12 precipitate morphologies in AZ alloys: (a) DP in lamellar form, |
|
(b) DP in ellipsoidal particle (globular) form, (c) CP in form of laths, and (d) |
|
CP in form of irregular plates. (a) and (b) are for AZ80 [92], while (c) and (d) |
|
are for AZ91 [91]. |
|
|
|
|
|
# Figure 2-18 |
|
|
|
|
|
(a-c) hot deformation of a cast-homogenized and pre-aged AZ91 sample at 300 °C, |
|
0.2 s-1, showing the occurrence of DRX only in regions previously occupied by |
|
DP [24]; (d-f) hot deformation of a cast-homogenized AZ91 sample at 400 °C, 0.2 |
|
s-1. At 400 °C, CP got precipitated in the homogenized material during heating |
|
to and holding at the test temperature. (e-f) shows extensive DRX occurred in |
|
the material in this case as well [23]. (a) and (d) correspond to non-deformed |
|
samples, while (b, c, e, and f) correspond to deformed samples. C and D in (b) |
|
and (c) refer to continuous precipitates and discontinuous precipitates, respectively. |
|
|
|
|
|
# Figure 2-19 |
|
|
|
|
|
Variation in (a) DRX ratio and (b) DRX grain size, with the Z parameter. Cast |
|
and extruded AZ31 alloys deformed in uniaxial compression for temperatures in |
|
the range of 300 - 450 °C, and strain rates in the range of 0.01 – 1 s-1. For |
|
the data presented in the graphs, DRXed grains were distinguished from the parent |
|
grains based on their size and shape [67]. |
|
|
|
|
|
# Figure 2-20 |
|
|
|
|
|
Change of DRX mechanism with deformation conditions for deformation of pure Mg. |
|
In area marked as I, DRX takes place via the twinning mechanism [35]' |
|
- '# 9.3.2. Microstructure to mechanical properties prediction model |
|
|
|
|
|
.......................................... 132 |
|
|
|
|
|
# 9.3.3. Evaluation |
|
|
|
|
|
......................................................................................................................... |
|
133 |
|
|
|
|
|
# 9.4. Results and discussion |
|
|
|
|
|
............................................................................................................. |
|
134 |
|
|
|
|
|
# 9.4.1. Microstructure and quasi-static tensile test of the cast-forged AZ80 |
|
magnesium alloy |
|
|
|
|
|
…134 |
|
|
|
|
|
# 9.4.2. Comparison of data balancing strategies |
|
|
|
|
|
........................................................................ 135 |
|
|
|
|
|
# 9.4.3. Mechanical properties prediction |
|
|
|
|
|
.................................................................................... |
|
136 |
|
|
|
|
|
# 9.5. Conclusion |
|
|
|
|
|
............................................................................................................................... |
|
139 |
|
|
|
|
|
# 10. Summary, Conclusions, limitations, and Future Work |
|
|
|
|
|
.......................................................... 141 |
|
|
|
|
|
# 10.1. Summary |
|
|
|
|
|
................................................................................................................................ |
|
141 |
|
|
|
|
|
# 10.2. Conclusions |
|
|
|
|
|
............................................................................................................................ |
|
141 |
|
|
|
|
|
# 10.3. Limitations |
|
|
|
|
|
............................................................................................................................. |
|
145 |
|
|
|
|
|
# References |
|
|
|
|
|
........................................................................................................................................... |
|
148' |
|
- '# Characterization of forged magnesium alloys |
|
|
|
|
|
# 3.4 Case study |
|
|
|
|
|
In industrially forged parts, control of forging parameters, such as temperature |
|
and cooling rate is often challenging compared to laboratory-scale studies involving |
|
uniaxial compression. As a result, these variations can significantly impact the |
|
microstructure and texture development. Furthermore, industrial forgings typically |
|
exhibit larger dimensions and involve more complex material flow in closed dies |
|
further influencing the microstructure and texture development during forging. |
|
|
|
|
|
To exemplify these points, we refer to the work of Jahed et al. who studied the |
|
forging of an I-beam at the University of Waterloo by using a cast AZ31B alloy |
|
billet. The investigation served as a precursor study before the forging of a |
|
more geometrically complex automotive suspension component (Gryguc et al., 2021a, |
|
2021b, 2021c; Toscano et al., 2018a, 2018b; Williams et al., 2021). The geometry |
|
was forged using a starting cylindrical billet with a diameter of 63.5 mm and |
|
length of 65 mm machined from an original cast AZ31B billet of 300 mm diameter |
|
(Toscano et al., 2018a, 2018b). |
|
|
|
|
|
Forging was carried out using a 500-ton hydraulic press at 250 °C at a ram displacement |
|
rate of 20 mm/s, equivalent to an initial strain rate of about 0.3 s−1. The dies |
|
were pre-heated to the same forging temperature. Graphite was used as a lubricant. |
|
Forging was carried out in a single step. After forging, the forged part was allowed |
|
to cool by itself in the air. Subsequently, the part was longitudinally sectioned |
|
along mid-plane (refer to Fig. 36c) and prepared for metallographic examination |
|
using the standard practice. Microstructure and texture were observed at three |
|
different locations, as marked in Fig. 36c. |
|
|
|
|
|
The texture was measured using a Brucker D8 Discover diffractometer using the |
|
process outlined in Pahlevanpour et al. (2018) and presented here based on the |
|
(0002) basal pole figures. Surface hardness measurements on the entire prepared |
|
cross-section were carried out using a United True Blue II Rockwell Hardness Tester |
|
using Rockwell 15T scale and a 1/16 ball indenter. |
|
|
|
|
|
The results presented in Fig. 36 show significantly different microstructures |
|
and textures at the three locations marked in Fig. 36c. This difference is attributed |
|
to local variations in strain and strain rate gradients, which are significantly |
|
impacted by the die geometry as it constrains the material flow. Additionally, |
|
the DRX grain size from the microstructures presented here is coarser than what |
|
is expected from a specimen-size hot compression test, which can be attributed |
|
to a slow cooling rate post-forging (Beer and Barnett, 2008). |
|
|
|
|
|
The differences in microstructure and texture are also reflected in the variation |
|
in the hardness map presented in Fig. 36c. Gryguc et al., have shown that such |
|
spatial variations in microstructure and texture in a forged part also result |
|
in variations in strength and ductility (Gryguc et al., 2018). Other authors have |
|
also reported similar spatial variations in microstructure, texture, and mechanical |
|
properties in industrially relevant magnesium forged parts. Specifically, such |
|
variations were reported in aircraft brackets forged using AZ31 alloy and automotive |
|
wheel hubs forged using AZ80 alloy (Dziubińska et al., 2015; Liao et al., 2023). |
|
|
|
|
|
Due to these variations in mechanical properties, it is crucial to assess them |
|
individually in all critical load areas when forging a new part, rather than relying |
|
on single measurements assuming it represents the overall behavior of the bulk |
|
material.' |
|
- source_sentence: 'content=''1. What was the purpose of using graphite lubricate |
|
on the die and billet during the forging process? \n2. How were the tensile monotonic |
|
and fatigue test samples extracted from the closed die forged components? '' |
|
additional_kwargs={''refusal'': None} response_metadata={''token_usage'': {''completion_tokens'': |
|
44, ''prompt_tokens'': 751, ''total_tokens'': 795, ''completion_tokens_details'': |
|
{''accepted_prediction_tokens'': 0, ''audio_tokens'': 0, ''reasoning_tokens'': |
|
0, ''rejected_prediction_tokens'': 0}, ''prompt_tokens_details'': {''audio_tokens'': |
|
0, ''cached_tokens'': 0}}, ''model_name'': ''gpt-4o-mini-2024-07-18'', ''system_fingerprint'': |
|
''fp_709714d124'', ''finish_reason'': ''stop'', ''logprobs'': None} id=''run-dc48d253-0c68-472b-9ba4-c148e6dc28c8-0'' |
|
usage_metadata={''input_tokens'': 751, ''output_tokens'': 44, ''total_tokens'': |
|
795, ''input_token_details'': {''audio'': 0, ''cache_read'': 0}, ''output_token_details'': |
|
{''audio'': 0, ''reasoning'': 0}}' |
|
sentences: |
|
- 'microstructure at the end of the forging stage. Because of the I-beam geometry, |
|
the effective strain distribution is nonuniform across the forged sample. The |
|
middle part of the I-beam component, referred to as the web region, undergoes |
|
a higher extent of deformation [63, 39]. The grains are therefore severely elongated |
|
in the web region. The higher effective strain results in an increase in the hardness |
|
due to work hardening. Moreover, higher deformation promotes DRX, which refines |
|
grain size and further increases the material hardness. |
|
|
|
|
|
# Table 5: Chemical composition of the cast AZ80. |
|
|
|
|
|
|Element|Mg|Al|Zn|Cu|Fe|Mn|Ni| |
|
|
|
|---|---|---|---|---|---|---|---| |
|
|
|
|Wt.%|Bal.|8.64|0.52|0.0017|0.0044|0.24|0.0012| |
|
|
|
|
|
Figure 49: SEM microstructure images of the AZ80 magnesium alloy cast at a cooling |
|
rate of: a) 1.5 °C/s and b) 6 °C/s. |
|
|
|
|
|
Typical representative alterations in the forged microstructure due to casting |
|
and forging process parameters are shown in Figure 50. Depending on the forging |
|
temperature, some dissolution of the eutectic and precipitated Mg17Al12 phase |
|
might occur during forging. Figure 50a and Figure 50b compare the effect of forging |
|
temperature on the extent of dissolution of β-Mg17Al12 particles during forging |
|
for samples cast at 1.5 °C/s cooling rate and forged at 300 °C and 350 °C, respectively. |
|
Moreover, the Mg17Al12 intermetallic can precipitate in discontinuous and/or continuous |
|
morphologies [153]. Figure 50c and Figure 50d show two different morphologies |
|
of precipitated β-Mg17Al12 formed in samples cast at 6 °C/s cooling rate and forged |
|
at 250 °C and 300 °C, respectively. In addition to the direct effect of intermetallic |
|
compounds on the material hardness [146], the β-Mg17Al12 morphologies are shown |
|
to control microstructure evolution and hardness via affecting DRX [44]. |
|
|
|
|
|
DRX occurs for all the studied cast-forge conditions; however, the extent and |
|
the size of DRXed grains vary for different samples. Both eutectic and precipitated |
|
Mg17Al12, as well as the higher amount of deformation promote DRX [44]. The size |
|
of DRXed grains is also controlled by deformation temperature [36] and the presence |
|
of fine intermetallic particles. Figure 50e and Figure 50f illustrate the effect |
|
of deformation temperature on the grain size of DRXed grains. Forging the sample |
|
cast at 1.5' |
|
- 'Metals 2021, 11, 1290 lower control arm (step 4). Graphite lubricate was used |
|
on the die and billet to reduce friction, sticking and promote material flow. |
|
The orientation of the billet to the press was such that the radial direction |
|
was along the direction of the press stroke (i.e., the direction of forging was |
|
parallel to the radial direction of the billet). Forging was carried out in a |
|
single step at a displacement rate of 2.1–4.2 mm/s, and they were allowed to air |
|
cool with no additional post-heat treatments. |
|
|
|
|
|
# Figure 1. Plan view of extruded billet during different stages of forging operation |
|
to achieve final component (prior to machining of the flash). |
|
|
|
|
|
Tensile monotonic and fatigue test samples were utilized with geometries according |
|
to Roostaei et al. [25] and a 4 mm thickness, these were machined from various |
|
locations throughout closed die forged components. There were 18 test samples |
|
that were extracted from 18 different locations throughout each of the closed |
|
die forging (see Figure 2), all with varying thermomechanical histories. One component |
|
forged at each temperature was dedicated to stress-controlled fatigue testing |
|
where 10 samples are presented for the 300 °C forging and 8 for the 450 °C forging. |
|
Strain-controlled tests were performed on 14 different samples extracted from |
|
another component forged at 300 °C. The remainder of the original 18 test samples |
|
from each component were used for monotonic testing to understand the spatial |
|
variation in strength and ductility. The test samples had a nominal surface finish |
|
of Ra ≤ 0.2 μm; however, in practice, the actual roughness was substantially less, |
|
around 0.05 μm. These samples were later utilized for quasi-static and cyclic |
|
(stress-controlled) testing. |
|
|
|
|
|
The quasi-static tensile tests were performed according to ASTM standard E8/E8M-15a |
|
using an MTS 810 Servo-Hydraulic test machine operating in displacement control |
|
mode with a displacement rate of 1 mm/min. Strain measurement was accomplished |
|
using a GOM ARAMIS 3D 5MP DIC system which passively functioned to measure the |
|
average axial strain on the surface of the gauge section of the sample throughout |
|
the duration of the test. The fatigue tests were performed as per ASTM E466-15 |
|
in an ambient environment using an MTS 810 Servo-Hydraulic test machine operating |
|
in stress control mode at a frequency range of 0.5 Hz to 60 Hz depending on the |
|
stress amplitude to maintain an approximately consistent loading rate between |
|
all tests. |
|
|
|
|
|
The strain was measured throughout the first 10,000 cycles using an MTS 632.26 |
|
extensometer with an 8-mm gauge and travel of ±1.2 mm until stabilization of the |
|
cyclic hysteresis loop was achieved. The tests were conducted at zero mean stress |
|
(i.e., RL = −1, fully reversed stress cycle) and stress amplitudes of between |
|
140 MPa and 180 MPa. The failure criteria for the tests were considered to be |
|
the final rupture of the specimen gauge section. |
|
|
|
|
|
# Figure 2. Sample extraction schematic from AZ80 Mg die-forged component. |
|
|
|
|
|
# 3. Results and Discussion |
|
|
|
|
|
# 3.1. Structure-Property Relationship (Effect of Material Modification) |
|
|
|
|
|
Some of the favourable aspects of selecting forging as a “near net-shape” manufacturing |
|
process include...' |
|
- 'The effect of initial cooling rate is apparent. Material cast with a higher cooling |
|
rate shows a higher peak compressive stress than material cast with a lower cooling |
|
rate, though this effect is most noticeable at low temperatures and/or high strain |
|
rates. Following the peak stress, the HCR samples experienced a higher rate of |
|
dynamic softening, lowering the stress to a similar value as the LCR samples. |
|
At high strains, after significant deformation has occurred, there is no effect |
|
of initial cast cooling rate. The peak compressive stress of all deformation conditions |
|
that were tested is shown in Figure 30. While some conditions show a higher difference |
|
in peak stress than others, the effect is consistent across all temperatures and |
|
strain rates. The higher flow stress that is seen in finer grained cast AZ80 has |
|
also been reported for similar alloys [100, 101]. |
|
|
|
|
|
|a)| |250 C-HCR|b)|I/s-HCR| | |
|
|
|
|---|---|---|---|---|---| |
|
|
|
|200| |250 C-LCR|200|I/s-LCR| | |
|
|
|
|180|300°C-HCR|180| |0.1/s-HCR| | |
|
|
|
| |300°C-LCR| | |0.1/s-LCR| | |
|
|
|
|160| |350 C-HCR|60|0.01/s-HCR| | |
|
|
|
| | |350 C-LCR| |0.01/s-LCR| | |
|
|
|
|140| | |9|140| | |
|
|
|
|8|120| |3|120| | |
|
|
|
|2|80|2|80| | | |
|
|
|
|60| |60| | | | |
|
|
|
|40| |40| | | | |
|
|
|
|20| |20| | | | |
|
|
|
|0.0|0.2|0.4|0.6|0.8|1.0| |
|
|
|
|
|
Figure 29: Measured compressive flow curves for degassed AZ80 cast with cooling |
|
rates of 3.4°C/s (LCR, dashed line) and 24.7°C/s (HCR, solid line) showing: a) |
|
the effect of temperature at a strain rate of 0.1s-1 and b) the effect of strain |
|
rate at a temperature of 300°C. Data from [78]. |
|
|
|
|
|
|250C-HCR|220|250C-LCR| |
|
|
|
|---|---|---| |
|
|
|
|300 C-HCR|200|300C-LCR| |
|
|
|
|350C-HCR|180|350C-LCR| |
|
|
|
|3|160| | |
|
|
|
|2|120| | |
|
|
|
|2|100| | |
|
|
|
|80| | | |
|
|
|
|0.01| | | |
|
|
|
|
|
Figure 30: Peak compressive stress at each strain rate for LCR and HCR degassed |
|
AZ80 at deformation temperatures of 250°C, 300°C, and 350°C. Data from [78]. |
|
|
|
|
|
52' |
|
- source_sentence: 'content=''1. What was the conclusion of Grygu´c et al. regarding |
|
the texture modification due to forging in AZ80 Mg materials?\n2. How does the |
|
forging temperature affect the yield strength of AZ80 Mg according to the studies |
|
presented?'' additional_kwargs={''refusal'': None} response_metadata={''token_usage'': |
|
{''completion_tokens'': 49, ''prompt_tokens'': 855, ''total_tokens'': 904, ''completion_tokens_details'': |
|
{''accepted_prediction_tokens'': 0, ''audio_tokens'': 0, ''reasoning_tokens'': |
|
0, ''rejected_prediction_tokens'': 0}, ''prompt_tokens_details'': {''audio_tokens'': |
|
0, ''cached_tokens'': 0}}, ''model_name'': ''gpt-4o-mini-2024-07-18'', ''system_fingerprint'': |
|
''fp_7fcd609668'', ''finish_reason'': ''stop'', ''logprobs'': None} id=''run-36dd8d40-0259-4d26-8db7-9eb705f3210d-0'' |
|
usage_metadata={''input_tokens'': 855, ''output_tokens'': 49, ''total_tokens'': |
|
904, ''input_token_details'': {''audio'': 0, ''cache_read'': 0}, ''output_token_details'': |
|
{''audio'': 0, ''reasoning'': 0}}' |
|
sentences: |
|
- 'Metals 2021, 11, 1290 |
|
|
|
|
|
(T5) treatment, with promising results [30]. In general, a significant component |
|
of the variability observed in the properties of the as-forged components stems |
|
from the air-cooling following forging, and the resulting variability in local |
|
cooling rates throughout the components. They also mention that the second phase |
|
sizes and morphologies which precipitate are highly dependent on the temperature |
|
and aging time, and in this study since air cooling was employed, both of these |
|
parameters would naturally spatially vary throughout the forged component and |
|
would be different in location #3 (Figure 8a) and location #13 (Figure 8b). Supporting |
|
this, is the observation that there is an apparent difference in the uniformity |
|
of the distribution in these continuous precipitates between the two locations. |
|
In location #3 (Figure 8a) the precipitate distribution is somewhat striated in |
|
the middle of the micrograph, with a non-uniform volume fraction of the second |
|
phase, as compared with location #13 (Figure 8b) which is much more uniform in |
|
nature across the entire image. Furthermore, from the higher magnification images, |
|
it can be observed that, in general, location #3 (Figure 8c) has comparatively |
|
larger precipitate sizes relative to location #13 (Figure 8d). A more refined |
|
precipitate structure with smaller spacing between particles leads to a higher |
|
propensity to resist dislocation movement when plastically deformed which is beneficial |
|
to the mechanical properties [34]. Both effects of this increase in uniformity |
|
in precipitation distribution and decrease in size in location #13 (Figure 8b) |
|
can contribute to an increase in strength, and fatigue life (as observed in the |
|
two samples tested at 160 MPa where there is a difference in life of about 6×) |
|
[1,34]. The source of this variation in life has many contributing effects, the |
|
most significant of which are: firstly, the effect of the thermal process history |
|
on the microstructure and precipitate morphology discussed here; secondly, the |
|
equivalent strain imposed by the mechanical deformation during forging which will |
|
be subsequently discussed; and finally, the well documented stochastic nature |
|
of fatigue. |
|
|
|
|
|
|T = 450 °C|y = 568.63x-0.13| |
|
|
|
|---|---| |
|
|
|
|T = 300 °C|y = 429.9x-0.088| |
|
|
|
|
|
[MPa] |
|
|
|
|
|
Amplitude cycles |
|
|
|
|
|
Stress |
|
|
|
|
|
Number of Cycles to Failure [N] |
|
|
|
|
|
|5,000|50,000|500,000| |
|
|
|
|---|---|---| |
|
|
|
|160|170|#13: 282,092| |
|
|
|
|140|130|#3: 46,428| |
|
|
|
|
|
Figure 7. Relationship between the forging temperature and stress-controlled fatigue |
|
behaviour for closed-die forged AZ80 Mg automotive suspension component. Location |
|
#3 & #13 are highlighted for reference (per Figure 2) as they are subsequently |
|
referred to in the SEM micrographs in Figure 8. |
|
|
|
|
|
Figure 8 illustrates SEM micrographs highlighting the precipitate morphology of |
|
two different locations within the forging that were tested at the same stress |
|
amplitude yet exhibited a significantly different number of cycles to failure. |
|
The precipitates are continuous in nature and range in both size and morphology, |
|
from smaller lozenge/blunted asymmetric hexagonal structure to larger slab-like |
|
structures, and they can be observed in the brighter contrast regions in the micrographs |
|
with the darker background being the.' |
|
- '# Chapter 4: Methodology |
|
|
|
|
|
Due to a relatively coarse grain size of the as-received cast material (compared |
|
to the starting extruded material), a larger area needed to be scanned for the |
|
cast-deformed samples, to be able to capture enough grains to produce statistically |
|
relevant results. Details of the area scanned for cast and extruded-deformed samples |
|
at various strain levels are provided in Figure 4-16. Note that for the sample |
|
deformed to the strain of 1.0, the scan area was limited to 0.75 x 0.6 mm even |
|
for the cast material, in order to avoid an overlap of the scan area with the |
|
dead metal zone (refer to Figure 4-16 (c) and (d)). As the results in later text |
|
show, hot deformation resulted in a considerable grain refinement in both cast |
|
and extruded materials, and this relatively small scan area was still found to |
|
adequately capture statistically relevant amount of data. |
|
|
|
|
|
|(a)|8 = 0.15|(b)|8 = 0.4|(c)|8 = 1.0| |
|
|
|
|---|---|---|---|---|---| |
|
|
|
|8|3 0.75 mm|!|8 0.75 mm|0.75 mm| | |
|
|
|
| |LQ|LO| | | | |
|
|
|
| | |(d)|17.5 mm|E| | |
|
|
|
|0 0.75 mm 2mm| | | | | | |
|
|
|
|Cast Az80 = 4008C, 0.1s-1 8 = 1.0| | | | | | |
|
|
|
|EBSD mapping area for cast mt| | | | | | |
|
|
|
|EBSD mapping area for extruded mt; Dead metal zone| | | | | | |
|
|
|
|
|
Figure 4-16 Location and size of EBSD scan for cast and extruded compressed samples. |
|
Sample size and the scan area are proportionally scaled. The dead metal zone location |
|
and extent is a simple schematic (and not to scale) and just for illustrative |
|
purposes. The true extent of the dead metal zone for one of the samples can be |
|
visualized from (d). |
|
|
|
|
|
The scanned data was processed and analysed using TSL OIM 8.0 software. The data |
|
was cleaned using standard data cleaning techniques [122]. This is illustrated |
|
through Figure 4-17 and Figure 4-18 for an extruded AZ80 sample deformed at 400 |
|
°C, 0.1 s-1 to a strain of 1.0. Initially, based on Image Quality (IQ) map and |
|
distribution, it was observed that areas with precipitates showed much lower IQ |
|
values. Therefore, the data pertaining to the precipitates could be filtered away |
|
from the data pertaining to the actual grains, using a lower threshold on the |
|
IQ values. For the case presented in Figure 4-17, an IQ value of 550 was found |
|
to be appropriate, though it should be noted.' |
|
- 'dominated by compressive loading) on the resulting texture of the as-forged material |
|
was previously explored by Grygu´c et al. [16, 17] in medium-scale I-beam style |
|
forgings. Those studies concluded that the favourable texture modification due |
|
to forging was the alignment of the c-axis of the hexagonal-closed-packed (HCP) |
|
crystal to be parallel with the local forging direction (or direction of dominant |
|
compressive loading), which is often normal to the die surface in relatively thin |
|
forgings at areas closer to the surface of the forged component. Depending on |
|
the process temperature, and local thermal history (accounting for the heat of |
|
deformation), the material recrystallization process may be highly non-uniform |
|
and heterogeneous in nature, resulting in spatially varying microstructure, texture |
|
and thus mechanical properties. |
|
|
|
|
|
# Figure 6 |
|
|
|
|
|
illustrates the relationship between the primary thermal process parameter (the |
|
target forging temperature) and the resulting yield strength for a variety of |
|
forged AZ80 Mg. Similar AZ80 Mg material conditions are presented, cast [3], extruded |
|
[17], cast-forged [3] and extruded forged [17], and they are once again grouped |
|
into categories of starting microstructure for clarity in illustrating the process-property |
|
relationship of forged AZ80 Mg. It can be observed that there is a general trend |
|
(albeit it, quite weak) where lower forging temperatures produce material with |
|
higher strength. This relatively poor correlation will subsequently be discussed |
|
in relation to the recrystallization mechanisms present during the forging process. |
|
For wrought forms of base material (i.e., extruded), forging at higher temperatures |
|
can result in grain growth (resulting in lower strength) and lower temperatures |
|
the potential for shear cracking (from reduced ductility) [30]. Of the studies |
|
presented, all were isothermal in nature (forging dies and billet are at the same |
|
temperature), and the forging temperature window ranges 250–450 °C. Beneath 250 |
|
°C, edge cracking and poor formability result as only a few deformation mechanisms |
|
become active, and large strain is not achievable [31]. At temperatures higher |
|
than 427 °C, incipient melting of the Mg17Al12 eutectic phase will occur near |
|
the grain boundary [13, 32]. This propensity towards incipient melting at higher |
|
processing temperatures for AZ series Mg alloys limits them from being extruded |
|
quickly, as compared to other structural metals, the processing window is quite |
|
narrow; however, this is not as limiting with die forging as the local deformation |
|
rates are typically smaller [33]. Chaudhury et al. suggest for AZ80 Mg, a non-isothermal |
|
forging process with a temperature window of 290–400 °C for the billet, and 205–290 |
|
°C for the forging dies to avoid these aforementioned types of defects that can |
|
result in the as-forged material [30] which agrees well with the trend seen in |
|
the current study. |
|
|
|
|
|
|Forging Temperature [°C]|Yield Strength| |
|
|
|
|---|---| |
|
|
|
|300|280| |
|
|
|
|260|240| |
|
|
|
|220|200| |
|
|
|
|180|160| |
|
|
|
|140|120| |
|
|
|
|100| | |
|
|
|
|
|
Figure 6. Relationship between primary thermal process parameter (forging temperature) |
|
and resulting yield strength of AZ80 Mg after forging (subdivided into various |
|
starting material conditions). The current study denotes properties extracted |
|
from various locations of the extruded then die-forged automotive suspension control |
|
arm forged at various temperatures (300 °C and 450 °C). For reference, extruded-forged |
|
data can found in the following studies [5–13,16], cast-forged from [3,16]. |
|
|
|
|
|
The stress-controlled fatigue response of the closed-die forged AZ80 Mg automotive |
|
suspension component is illustrated in Figure 7 for the two target forging temperatures |
|
that were investigated in this current study. It can be easily observed that for |
|
virtually all of the tests performed, for a given stress amplitude, the lower |
|
forging temperature (300 °C) pro-' |
|
- source_sentence: 'content=''1. What does MSE stand for in the context provided?\n2. |
|
Which technique is abbreviated as OES?'' additional_kwargs={''refusal'': None} |
|
response_metadata={''token_usage'': {''completion_tokens'': 24, ''prompt_tokens'': |
|
196, ''total_tokens'': 220, ''completion_tokens_details'': {''accepted_prediction_tokens'': |
|
0, ''audio_tokens'': 0, ''reasoning_tokens'': 0, ''rejected_prediction_tokens'': |
|
0}, ''prompt_tokens_details'': {''audio_tokens'': 0, ''cached_tokens'': 0}}, ''model_name'': |
|
''gpt-4o-mini-2024-07-18'', ''system_fingerprint'': ''fp_709714d124'', ''finish_reason'': |
|
''stop'', ''logprobs'': None} id=''run-798f8f65-4b52-4198-a267-dc6f61a192d0-0'' |
|
usage_metadata={''input_tokens'': 196, ''output_tokens'': 24, ''total_tokens'': |
|
220, ''input_token_details'': {''audio'': 0, ''cache_read'': 0}, ''output_token_details'': |
|
{''audio'': 0, ''reasoning'': 0}}' |
|
sentences: |
|
- '# References |
|
|
|
|
|
1. J.-H. Jun, "Dependence of hardness on interlamellar spacing of discontinuous |
|
precipitates in cast AZ91 magnesium alloy," Journal of Alloys and Compounds, vol. |
|
725, pp. 237-241, 2017. |
|
|
|
2. H. Jafari, M.H. Idris, A. Ourdjini, and G. Payganeh, "Effect of thermomechanical |
|
treatment on microstructure and hardness behavior of AZ63 magnesium alloy," Acta |
|
Metallurgica Sinica, vol. 22, pp. 401-407, 2009. |
|
|
|
3. Fuan Wei, Zhengang Zhang, Bo Shi, Chen Yang, and Jinhui Wang, "Effect of rolling |
|
deformation on microstructure and mechanical properties of Mg-6Sn-3Al-1Zn alloy," |
|
Materials Research Express, vol. 7, p. 026516, 2020. |
|
|
|
4. Ondřej Hilšer, Stanislav Rusz, Pavel Szkandera, Lubomír Čížek, Martin Kraus, |
|
Jan Džugan, and Wojciech Maziarz, "Study of the Microstructure, Tensile Properties |
|
and Hardness of AZ61 Magnesium Alloy Subjected to Severe Plastic Deformation," |
|
Metals, vol. 8, p. 776, 2018. |
|
|
|
5. Longqing Tang, Guowei Bo, Fulin Jiang, Shiwei Xu, Jie Teng, Dingfa Fu, Hui |
|
Zhang, "Unravelling the precipitation evolutions of AZ80 magnesium alloy during |
|
non-isothermal and isothermal processes," Journal of Materials Science & Technology, |
|
vol. 75, pp. 184-195, 2021. |
|
|
|
6. Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, Mark Chen, "Hierarchical |
|
text-conditional image generation with clip latents," arXiv preprint arXiv:2204.06125, |
|
p. https://doi.org/10.48550/arXiv.2204.06125, 2022. |
|
|
|
7. P. Acar, "Machine Learning Approach for Identification of Microstructure–Process |
|
Linkages," AIAA Journal, vol. 57, no. 8, pp. 3608-3614; https://doi.org/10.2514/1.J058244, |
|
2019. |
|
|
|
8. Yanming Liu, Shu Jian Chen, Kwesi Sagoe-Crentsil, Wenhui Duan, "Large set microstructure |
|
reconstruction mimicking quantum computing approach via deep learning," Acta Materialia, |
|
vol. 230, p. 117860; https://doi.org/10.1016/j.actamat.2022.117860, 2022. |
|
|
|
9. Yue Liu, Tianlu Zhao, Wangwei Ju, Siqi Shi, "Materials discovery and design |
|
using machine learning," Journal of Materiomics, vol. 3, no. 3, pp. 159-177; https://doi.org/10.1016/j.jmat.2017.08.002, |
|
2017. |
|
|
|
10. Brian L. DeCost, Toby Francis, Elizabeth A. Holm, "Exploring the microstructure |
|
manifold: Image texture representations applied to ultrahigh carbon steel microstructures," |
|
Acta Materialia, vol. 133, pp. 30-40; https://doi.org/10.1016/j.actamat.2017.05.014, |
|
2017. |
|
|
|
11. Daniel A. Cogswell and W. Craig Carter, "Thermodynamic phase-field model for |
|
microstructure with multiple components and phases: The possibility of metastable |
|
phases," Phys. Rev. E, vol. 83, no. 6, p. 061602; https://doi.org/10.1103/PhysRevE.83.061602, |
|
2011. |
|
|
|
12. S. Florez, M. Shakoor, T. Toulorge, M. Bernacki, "A new finite element strategy |
|
to simulate microstructural evolutions," Computational Materials Science, vol. |
|
172, p. 109335; https://doi.org/10.1016/j.commatsci.2019.109335, 2020. |
|
|
|
13. Damien Tourret, Hong Liu, Javier LLorca, "Phase-field modeling of microstructure |
|
evolution: Recent applications, perspectives and challenges," Progress in Materials |
|
Science, vol. 123, p. 100810; https://doi.org/10.1016/j.pmatsci.2021.100810, 2022.' |
|
- '# MSE – Mean Squared Error |
|
|
|
|
|
# NRMSE – Normalized Root Mean Squared Error |
|
|
|
|
|
# OES – Optical Emission Spectrometry |
|
|
|
|
|
# PSN – Particle Stimulated Nucleation |
|
|
|
|
|
# ReLU – Rectified linear unit |
|
|
|
|
|
# RT – Room Temperature |
|
|
|
|
|
# SDAS – Secondary Dendrite Arm Spacing |
|
|
|
|
|
# SEM – Scanning Electron Microscopy |
|
|
|
|
|
# SGD – Stochastic Gradient Descent |
|
|
|
|
|
# SVM – Support Vector Machine |
|
|
|
|
|
# TD – Transverse Direction |
|
|
|
|
|
# UTS – Ultimate Tensile Strength |
|
|
|
|
|
# VAE – Variational Autoencoder |
|
|
|
|
|
# XCT – X-ray Computed Tomography |
|
|
|
|
|
# YS – Yield Strength |
|
|
|
|
|
# xxi' |
|
- 'analytical or even empirical model that could relate a specific process parameter |
|
to a selected property. Also, compared to steel or aluminum alloys, there are |
|
still not enough experimental results that guide researchers and manufacturers |
|
in selecting proper processing routes and finding the optimum combination of process |
|
parameters. |
|
|
|
|
|
Machine learning techniques, as computational tools that use data gathered from |
|
experience to improve performance, have shown to be successful in establishing |
|
such relationships between interacting variables. Machine learning methods can |
|
be used to connect sparse findings on production of magnesium alloys to provide |
|
some insight into missing pieces of information. Such a link between different |
|
parameters not only help researchers to predict desired properties or select optimum |
|
parameters, but also can be used in theoretical studies of controlling mechanisms, |
|
which can finally give rise to more robust analytical or empirical models. |
|
|
|
|
|
In the current study, the feasibility of using machine learning approaches in |
|
establishing process-microstructure-property relationship would be put to the |
|
test. For this purpose, cast-forging process of AZ80 magnesium alloy, as a cost-effective |
|
hybrid manufacturing method, is going to be studied through advanced characterization |
|
methods. The effect of different process parameters of casting, forging, and intermediate |
|
thermo-mechanical processes and microstructural and mechanical properties of material |
|
in various stages of manufacturing are parameters of interest which will also |
|
be linked through machine learning models. |
|
|
|
|
|
# 1.2 Challenges and Opportunities |
|
|
|
|
|
To establish a link between process, microstructure, and property, a complete |
|
study of all the controlling parameters is required. Previous studies on the viability |
|
of production of magnesium-based structural components cover a variety of parameters |
|
and properties [3,4,7,8], but since these studies have focused on different aspects |
|
of production, a complete experimental study on the effect of parameters of casting |
|
and forging processes for AZ80 magnesium alloy is selected as a basis for the |
|
establishment of this relationship. It also contributes to broadening the knowledge |
|
of cast-forging manufacturing of magnesium alloys, where thorough experimental |
|
work is unavailable. |
|
|
|
|
|
Such an experimental study allows us to measure all properties in a controlled, |
|
consistent approach, while it also allows us to incorporate different processing |
|
parameters and, if necessary, append new variables based on the performance of |
|
machine learning models. |
|
|
|
|
|
As the experimental part progresses through different processing steps like wedge |
|
casting, cylinder casting, homogenization, I-beam forging, optimized preform casting, |
|
and forging the final structural components, data-driven models are developed |
|
to link process parameters to microstructural features.' |
|
pipeline_tag: sentence-similarity |
|
library_name: sentence-transformers |
|
metrics: |
|
- cosine_accuracy@1 |
|
- cosine_accuracy@3 |
|
- cosine_accuracy@5 |
|
- cosine_accuracy@10 |
|
- cosine_precision@1 |
|
- cosine_precision@3 |
|
- cosine_precision@5 |
|
- cosine_precision@10 |
|
- cosine_recall@1 |
|
- cosine_recall@3 |
|
- cosine_recall@5 |
|
- cosine_recall@10 |
|
- cosine_ndcg@10 |
|
- cosine_mrr@10 |
|
- cosine_map@100 |
|
model-index: |
|
- name: SentenceTransformer based on Snowflake/snowflake-arctic-embed-l |
|
results: |
|
- task: |
|
type: information-retrieval |
|
name: Information Retrieval |
|
dataset: |
|
name: Unknown |
|
type: unknown |
|
metrics: |
|
- type: cosine_accuracy@1 |
|
value: 0.9948979591836735 |
|
name: Cosine Accuracy@1 |
|
- type: cosine_accuracy@3 |
|
value: 1.0 |
|
name: Cosine Accuracy@3 |
|
- type: cosine_accuracy@5 |
|
value: 1.0 |
|
name: Cosine Accuracy@5 |
|
- type: cosine_accuracy@10 |
|
value: 1.0 |
|
name: Cosine Accuracy@10 |
|
- type: cosine_precision@1 |
|
value: 0.9948979591836735 |
|
name: Cosine Precision@1 |
|
- type: cosine_precision@3 |
|
value: 0.33333333333333337 |
|
name: Cosine Precision@3 |
|
- type: cosine_precision@5 |
|
value: 0.19999999999999998 |
|
name: Cosine Precision@5 |
|
- type: cosine_precision@10 |
|
value: 0.09999999999999999 |
|
name: Cosine Precision@10 |
|
- type: cosine_recall@1 |
|
value: 0.9948979591836735 |
|
name: Cosine Recall@1 |
|
- type: cosine_recall@3 |
|
value: 1.0 |
|
name: Cosine Recall@3 |
|
- type: cosine_recall@5 |
|
value: 1.0 |
|
name: Cosine Recall@5 |
|
- type: cosine_recall@10 |
|
value: 1.0 |
|
name: Cosine Recall@10 |
|
- type: cosine_ndcg@10 |
|
value: 0.9981169885386298 |
|
name: Cosine Ndcg@10 |
|
- type: cosine_mrr@10 |
|
value: 0.9974489795918368 |
|
name: Cosine Mrr@10 |
|
- type: cosine_map@100 |
|
value: 0.9974489795918368 |
|
name: Cosine Map@100 |
|
--- |
|
|
|
# SentenceTransformer based on Snowflake/snowflake-arctic-embed-l |
|
|
|
This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [Snowflake/snowflake-arctic-embed-l](https://huggingface.co/Snowflake/snowflake-arctic-embed-l). It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more. |
|
|
|
## Model Details |
|
|
|
### Model Description |
|
- **Model Type:** Sentence Transformer |
|
- **Base model:** [Snowflake/snowflake-arctic-embed-l](https://huggingface.co/Snowflake/snowflake-arctic-embed-l) <!-- at revision d8fb21ca8d905d2832ee8b96c894d3298964346b --> |
|
- **Maximum Sequence Length:** 512 tokens |
|
- **Output Dimensionality:** 1024 dimensions |
|
- **Similarity Function:** Cosine Similarity |
|
<!-- - **Training Dataset:** Unknown --> |
|
<!-- - **Language:** Unknown --> |
|
<!-- - **License:** Unknown --> |
|
|
|
### Model Sources |
|
|
|
- **Documentation:** [Sentence Transformers Documentation](https://sbert.net) |
|
- **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers) |
|
- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers) |
|
|
|
### Full Model Architecture |
|
|
|
``` |
|
SentenceTransformer( |
|
(0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel |
|
(1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True}) |
|
(2): Normalize() |
|
) |
|
``` |
|
|
|
## Usage |
|
|
|
### Direct Usage (Sentence Transformers) |
|
|
|
First install the Sentence Transformers library: |
|
|
|
```bash |
|
pip install -U sentence-transformers |
|
``` |
|
|
|
Then you can load this model and run inference. |
|
```python |
|
from sentence_transformers import SentenceTransformer |
|
|
|
# Download from the 🤗 Hub |
|
model = SentenceTransformer("tabesink92/mg_alloy-snowflake-arctic-embed-l-ft-v2") |
|
# Run inference |
|
sentences = [ |
|
"content='1. What does MSE stand for in the context provided?\\n2. Which technique is abbreviated as OES?' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 24, 'prompt_tokens': 196, 'total_tokens': 220, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-mini-2024-07-18', 'system_fingerprint': 'fp_709714d124', 'finish_reason': 'stop', 'logprobs': None} id='run-798f8f65-4b52-4198-a267-dc6f61a192d0-0' usage_metadata={'input_tokens': 196, 'output_tokens': 24, 'total_tokens': 220, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}", |
|
'# MSE – Mean Squared Error\n\n# NRMSE – Normalized Root Mean Squared Error\n\n# OES – Optical Emission Spectrometry\n\n# PSN – Particle Stimulated Nucleation\n\n# ReLU – Rectified linear unit\n\n# RT – Room Temperature\n\n# SDAS – Secondary Dendrite Arm Spacing\n\n# SEM – Scanning Electron Microscopy\n\n# SGD – Stochastic Gradient Descent\n\n# SVM – Support Vector Machine\n\n# TD – Transverse Direction\n\n# UTS – Ultimate Tensile Strength\n\n# VAE – Variational Autoencoder\n\n# XCT – X-ray Computed Tomography\n\n# YS – Yield Strength\n\n# xxi', |
|
'analytical or even empirical model that could relate a specific process parameter to a selected property. Also, compared to steel or aluminum alloys, there are still not enough experimental results that guide researchers and manufacturers in selecting proper processing routes and finding the optimum combination of process parameters.\n\nMachine learning techniques, as computational tools that use data gathered from experience to improve performance, have shown to be successful in establishing such relationships between interacting variables. Machine learning methods can be used to connect sparse findings on production of magnesium alloys to provide some insight into missing pieces of information. Such a link between different parameters not only help researchers to predict desired properties or select optimum parameters, but also can be used in theoretical studies of controlling mechanisms, which can finally give rise to more robust analytical or empirical models.\n\nIn the current study, the feasibility of using machine learning approaches in establishing process-microstructure-property relationship would be put to the test. For this purpose, cast-forging process of AZ80 magnesium alloy, as a cost-effective hybrid manufacturing method, is going to be studied through advanced characterization methods. The effect of different process parameters of casting, forging, and intermediate thermo-mechanical processes and microstructural and mechanical properties of material in various stages of manufacturing are parameters of interest which will also be linked through machine learning models.\n\n# 1.2 Challenges and Opportunities\n\nTo establish a link between process, microstructure, and property, a complete study of all the controlling parameters is required. Previous studies on the viability of production of magnesium-based structural components cover a variety of parameters and properties [3,4,7,8], but since these studies have focused on different aspects of production, a complete experimental study on the effect of parameters of casting and forging processes for AZ80 magnesium alloy is selected as a basis for the establishment of this relationship. It also contributes to broadening the knowledge of cast-forging manufacturing of magnesium alloys, where thorough experimental work is unavailable.\n\nSuch an experimental study allows us to measure all properties in a controlled, consistent approach, while it also allows us to incorporate different processing parameters and, if necessary, append new variables based on the performance of machine learning models.\n\nAs the experimental part progresses through different processing steps like wedge casting, cylinder casting, homogenization, I-beam forging, optimized preform casting, and forging the final structural components, data-driven models are developed to link process parameters to microstructural features.', |
|
] |
|
embeddings = model.encode(sentences) |
|
print(embeddings.shape) |
|
# [3, 1024] |
|
|
|
# Get the similarity scores for the embeddings |
|
similarities = model.similarity(embeddings, embeddings) |
|
print(similarities.shape) |
|
# [3, 3] |
|
``` |
|
|
|
<!-- |
|
### Direct Usage (Transformers) |
|
|
|
<details><summary>Click to see the direct usage in Transformers</summary> |
|
|
|
</details> |
|
--> |
|
|
|
<!-- |
|
### Downstream Usage (Sentence Transformers) |
|
|
|
You can finetune this model on your own dataset. |
|
|
|
<details><summary>Click to expand</summary> |
|
|
|
</details> |
|
--> |
|
|
|
<!-- |
|
### Out-of-Scope Use |
|
|
|
*List how the model may foreseeably be misused and address what users ought not to do with the model.* |
|
--> |
|
|
|
## Evaluation |
|
|
|
### Metrics |
|
|
|
#### Information Retrieval |
|
|
|
* Evaluated with [<code>InformationRetrievalEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) |
|
|
|
| Metric | Value | |
|
|:--------------------|:-----------| |
|
| cosine_accuracy@1 | 0.9949 | |
|
| cosine_accuracy@3 | 1.0 | |
|
| cosine_accuracy@5 | 1.0 | |
|
| cosine_accuracy@10 | 1.0 | |
|
| cosine_precision@1 | 0.9949 | |
|
| cosine_precision@3 | 0.3333 | |
|
| cosine_precision@5 | 0.2 | |
|
| cosine_precision@10 | 0.1 | |
|
| cosine_recall@1 | 0.9949 | |
|
| cosine_recall@3 | 1.0 | |
|
| cosine_recall@5 | 1.0 | |
|
| cosine_recall@10 | 1.0 | |
|
| **cosine_ndcg@10** | **0.9981** | |
|
| cosine_mrr@10 | 0.9974 | |
|
| cosine_map@100 | 0.9974 | |
|
|
|
<!-- |
|
## Bias, Risks and Limitations |
|
|
|
*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* |
|
--> |
|
|
|
<!-- |
|
### Recommendations |
|
|
|
*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* |
|
--> |
|
|
|
## Training Details |
|
|
|
### Training Dataset |
|
|
|
#### Unnamed Dataset |
|
|
|
* Size: 892 training samples |
|
* Columns: <code>sentence_0</code> and <code>sentence_1</code> |
|
* Approximate statistics based on the first 892 samples: |
|
| | sentence_0 | sentence_1 | |
|
|:--------|:--------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------| |
|
| type | string | string | |
|
| details | <ul><li>min: 340 tokens</li><li>mean: 371.01 tokens</li><li>max: 410 tokens</li></ul> | <ul><li>min: 29 tokens</li><li>mean: 439.26 tokens</li><li>max: 512 tokens</li></ul> | |
|
* Samples: |
|
| sentence_0 | sentence_1 | |
|
|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| |
|
| <code>content='1. What experimental method was used to investigate the fatigue properties of extruded AZ80 magnesium alloys in the study? \n2. How does the crack initiation mechanism change with varying stress amplitudes in the AZ80 magnesium alloys?' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 46, 'prompt_tokens': 731, 'total_tokens': 777, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-mini-2024-07-18', 'system_fingerprint': 'fp_7fcd609668', 'finish_reason': 'stop', 'logprobs': None} id='run-d649e508-ccc4-420e-b812-67719c92f2cb-0' usage_metadata={'input_tokens': 731, 'output_tokens': 46, 'total_tokens': 777, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}</code> | <code># Fatigue behaviour and fractography of extruded AZ80 magnesium alloys in very high cycle regime<br><br># Kazuaki Shiozawaa *, Tomoki Kashiwagi b, Tutomu Murai c, Tooru Takahashi c<br><br>b Tohoku Electric Power Co.Inc., Sendai980-8550, Japan<br><br>c Sankyo-Tateyama Aluminum Industry Co. Ltd., Imizu934-8577, Japan<br><br>Received 26 February 2010; revised 11 March 2010; accepted 15 March 2010<br><br># Abstract<br><br>In order to investigate the fatigue properties of extruded magnesium alloy in very high-cycle regime, rotary bending fatigue test was performed in ambient atmosphere at room temperature using the hourglass shaped specimens of AZ80 alloys extruded (F-specimen) and treated by an artificial aging after extrusion (T5-specimen). From the experimental results, both specimens show a clear step-wise S-N curve on which two knees exists. Specific stress amplitude formed the knee corresponded to the 0.2% offset proof stress of 160MPa in compression. From the detailed observation of fracture surface, small facet-like s...</code> | |
|
| <code>content='1. What experimental method was used to investigate the fatigue properties of extruded AZ80 magnesium alloys in the study? \n2. How does the crack initiation mechanism change with varying stress amplitudes in the AZ80 magnesium alloys?' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 46, 'prompt_tokens': 731, 'total_tokens': 777, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-mini-2024-07-18', 'system_fingerprint': 'fp_7fcd609668', 'finish_reason': 'stop', 'logprobs': None} id='run-5e3c339c-30e6-4dc7-bacb-24b9143312ad-0' usage_metadata={'input_tokens': 731, 'output_tokens': 46, 'total_tokens': 777, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}</code> | <code># Fatigue behaviour and fractography of extruded AZ80 magnesium alloys in very high cycle regime<br><br># Kazuaki Shiozawaa *, Tomoki Kashiwagi b, Tutomu Murai c, Tooru Takahashi c<br><br>b Tohoku Electric Power Co.Inc., Sendai980-8550, Japan<br><br>c Sankyo-Tateyama Aluminum Industry Co. Ltd., Imizu934-8577, Japan<br><br>Received 26 February 2010; revised 11 March 2010; accepted 15 March 2010<br><br># Abstract<br><br>In order to investigate the fatigue properties of extruded magnesium alloy in very high-cycle regime, rotary bending fatigue test was performed in ambient atmosphere at room temperature using the hourglass shaped specimens of AZ80 alloys extruded (F-specimen) and treated by an artificial aging after extrusion (T5-specimen). From the experimental results, both specimens show a clear step-wise S-N curve on which two knees exists. Specific stress amplitude formed the knee corresponded to the 0.2% offset proof stress of 160MPa in compression. From the detailed observation of fracture surface, small facet-like s...</code> | |
|
| <code>content='1. What is the chemical composition of the AZ80 magnesium alloy used in the study?\n2. How was the T5 treatment applied to the specimens, and what was its purpose in the experiment?' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 41, 'prompt_tokens': 617, 'total_tokens': 658, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-mini-2024-07-18', 'system_fingerprint': 'fp_7fcd609668', 'finish_reason': 'stop', 'logprobs': None} id='run-1ea8b034-255f-4994-b459-394267605838-0' usage_metadata={'input_tokens': 617, 'output_tokens': 41, 'total_tokens': 658, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}</code> | <code># 2. Experimental Procedures<br><br># 2.1. Testing materials and specimen<br><br>The material used in this study was commercial Mg-Al-Zn magnesium alloy, AZ80. The chemical composition of these materials (in mass percentage) is 8.24Al, 0.67Zn, 0.20Mn, 0.005Fe, 0.012Si, 0.0008Cu, 0.0007Ni and balanced Mg. The bar with 16mm in diameter was extruded from a billet of 160mm diameter (an extrusion ratio of 99.4:1) under the extrusion ram speed of 1.5m/min at temperature of 623K. Hour-glass shaped specimens with a grip diameter of 10 mm and minimum diameter of 5 mm (Fig. 1) was machined from the extruded bar with the loading axis parallel to their extrusion directions. The elastic stress concentration factor, Kt, of these specimens was 1.065. On the other hand, to investigate the effect of an aging treatment on fatigue behavior, specimens were prepared from the extruded bar which was heated at 473K for 32h and then air-cooled (T5 treatment). From now on, the specimen extruded and no heat-treated is refer...</code> | |
|
* Loss: [<code>MatryoshkaLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#matryoshkaloss) with these parameters: |
|
```json |
|
{ |
|
"loss": "MultipleNegativesRankingLoss", |
|
"matryoshka_dims": [ |
|
768, |
|
512, |
|
256, |
|
128, |
|
64 |
|
], |
|
"matryoshka_weights": [ |
|
1, |
|
1, |
|
1, |
|
1, |
|
1 |
|
], |
|
"n_dims_per_step": -1 |
|
} |
|
``` |
|
|
|
### Training Hyperparameters |
|
#### Non-Default Hyperparameters |
|
|
|
- `eval_strategy`: steps |
|
- `per_device_train_batch_size`: 16 |
|
- `per_device_eval_batch_size`: 16 |
|
- `num_train_epochs`: 25 |
|
- `multi_dataset_batch_sampler`: round_robin |
|
|
|
#### All Hyperparameters |
|
<details><summary>Click to expand</summary> |
|
|
|
- `overwrite_output_dir`: False |
|
- `do_predict`: False |
|
- `eval_strategy`: steps |
|
- `prediction_loss_only`: True |
|
- `per_device_train_batch_size`: 16 |
|
- `per_device_eval_batch_size`: 16 |
|
- `per_gpu_train_batch_size`: None |
|
- `per_gpu_eval_batch_size`: None |
|
- `gradient_accumulation_steps`: 1 |
|
- `eval_accumulation_steps`: None |
|
- `torch_empty_cache_steps`: None |
|
- `learning_rate`: 5e-05 |
|
- `weight_decay`: 0.0 |
|
- `adam_beta1`: 0.9 |
|
- `adam_beta2`: 0.999 |
|
- `adam_epsilon`: 1e-08 |
|
- `max_grad_norm`: 1 |
|
- `num_train_epochs`: 25 |
|
- `max_steps`: -1 |
|
- `lr_scheduler_type`: linear |
|
- `lr_scheduler_kwargs`: {} |
|
- `warmup_ratio`: 0.0 |
|
- `warmup_steps`: 0 |
|
- `log_level`: passive |
|
- `log_level_replica`: warning |
|
- `log_on_each_node`: True |
|
- `logging_nan_inf_filter`: True |
|
- `save_safetensors`: True |
|
- `save_on_each_node`: False |
|
- `save_only_model`: False |
|
- `restore_callback_states_from_checkpoint`: False |
|
- `no_cuda`: False |
|
- `use_cpu`: False |
|
- `use_mps_device`: False |
|
- `seed`: 42 |
|
- `data_seed`: None |
|
- `jit_mode_eval`: False |
|
- `use_ipex`: False |
|
- `bf16`: False |
|
- `fp16`: False |
|
- `fp16_opt_level`: O1 |
|
- `half_precision_backend`: auto |
|
- `bf16_full_eval`: False |
|
- `fp16_full_eval`: False |
|
- `tf32`: None |
|
- `local_rank`: 0 |
|
- `ddp_backend`: None |
|
- `tpu_num_cores`: None |
|
- `tpu_metrics_debug`: False |
|
- `debug`: [] |
|
- `dataloader_drop_last`: False |
|
- `dataloader_num_workers`: 0 |
|
- `dataloader_prefetch_factor`: None |
|
- `past_index`: -1 |
|
- `disable_tqdm`: False |
|
- `remove_unused_columns`: True |
|
- `label_names`: None |
|
- `load_best_model_at_end`: False |
|
- `ignore_data_skip`: False |
|
- `fsdp`: [] |
|
- `fsdp_min_num_params`: 0 |
|
- `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False} |
|
- `fsdp_transformer_layer_cls_to_wrap`: None |
|
- `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None} |
|
- `deepspeed`: None |
|
- `label_smoothing_factor`: 0.0 |
|
- `optim`: adamw_torch |
|
- `optim_args`: None |
|
- `adafactor`: False |
|
- `group_by_length`: False |
|
- `length_column_name`: length |
|
- `ddp_find_unused_parameters`: None |
|
- `ddp_bucket_cap_mb`: None |
|
- `ddp_broadcast_buffers`: False |
|
- `dataloader_pin_memory`: True |
|
- `dataloader_persistent_workers`: False |
|
- `skip_memory_metrics`: True |
|
- `use_legacy_prediction_loop`: False |
|
- `push_to_hub`: False |
|
- `resume_from_checkpoint`: None |
|
- `hub_model_id`: None |
|
- `hub_strategy`: every_save |
|
- `hub_private_repo`: None |
|
- `hub_always_push`: False |
|
- `gradient_checkpointing`: False |
|
- `gradient_checkpointing_kwargs`: None |
|
- `include_inputs_for_metrics`: False |
|
- `include_for_metrics`: [] |
|
- `eval_do_concat_batches`: True |
|
- `fp16_backend`: auto |
|
- `push_to_hub_model_id`: None |
|
- `push_to_hub_organization`: None |
|
- `mp_parameters`: |
|
- `auto_find_batch_size`: False |
|
- `full_determinism`: False |
|
- `torchdynamo`: None |
|
- `ray_scope`: last |
|
- `ddp_timeout`: 1800 |
|
- `torch_compile`: False |
|
- `torch_compile_backend`: None |
|
- `torch_compile_mode`: None |
|
- `dispatch_batches`: None |
|
- `split_batches`: None |
|
- `include_tokens_per_second`: False |
|
- `include_num_input_tokens_seen`: False |
|
- `neftune_noise_alpha`: None |
|
- `optim_target_modules`: None |
|
- `batch_eval_metrics`: False |
|
- `eval_on_start`: False |
|
- `use_liger_kernel`: False |
|
- `eval_use_gather_object`: False |
|
- `average_tokens_across_devices`: False |
|
- `prompts`: None |
|
- `batch_sampler`: batch_sampler |
|
- `multi_dataset_batch_sampler`: round_robin |
|
|
|
</details> |
|
|
|
### Training Logs |
|
| Epoch | Step | Training Loss | cosine_ndcg@10 | |
|
|:-------:|:----:|:-------------:|:--------------:| |
|
| 0.8929 | 50 | - | 0.9525 | |
|
| 1.0 | 56 | - | 0.9638 | |
|
| 1.7857 | 100 | - | 0.9801 | |
|
| 2.0 | 112 | - | 0.9743 | |
|
| 2.6786 | 150 | - | 0.9849 | |
|
| 3.0 | 168 | - | 0.9812 | |
|
| 3.5714 | 200 | - | 0.9868 | |
|
| 4.0 | 224 | - | 0.9962 | |
|
| 4.4643 | 250 | - | 0.9925 | |
|
| 5.0 | 280 | - | 0.9944 | |
|
| 5.3571 | 300 | - | 0.9944 | |
|
| 6.0 | 336 | - | 0.9944 | |
|
| 6.25 | 350 | - | 0.9944 | |
|
| 7.0 | 392 | - | 0.9981 | |
|
| 7.1429 | 400 | - | 0.9962 | |
|
| 8.0 | 448 | - | 0.9981 | |
|
| 8.0357 | 450 | - | 0.9981 | |
|
| 8.9286 | 500 | 0.3646 | 0.9981 | |
|
| 9.0 | 504 | - | 0.9981 | |
|
| 9.8214 | 550 | - | 0.9981 | |
|
| 10.0 | 560 | - | 0.9981 | |
|
| 10.7143 | 600 | - | 0.9981 | |
|
| 11.0 | 616 | - | 0.9981 | |
|
| 11.6071 | 650 | - | 0.9981 | |
|
| 12.0 | 672 | - | 0.9944 | |
|
| 12.5 | 700 | - | 0.9981 | |
|
| 13.0 | 728 | - | 0.9981 | |
|
| 13.3929 | 750 | - | 0.9981 | |
|
| 14.0 | 784 | - | 0.9981 | |
|
| 14.2857 | 800 | - | 0.9981 | |
|
| 15.0 | 840 | - | 0.9981 | |
|
| 15.1786 | 850 | - | 0.9981 | |
|
| 16.0 | 896 | - | 0.9981 | |
|
| 16.0714 | 900 | - | 0.9981 | |
|
| 16.9643 | 950 | - | 0.9981 | |
|
| 17.0 | 952 | - | 0.9981 | |
|
| 17.8571 | 1000 | 0.064 | 0.9981 | |
|
| 18.0 | 1008 | - | 0.9981 | |
|
| 18.75 | 1050 | - | 0.9981 | |
|
| 19.0 | 1064 | - | 0.9981 | |
|
| 19.6429 | 1100 | - | 0.9981 | |
|
| 20.0 | 1120 | - | 0.9981 | |
|
| 20.5357 | 1150 | - | 0.9981 | |
|
| 21.0 | 1176 | - | 0.9981 | |
|
| 21.4286 | 1200 | - | 0.9981 | |
|
| 22.0 | 1232 | - | 0.9981 | |
|
| 22.3214 | 1250 | - | 0.9981 | |
|
| 23.0 | 1288 | - | 0.9981 | |
|
| 23.2143 | 1300 | - | 0.9981 | |
|
| 24.0 | 1344 | - | 0.9981 | |
|
| 24.1071 | 1350 | - | 0.9981 | |
|
| 25.0 | 1400 | - | 0.9981 | |
|
|
|
|
|
### Framework Versions |
|
- Python: 3.11.11 |
|
- Sentence Transformers: 3.4.1 |
|
- Transformers: 4.48.3 |
|
- PyTorch: 2.5.1+cu124 |
|
- Accelerate: 1.3.0 |
|
- Datasets: 3.3.2 |
|
- Tokenizers: 0.21.0 |
|
|
|
## Citation |
|
|
|
### BibTeX |
|
|
|
#### Sentence Transformers |
|
```bibtex |
|
@inproceedings{reimers-2019-sentence-bert, |
|
title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks", |
|
author = "Reimers, Nils and Gurevych, Iryna", |
|
booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing", |
|
month = "11", |
|
year = "2019", |
|
publisher = "Association for Computational Linguistics", |
|
url = "https://arxiv.org/abs/1908.10084", |
|
} |
|
``` |
|
|
|
#### MatryoshkaLoss |
|
```bibtex |
|
@misc{kusupati2024matryoshka, |
|
title={Matryoshka Representation Learning}, |
|
author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi}, |
|
year={2024}, |
|
eprint={2205.13147}, |
|
archivePrefix={arXiv}, |
|
primaryClass={cs.LG} |
|
} |
|
``` |
|
|
|
#### MultipleNegativesRankingLoss |
|
```bibtex |
|
@misc{henderson2017efficient, |
|
title={Efficient Natural Language Response Suggestion for Smart Reply}, |
|
author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil}, |
|
year={2017}, |
|
eprint={1705.00652}, |
|
archivePrefix={arXiv}, |
|
primaryClass={cs.CL} |
|
} |
|
``` |
|
|
|
<!-- |
|
## Glossary |
|
|
|
*Clearly define terms in order to be accessible across audiences.* |
|
--> |
|
|
|
<!-- |
|
## Model Card Authors |
|
|
|
*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.* |
|
--> |
|
|
|
<!-- |
|
## Model Card Contact |
|
|
|
*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.* |
|
--> |
